**Mice Microbiota Composition Changes by Inulin Feeding with a Long Fasting Period under a Two-Meals-Per-Day Schedule**

**Hiroyuki Sasaki 1,2,**†**, Hiroki Miyakawa 1,**†**, Aya Watanabe 1, Yuki Nakayama 1, Yijin Lyu 1, Koki Hama <sup>1</sup> and Shigenobu Shibata 1,\***


Received: 23 October 2019; Accepted: 14 November 2019; Published: 16 November 2019

**Abstract:** Water-soluble dietary fiber is known to modulate fecal microbiota. Although there are a few reports investigating the effects of fiber intake timing on metabolism, there are none on the effect of intake timing on microbiota. Therefore, in this study, we examined the timing effects of inulin-containing food on fecal microbiota. Mice were housed under conditions with a two-meals-per-day schedule, with a long fasting period in the morning and a short fasting period in the evening. Then, 10–14 days after inulin intake, cecal content and feces were collected, and cecal pH and short-chain fatty acids (SCFAs) were measured. The microbiome was determined using 16S rDNA sequencing. Inulin feeding in the morning rather than the evening decreased the cecal pH, increased SCFAs, and changed the microbiome composition. These data suggest that inulin is more easily digested by fecal microbiota during the active period than the inactive period. Furthermore, to confirm the effect of fasting length, mice were housed under a one-meal-per-day schedule. When the duration of fasting was equal, the difference between morning and evening nearly disappeared. Thus, our study demonstrates that consuming inulin at breakfast, which is generally after a longer fasting period, has a greater effect on the microbiota.

**Keywords:** microbiota; inulin; circadian rhythm; feeding timing

#### **1. Introduction**

In the gut of mammals, the microbiota includes 100 trillion bacteria. Disordered microbiota alteration is involved in the development of various diseases [1]. *Firmicutes* are bacteria related to obesity, while *Bacteroidetes* suppress fat accumulation in mice fed a high-fat diet (HFD) [2]. When the feces of obese mice are transplanted into germ-free mice, obesity develops [3]. Moreover, *Fusobacterium*, including *Fusobacterium nucleatum*, are increased in patients with colorectal cancer compared with healthy subjects [4,5]. In addition to physical illnesses, a relationship of microbiota with psychological illness has also been reported. In patients with major depression, *Bacteroidetes*, *Proteobacteria*, and *Actinobacteria* are significantly increased compared with healthy subjects [6]. These results suggest that intestinal bacteria are related to the development of diseases and that maintaining homeostasis of the microbiota is important for the mental and physical health of the host.

Short-chain fatty acids (SCFAs) are produced when the microbiota ferments and degrades non-digestible food components [7]. The SCFAs lower intestinal pH, suppress the growth of pathogenic

bacteria in the gut, and function as a regulator of metabolism and immunity [8]. Among SCFAs, acetic acid is a liver energy substrate used for fat synthesis, and propionic acid is used as a material for gluconeogenesis in the liver. Butyrate promotes the induction of regulatory T cells in the large intestine [9,10]. Furthermore, SCFAs also increase insulin sensitivity in the liver and muscles through GPR43, a receptor for SCFAs in white fat, as well as increase energy efficiency [11].

The circadian rhythm, controlled by clock genes, plays an important role in daily locomotor activity rhythms and physiological events, such as the sleep–wake cycle, hormone secretion, and the sympathetic nervous system [12,13]. Clocks in peripheral tissues are regulated by the central clock in the suprachiasmatic nucleus and external cues such as food, temperature, and exercise [14–16]. It has been reported that circadian rhythms are also present in the intestinal flora and controlled by dietary composition [17–20]. Furthermore, disturbance of the circadian clock due to jet lag alters microbial populations. For example, when the stool of jet-lagged mice is transplanted into germ-free mice, the recipient mice become obese [19]. Recently, however, it has been reported that SCFAs produced by gut microbiota can synchronize the circadian clock [21].

The microbiota composition changes depending on food components. In particular, foods rich in dietary fiber have a strong effect on the microbiota and are known as prebiotics [22]. According to Gibson et al., prebiotics are defined as "nondigestible food ingredients that beneficially affect the host by selectively stimulating the growth and/or activity of one or a limited number of bacterial species already resident in the colon, and thus attempt to improve host health" [23]. Inulin is a water-soluble dietary fiber and, thus, a prebiotic. It is particularly involved in the growth of bacteria that produce lactic acid [24] and promote the absorption of minerals such as calcium and magnesium [25,26].

It has been suggested that meal timing and daily eating habits may affect the development and prevention of lifestyle-related diseases such as obesity. A study by Hatori et al. demonstrates that restricted feeding in an activity period for mice without reducing calorific intake prevents metabolic diseases in mice fed a HFD [27]. Mice consuming milk fat late in the activity period have elevated hepatic fat and increased serum triglycerides and free fatty acids [28]. In addition, scheduled access to a HFD during the inactivity period increases body weight in mice compared with access during the activity period [29,30]. Moreover, in human experiments, the combination of a late dinner with a short sleep duration is associated with the risk of obesity [31]. In addition, the risk of obesity has been related to eating supper after 20:00 in the evening [32]. In recent years, it has been suggested that the influence of food on lipid metabolism is different depending on the time of food intake. In mice fed a high-fructose diet, fish oil given earlier in the activity period rather than later more effectively lowered lipids [33].

There are many reports indicating that time of food intake affects energy metabolism, but there are still relatively few reports describing the effect of eating time on microbiota. Furthermore, there are few reports on the dual effect of food type, particularly dietary fiber, and intake time on microbiota. Therefore, in the current study, we investigated whether inulin intake during the morning has a stronger effect on the microbiota than inulin intake during the evening with a two-meals-per-day schedule in mice.

#### **2. Materials and Methods**

#### *2.1. Mice*

In this study, we used eight-week-old male ICR mice (Tokyo Laboratory Animals, Tokyo, Japan). The mice were kept in a room maintained on a 12 h light/12 h dark (LD) cycle (lights on from 08:00 to 20:00). Zeitgeber time 0 (ZT0) was designated as lights-on time and ZT12 as lights-off time under the LD cycle. The mice were housed either in groups (five mice per cage; experiments 1 and 2) or individually (experiments 3 and 4) in plastic cages. The cages were maintained at a temperature of 22 ± 2 ◦C, humidity of 60 ± 5%, and light intensity of 100–150 lux. The mice were provided with a HFD containing 45% kcal of fat (Diet 12451; Research Diets Inc., New Brunswick, NJ, USA) with cellulose

(Oriental Yeast Co., Ltd., Tokyo, Japan) or inulin (Fuji FF; Fuji Nihon Seito Co., Tokyo, Japan) [34,35] and water ad libitum. This HFD is a diet used as a model for obesity, diabetes, and fatty liver in rodents [36,37]. Inulin has been reported to attenuate HFD-induced lipid metabolism and microbiota change [38]. In addition, the metabolic syndrome caused by obesity and abnormal lipid metabolism in the liver are related to microbiota change [39–41]. Therefore, we used an HFD to enhance the attenuating effects of inulin with the condition of microbiota change. The animal experiment was conducted with permission from the Committee for Animal Experimentation of the School of Science and Engineering at Waseda University (permission # 09A11, 10A11) and in accordance with the law (No. 105) and notification (No. 6) of the Japanese government.

#### *2.2. Scheduled Feeding*

We prepared two types of feeding conditions. In type 1 (experiments 1 and 2), only the feeding time was controlled, while in type 2 (experiments 3 and 4), both the start time of feeding and the amount of food were controlled.

In type 1 feeding, all of the mice could approach the feed box during the permitted time. We defined the morning as ZT12–20 and the evening as ZT20–4. The mice had free access to the feed box for predetermined four-hour periods (morning meal as ZT12–16 and evening meal as ZT20–0). Throughout the remaining time, the feed box was locked. Food intake was calculated by measuring the weight of the food in the feed box at the start and end of the experiment. The total consumed food was divided by the number of mice and the number of days in the experiment. In the type 1 experiments, we housed the mice as a group to avoid the stress induced by individual housing.

In type 2 feeding, all of the mice were housed in cages containing food dispensers that released food pellets under the regulation of a timer. The mice were fed 90% of the amount of food that was consumed in experiment 1 (Figure 1c). In experiment 3, the mice were fed two meals per day at ZT12 (defined as morning) or ZT20 (defined as evening); the meal size was 1.8 g. In experiment 4, the mice were fed one meal per day at ZT12 (morning) or ZT20 (evening); the meal size was 3.6 g.

We adjusted the concentration of dietary fiber so that the amount of inulin was approximately equal between experiments.

#### *2.3. Cecal pH Measurement*

The cecal pH was measured by inserting the glass tip of an electrode of a pH meter (pH Spear; Eutech Instruments, Vernon Hills, IL, USA) directly into the cecum.

#### *2.4. Measurement of SCFAs*

The SCFAs were measured via gas chromatography and flame ionization detection (Shimadzu Corp., Kyoto, Japan) as described by a previous report [42] with some modifications. A total of 0.05 g of cecal content was acidified with 0.05 mL sulfuric acid (FUJIFILM Wako Pure Chemical Corp., Osaka, Japan). Then, the SCFAs were extracted by shaking with 0.4 mL of diethyl ether (FUJIFILM Wako Pure Chemical Corp., Osaka, Japan) and 0.2 mL of ethanol (FUJIFILM Wako Pure Chemical Corp., Osaka, Japan), which was then centrifuged at 14,000 rpm at room temperature for 30 s. A total of 1 μL of the organic phase was injected into the capillary column (InertCap Pure WAX (30 m × 0.25 mm, df = 0.5 μm); GL Science Inc., Tokyo, Japan). The initial temperature was 80 ◦C, and the final temperature was 200 ◦C. Helium was used as a carrier gas. Quantification of the samples was performed using calibration curves for acetic, lactic, propionic, and butyric acids. A standard curve for each acid was conducted for their quantitation in the samples.

#### *2.5. Fecal DNA Extraction*

The fecal DNA was extracted as previously described, with some modifications [43]. About 0.2 g of the fecal sample was suspended in a 50 mL Falcon tube containing 20 mL PBS. The suspension was filtered through a 100-μm mesh nylon filter (Corning Inc., New York NY, USA). The debris on the filter was washed with 10 mL of Phosphate buffered salts (PBS). The filtrates were centrifuged at 4000 rpm for 20 min at 4 ◦C, and each precipitate was then suspended with 1.5 mL of TE10 buffer (10 mM Tris-HCl (FUJIFILM Wako Pure Chemical Corp., Osaka, Japan)/10 mM ethylenediaminetetraacetic acid (EDTA) (DOJINDO, Tokyo, Japan)). The suspensions were transferred to 2-mL microtubes before being centrifuged at 10,000 rpm for five minutes at 4 ◦C. Following this, each precipitate was suspended again with 0.8 mL of TE10 buffer. The DNA was extracted using 1 mL of PCI (Invitrogen, Carlsbad, CA, USA) and isolated with 0.1 mL of lysozyme (FUJIFILM Wako Pure Chemical Corp., Osaka, Japan) and 0.02 mL of achromopeptidase (FUJIFILM Wako Pure Chemical Corp., Osaka, Japan). The DNA was purified via treatment with RNase (Promega Corp., Madison, WI, USA), followed by precipitation with 20% PEG solution (Tokyo Chemical Industry Co., Ltd., Tokyo, Japan). Finally, the DNA was rinsed with 70% ethanol and dissolved in 50 μL TE buffer.

#### *2.6. 16 S rDNA Gene Sequencing*

The 16S rDNA gene sequencing was performed according to the instructions of Illumina. V3–V4 variable regions of the 16S rDNA gene were amplified by PCR using the following primers:

#### forward primer = 5 -TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTA

CGGGNGGCWGCAG-3 ;

#### reverse primer = 5 -GTCTCGTGGGCTCGGAGATGTGTATAAGA

#### GACAGGACTACHVGGGTATCTAATCC-3 .

The PCR amplification was performed with 2.5 μL microbial DNA (5 ng/μL), 5 μL of each primer (1 μmol/L), and 12.5 μL 2 × KAPA HiFi HotStart Ready Mix (Kapa Biosystems Inc., Wilmington, MA, USA). The following PCR procedure was used: 95 ◦C for three minutes, followed by 25 cycles of 95 ◦C for 30 s, 55 ◦C for 30 s, and 72 ◦C for 30 s. Finally, an extension was performed at 72 ◦C for five minutes. The Amplicon PCR products were purified using AMPure XP beads (Beckman Coulter, Inc., Brea, CA, USA), according to the manufacturer's instructions. A Nextera XT Index Kit v2 (Illumina Inc., Hayward, CA, USA) was used for the Illumina sequencing adapters and attachment of the dual indices. An index PCR was performed with 5.0 μL PCR product, 5.0 μL of each of the Nextera XT Index Primers, 25 μL 2× KAPA HiFi HotStart Ready Mix, and 10 μL PCR-Grade Water. The PCR was performed via the following procedure: 95 ◦C for three minutes, followed by eight cycles of 95 ◦C for 30 s, 55 ◦C for 30 s, and 72 ◦C for 30 s. Finally, an extension was performed at 72 ◦C for five minutes. The index PCR products were purified using AMPure XP beads (Beckman Coulter, Inc., Brea, CA, USA). The quality of the purifications was checked using the Agilent 2100 Bioanalyzer with a DNA1000 Kit (Agilent Technologies Inc., Santa Clara, CA, USA). Finally, the DNA library was diluted to 4 nmol/L.

Then, the DNA library was sequenced using the Miseq Reagent Kit v3 (Illumina Inc.) in the Illumina Miseq 2 × 300 bp platform, according to the manufacturer's instructions.

#### *2.7. Analysis of 16S rDNA Gene Sequences*

The 16S rDNA sequence reads were processed by the Quantitative Insights into Microbial Ecology (QIIME) pipeline version 1.9.1 [44]. The quality-filtered sequence reads were assigned to operational taxonomic units at 97% identity with the UCLUST algorithm [45]; these reads were then compared with reference sequence collections in the Greengenes database (August 2013 version). A total of 6,680,549 reads were obtained from the 91 samples. On average, 73,412 ± 4606 reads were obtained per sample. The taxonomy summary at the phylum to genus levels, alpha diversity such as the Simpson diversity index, beta diversity, and principal coordinate analysis (PCoA) were calculated and generated using QIIME. A PCoA analysis was also calculated using weighted UniFrac distances.

#### *2.8. Predicted Metagenomes*

In experiments 3 and 4, the functional profiles of microbial communities were predicted by the Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) [46]. The functional predictions were assigned to the Kyoto Encyclopedia of Genes and Genomes (KEGG) ortholog functional profiles of microbial communities via 16S sequences. We selected and examined categories related to "carbohydrate metabolism" for simplification and clarity of the analysis.

#### *2.9. Statistical Analysis*

The data were expressed as means ± standard error of the mean (SEM). All statistical analyses were performed using GraphPad Prism (version 6.03, GraphPad Software Inc., San Diego, CA, USA). We checked whether the data showed a normal or non-normal distribution and equal or biased variation via the D'Agostino-Pearson test/Kolmogorov–Smirnov test and F-value test/Bartlett's test, respectively. If the data showed a normal distribution and equal variation, the statistical significance was determined by the Student's *t*-test or one-way ANOVA with a Tukey's test or two-way ANOVA with a Tukey's post-hoc analysis if the interaction was significant. If the interaction was not significant but the main effect was, Sidak's post-hoc analysis was used. If the data showed a non-normal distribution or biased variation, the statistical significance was determined by the Mann–Whitney test or Kruskal–Wallis test with a Dunn post-hoc analysis and a two-stage linear step-up procedure of the Benjamini, Krieger, and Yekutieli test for multiple comparisons. The permutational multivariate analysis of variance (PERMANOVA) was used to assess the change of the microbiota composition. The PERMANOVA was analyzed by QIIME.

#### **3. Results**

#### *3.1. Inulin Intake Changed Microbiota Composition under Both Morning and Evening Timings*

In this experiment, cellulose, an insoluble dietary fiber, was added to the HFD as a control for inulin because the dietary fiber contained in this HFD was cellulose. The amount of dietary fiber in the food was kept the same between each group. Indeed, when comparing the HFD with and without cellulose, there was no significant difference in the body weight, food intake, cecal pH, amount of SCFAs, or microbiota composition of the mice (Figure S1). Therefore, this concentration of cellulose did not appear to affect these factors.

Some reports have suggested that inulin consumption induces changes in the microbiota composition [47–50]. Here, we divided the mice into two groups. Group 1 received cellulose and was fed an HFD with 2.5% cellulose in the morning and evening. Group 2 received inulin and was fed HFD with 2.5% inulin in the morning and evening. The mice were housed under each condition for 10 days, after which they were sacrificed at ZT20 (four hours after the morning intake) or ZT4 (four hours after the evening intake) on days 10–11 (Figure 1a). We sampled the cecal content and feces from the rectum and measured the cecal pH. There was no significant difference in the body weight between the two groups before sacrifice (Figure 1b), nor was there was a difference in the food intake between them (Figure 1c). There were no standard error bars in the food intake volume because of the group housing.

The cecal pH was significantly lower in the inulin group than in the cellulose group at both ZT20 and ZT4 (Figure 1d). The propionic acid level was significantly higher in the inulin group than in the cellulose group at ZT4, and the lactic acid level was significantly higher in the inulin group than in the cellulose group at ZT20. At ZT4, there was only a slight increase in the inulin group compared with the cellulose group. In the cellulose group, the lactic acid level was significantly different between ZT20 and ZT4. There were no significant differences in the acetic acid, propionic acid, and total SCFA levels between the cellulose and inulin group at either ZT20 or ZT4 (Figure 1e–i).

As the propionic and lactic acid increased and the cecal pH decreased, the microbiota may have changed due to inulin feeding. Therefore, we extracted 16S rDNA from the mice feces and analyzed the microbiota. In the cellulose group, the values of alpha-diversity as described by the Simpson index were significantly higher at ZT4 than at ZT20. The Simpson index was significantly higher in the inulin group than in the cellulose group at ZT20, but there was no significant difference observed at ZT4 (Figure 1j). Next, we examined the differences in the changes of the relative abundance of taxa between the inulin group and cellulose group. Some of the detected bacteria are shown in Figure 2. At the phylum level, the relative abundance of *Firmicutes* was significantly lower in the inulin group than in the cellulose group at ZT20. However, there was no significant difference in the relative abundance of *Bacteroidetes* between the inulin and cellulose groups, though the levels were increased slightly in the former (Figure 2a). At the genus level, the relative abundance of *Lactococcus* and *Streptococcus* significantly decreased in the inulin group at ZT20, and the relative abundance of *Oscillospira* significantly decreased in the inulin group at ZT4 (Figure 2b). We analyzed the PCoA of the weighted UniFrac distances and determined the beta-diversity of the microbiota composition (Figure S2a). In this experiment, we focused on the influence of inulin on the microbiota; thus, we primarily compared cellulose and inulin feeding. The beta-diversity of the microbiota composition was significantly different between the cellulose and inulin groups at ZT20 but not at ZT4 (Figure S2b,c).

These results suggest that inulin consumption changes microbiota composition. In addition, the inulin feeding time may have different effects on the microbiota because changes in the microbiota were more prominent at ZT20 (morning) than at ZT4 (evening), which showed significant and non-significant differences, respectively, compared with the cellulose group.

**Figure 1.** Inulin feeding decreased cecal pH and increased short-chain-fatty-acids. (**a**) Experimental schedule, where the white and black bars indicate environmental 12 h light and dark conditions, respectively. The gray bar indicates feeding with a high-fat-diet (HFD) and 2.5% cellulose. The yellow bar indicates feeding with HFD and 2.5% inulin. The black arrowhead indicates the sampling time. (**b**) Body

weight before sampling. (**c**) Average daily food intake. (**d**) Cecal pH of mice housed for 10 days for each group. (**e**–**i**) The short-chain fatty acids (SCFAs) of mice, including (**e**) acetic acid, (**f**) propionic acid, (**g**) lactic acid, (**h**) butyric acid, and (**i**) total SCFAs. (**j**) Bacterial alpha diversity. Comparison of the Simpson index estimation of the 16S rDNA gene libraries at 97% similarity from the sequencing analysis. All values except (**c**) are represented as mean ± SEM (cellulose at ZT20 (n = 5) and 4 (n = 5); inulin at ZT20 (n = 5) and 4 (n = 5)). \* *p* < 0.05, evaluated using the two-way ANOVA with Tukey's post hoc test. \$\$ *p* < 0.01, \$ *p* < 0.05, evaluated using the two-way ANOVA with Sidak's post hoc test. # *p* < 0.05, evaluated using the Mann–Whitney test with a two-stage linear step-up procedure of the Benjamini, Krieger, and Yekutieli test for multiple comparisons.

**Figure 2.** Inulin feeding changed the relative abundance of some bacteria. (**a**) Phylum level. (**b**) Genus level. All values are represented as mean ± SEM (cellulose at ZT20 (n = 5) and 4 (n = 5); inulin at ZT20 (n = 5) and 4 (n = 5)). \*\* *p* < 0.01, \* *p* < 0.05, evaluated using the two-way ANOVA with Tukey's post hoc test. # *p* < 0.05, evaluated using the Mann–Whitney test with a two-stage linear step-up procedure of the Benjamini, Krieger, and Yekutieli test for multiple comparisons.

#### *3.2. Inulin Intake in the Morning Rather than the Evening Strongly A*ff*ected the Microbiota Composition under Time-Restricted Feeding Conditions*

In this study, inulin may have had different effects on the microbiota depending on the feeding times. However, in experiment 1, we did not measure the effect of the feeding pattern. It is possible that the effect of inulin was increased at ZT20 (four hours after the morning intake) due to the high consumption in the morning. In the next experiment, we examined whether morning or evening inulin feeding affected the microbiota under the two meals-per-day schedule. The mice were divided into three groups. Group 1 received cellulose and was fed an HFD with 5% cellulose in the morning and evening. Group 2 received inulin in the morning and was fed an HFD with 5% inulin in the morning and an HFD with 5% cellulose in the evening. Group 3 received inulin in the evening and was fed an HFD with 5% cellulose in the morning and an HFD with 5% inulin in the evening. The mice were housed under each condition for 10 days, after which they were sacrificed at ZT20 and ZT4 on days 10–11 (Figure 3a). We sampled cecal contents and feces and measured the cecal pH. There was no significant difference in body weight between any group before sacrifice (Figure 3b), nor was there a large difference in total food intake between them. However, the total food intake was slightly higher if inulin intake was in the morning rather than in the evening (Figure 3c). The cecal pH was significantly lower in the morning inulin group than in the morning cellulose or evening inulin groups at ZT20. On the contrary, the pH was significantly lower in the evening inulin group than in the

evening cellulose and morning inulin groups at ZT4 (Figure 3d). The acetic acid, propionic acid, lactic acid, butyric acid, and total SCFA levels were significantly higher in the morning inulin group than in the morning cellulose or evening inulin groups at ZT20. However, the acetic acid, propionic acid, lactic acid, butyric acid, and total SCFA levels were significantly higher in the evening inulin group than in the evening cellulose or morning inulin groups at ZT4 (Figure 3e–i).

**Figure 3.** Morning inulin feeding decreased cecal pH and increased short-chain-fatty-acids more than evening inulin feeding. (**a**) Experimental schedule, where white and black bars indicate environmental 12 h light and dark conditions, respectively. The gray bar indicates feeding with a high-fat-diet (HFD) and 5% cellulose. The yellow bar indicates feeding with HFD and 5% inulin. The black arrowhead indicates the sampling time. (**b**) Body weight before sampling. (**c**) Average daily food intake. The gray bar indicates the average daily food intake of cellulose, and the yellow bar indicates the average daily food intake of inulin. (**d**) Cecal pH of mice housed for 10 days for each group. (**e**–**i**) SCFAs of mice, including (**e**) acetic acid, (**f**) propionic acid, (**g**) lactic acid, (**h**) butyric acid, and (**i**) total SCFAs. (**j**) Bacterial alpha diversity. Comparison of the Simpson index estimation of the 16S rDNA gene libraries at 97% similarity from the sequencing analysis. All values except (**c**) are represented as mean ± SEM (cellulose at ZT20 (n = 5) and 4 (n = 5); morning inulin at ZT20 (n = 5) and 4 (n = 5); evening inulin at ZT20 (n = 5) and 4 (n = 5)). \*\* *p* < 0.01, \* *p* < 0.05, evaluated using the two-way ANOVA with Tukey's post hoc test. # *p* < 0.05, evaluated using the Kruskal–Wallis test with Dunn post hoc test with a two-stage linear step-up procedure of the Benjamini, Krieger, and Yekutieli test for multiple comparisons. Cellulose, morning inulin, and evening inulin are C, M, or E, respectively.

Next, we extracted 16S rDNA from the mice feces and analyzed the microbiota. The value of alpha-diversity as determined by the Simpson index was significantly higher in the morning inulin group than in the morning cellulose or evening inulin groups at both ZT20 and ZT4 (Figure 3j). We also examined the differences between the changes of the relative abundance of taxa between the inulin and cellulose groups. Some of the detected bacteria are shown in Figure 4. At the phylum level, the relative abundance of *Bacteroidetes* was significantly higher in the morning inulin group than in the morning cellulose group at ZT20. Meanwhile, the relative abundance of *Firmicutes* was significantly lower in the morning inulin group than in the morning cellulose and evening inulin groups at ZT20 as well as significantly lower in the evening inulin group than in the evening cellulose group at ZT4

(Figure 4a). At the genus level, the relative abundance of *Lactococcus* was significantly decreased in the morning inulin group at both ZT20 and ZT4 and in the evening inulin group at ZT4, while the relative abundance of *Dorea* and *Allobaculum* was significantly increased in the morning inulin group (Figure 4b). We analyzed the PCoA of the weighted UniFrac distances and determined the beta-diversity of the microbiota composition (Figure S3a). At ZT20, the beta-diversity of the microbiota was significantly different between the cellulose and morning inulin groups and the morning and evening inulin groups (Figure S3b). At ZT4, the beta-diversity of the microbiota composition was significantly different among all of the groups (Figure S3c).

**Figure 4.** Morning inulin feeding changed the relative abundance of some bacteria. (**a**) Phylum level. (**b**) Genus level. All values are represented as mean ± SEM (cellulose at ZT20 (n = 5) and 4 (n = 5); morning inulin at ZT20 (n = 5) and 4 (n = 5); evening inulin at ZT20 (n = 5) and 4 (n = 5)). \*\* *p* < 0.01, evaluated using the two-way ANOVA with Tukey's post hoc test. # *p* < 0.05, evaluated using the Kruskal–Wallis test with Dunn post hoc test with a two-stage linear step-up procedure of the Benjamini, Krieger, and Yekutieli test for multiple comparisons. Cellulose, morning inulin, and evening inulin are C, M, or E, respectively.

These results suggest that morning inulin feeding affected the microbiota more than evening inulin feeding. However, the inulin intake was higher in the morning inulin group. Therefore, the increased consumption of the morning inulin group may have had more of an impact on the microbiota. To eliminate the effects of different food intakes, we prepared an apparatus to supply equal food amounts at two meals per day in the next experiment.

#### *3.3. Inulin Feeding in the Morning A*ff*ected the Microbiota Composition More than that in the Evening under Restricted Food Amount Conditions*

In this experiment, we provided the mice with two meals per day of 1.8 g of food at ZT12 (morning) and ZT20 (evening) to achieve equal food intake. The mice were divided into three groups. Group 1 received cellulose and was fed 1.8 g of an HFD with 5% cellulose in both the morning and evening. Group 2 received inulin in the morning and was fed 1.8 g of an HFD with 5% inulin in the morning and 1.8 g of an HFD with 5% cellulose in the evening. Group 3 received inulin in the evening and was fed 1.8 g of an HFD with 5% cellulose in the morning and 1.8 g of an HFD with 5% inulin in the evening. The mice were housed under each condition for 14 days, after which they were sacrificed at

ZT20 and ZT4 on days 14–15 (Figure 5a). We sampled cecal content and feces and measured the cecal pH. There was no significant difference in body weight between any group before sacrifice (Figure 5b). The cecal pH was significantly lower in the morning inulin group than in the morning cellulose group and significantly lower in the evening inulin group than in the evening cellulose group. Moreover, the cecal pH was significantly lower in the morning inulin group than in the evening inulin group (Figure 5c). The propionic acid, lactic acid, butyric acid, and total SCFA levels were significantly higher in the morning inulin group than in the morning cellulose group, while the propionic acid level was significantly higher in the evening inulin group than in the evening cellulose group (Figure 5d–h).

**Figure 5.** Morning inulin feeding decreased cecal pH and increased short-chain-fatty-acids more than evening inulin feeding under equivalent feeding conditions. (**a**) Experimental schedule, where white and black bars indicate environmental 12 h light and dark conditions, respectively. The gray cylinder indicates the 1.8 g of high-fat-diet (HFD) with 5% cellulose. The yellow cylinder indicates the 1.8 g of HFD with 5% inulin. The black arrowhead indicates the sampling time. (**b**) Body weight before sampling. (**c**) Cecal pH of mice housed for 14 days for each group. (**d**–**h**) SCFAs of mice, including (**d**) acetic acid, (**e**) propionic acid, (**f**) lactic acid, (**g**) butyric acid, and (**h**) total SCFAs. (**i**) Bacterial alpha diversity. Comparison of the Simpson index estimation of the 16S rDNA gene libraries at 97% similarity from the sequencing analysis. All values are represented as mean ± SEM (cellulose at ZT20 (n = 4) and 4 (n = 4); morning inulin (n = 5); evening inulin (n = 5)). \$\$ *p* < 0.01, \$ *p* < 0.05, evaluated using the two-way ANOVA with Sidak's post hoc test. # *p* < 0.05, evaluated using the Mann–Whitney test with a two-stage linear step-up procedure of the Benjamini, Krieger, and Yekutieli test for multiple comparisons.

Next, we extracted 16S rDNA from the mice feces and analyzed the microbiota. The value of alpha-diversity as determined by the Simpson index showed no significant difference between any group (Figure 5i). We also examined the differences in the changes of the relative abundance of taxa. Bacteria detected in over half of all samples are shown in Table 1. At the phylum level, the relative abundance of *Proteobacteria* was significantly increased in the morning inulin group, while the relative abundance of *TM7* was significantly decreased in the morning and evening inulin groups (Table 1a). At the genus level, the relative abundance of *Butyricimonas* was significantly increased in the morning

inulin group, while the relative abundance of *AF12*, *Staphylococcus*, *Lactococcus*, *Oscillospira*, *Bilophila*, and *Desulfovibrio* was significantly decreased in the morning inulin group. Meanwhile, the relative abundance of *AF12*, *Odoribacter*, and *Oscillospira* was significantly decreased in the evening inulin group (Table 1b). The number of bacteria changed by inulin feeding in the morning was higher than that changed by inulin feeding in the evening. We analyzed the PCoA of the weighted UniFrac distances and determined the beta-diversity of the microbiota composition (Figure S4a). The beta-diversity of the microbiota was significantly different between the cellulose group and morning inulin group (Figure S4b), but no significant difference was observed between the cellulose group and evening inulin group (Figure S4c). We predicted the functional profiles from sequencing data with PICRUSt. Among the categories related to "carbohydrate metabolism", the relative abundance of fructose and mannose metabolism was significantly increased in the morning inulin group but not in the evening inulin group (Figure S4d).

These results suggest that inulin feeding in the morning may affect the microbiota, even if the food intake amount is the same in the morning and evening.


**Table 1.** The relative abundance of some bacteria under the condition of two meals per day. (**a**). Phylum level. (**b**). Genus level.

**Table 1.** *Cont.*


(**a**) Number of bacteria significantly changed by M-inulin/all number of bacteria well-detected = 2/7. Number of bacteria significantly changed by E-inulin/all number of bacteria well-detected = 1/7. # *p* < 0.05, evaluated using the Mann–Whitney test with a two-stage linear step-up procedure of the Benjamini, Krieger, and Yekutieli test for multiple comparisons. (**b**) Number of bacteria significantly changed by M-inulin/all number of bacteria well-detected = 7/24. Number of bacteria significantly changed by E-inulin/all number of bacteria well-detected = 3/24. \$ *p* < 0.05, evaluated using the two-way ANOVA with Sidak post hoc test. # *p* < 0.05, evaluated using the Mann–Whitney test with a two-stage linear step-up procedure of the Benjamini, Krieger, and Yekutieli test for multiple comparisons.

#### *3.4. A Relationship Was Observed between the Length of Fasting Time and Inulin Feeding Stimulation*

In experiment 3, it was observed that inulin intake in the morning may have an effect on the microbiota and that the fasting time factored into this effect in the morning. The morning inulin group fasted for 16 h after the previous feeding, while the evening inulin group fasted for 8 h after the previous feeding, meaning that the time until breakfast was longer than the time until dinner. Therefore, the difference in the length of fasting time may have changed the effect on the microbiota. To test this hypothesis, we prepared an experiment with equal fasting times based on one meal a day, in which 3.6 g of food was given to the mice at either ZT12 (morning) or ZT20 (evening). The mice were divided into four groups. Group 1 received cellulose in the morning and was fed 3.6 g of an HFD with 5% cellulose in the morning. Group 2 received inulin in the morning and was fed 3.6 g of an HFD with 5% inulin in the morning. Group 3 received cellulose in the evening and was fed 3.6 g of an HFD with 5% cellulose in the evening. Group 4 received inulin in the evening and was fed 3.6 g of an HFD with 5% inulin in the evening. The mice were housed under each condition for 14 days, after which they were sacrificed at ZT20 and ZT4 on days 14–15 (Figure 6a). We sampled cecal content and feces and measured the cecal pH. The body weight was significantly increased in the evening cellulose and inulin groups compared with the morning cellulose and inulin groups. (Figure 6b). The cecal pH was significantly lower in the morning and evening inulin groups than in the morning and evening cellulose groups (Figure 6c). The propionic and lactic acid levels were significantly higher in the morning inulin group than in the morning cellulose group. In addition, the butyric acid level was higher, albeit not significantly, in the morning inulin group than in the morning cellulose group. Meanwhile, the lactic and butyric acid levels were significantly higher in the evening inulin group than in the evening cellulose group, and the propionic acid level was higher, albeit not significantly, in the evening inulin group than in the evening cellulose group (Figure 6d–h).

Next, we extracted 16S rDNA from the mice feces and analyzed the microbiota. The value of alpha-diversity as determined by the Simpson index was significantly higher in the morning cellulose group than in the evening cellulose group (Figure 6i). We also examined the differences in the changes of the relative abundance of taxa. Bacteria detected in over half of all samples are shown in Table 2. At the phylum level, the relative abundance of *Actinobacteria* was increased in the morning inulin group, but there was no significant difference in the relative abundance in the evening inulin group (Table 2a). At the genus level, the relative abundance of *Bifidobacterium* and *Allobaculum* was significantly increased in the morning inulin group, while the relative abundance of *Streptococcus*, *Oscillospira*, and *Ruminococcus* was significantly decreased in the morning inulin group. Meanwhile, the relative abundance of *Dorea* and *Allobaculum* was significantly increased in the evening inulin group, and the relative abundance of *Staphylococcus* and *Lactococcus* was significantly decreased in the evening inulin group (Table 2b). The number of bacteria changed by inulin feeding in either the morning or the evening was similar. We analyzed the PCoA of the weighted UniFrac distances and determined the beta-diversity of the microbiota composition (Figure S5a). The beta-diversity of the

microbiota was not significantly different between the cellulose and inulin groups in either the morning or evening (Figure S5b,c). We predicted the functional profiles from sequencing data with PICRUSt. Among the categories related to "carbohydrate metabolism", the relative abundance of fructose and mannose metabolism was not significantly different between the cellulose and inulin groups in either the morning or evening (Figure S5d). These results suggest that inulin intake in either the morning or evening with equal fasting periods does not change microbiota beta-diversity.

**Figure 6.** When fasting times are equal, the difference between morning and evening inulin feeding disappears. (**a**) Experimental schedule, where white and black bars indicate environmental 12 h light and dark conditions, respectively. The gray cylinder indicates the 3.6 g high-fat-diet (HFD) with 2.5% cellulose. The yellow cylinder indicates the 3.6 g of HFD with 2.5% inulin. The black arrowhead indicates the sampling time. (**b**) Body weight before sampling. (**c**) Cecal pH of mice housed for 14 days for each group. (**d**–**h**) SCFAs of mice, including (**d**) acetic acid, (**e**) propionic acid, (**f**) lactic acid, (**g**) butyric acid, and (**h**) total SCFAs. (**i**) Bacterial alpha diversity. Comparison of the Simpson index estimation of the 16S rDNA gene libraries at 97% similarity from the sequencing analysis. All values are represented as mean ± SEM (morning cellulose (n = 6); morning inulin (n = 6); evening cellulose (n = 6); evening inulin (n = 6)). \$ *p* < 0.05, evaluated using the two-way ANOVA with Sidak's post hoc test. # *p* < 0.05, evaluated using the Mann–Whitney test with a two-stage linear step-up procedure of the Benjamini, Krieger, and Yekutieli test for multiple comparisons. The table in (**j**) indicates the results using permutational multivariate analysis of variance (PERMANOVA). Morning cellulose, morning inulin, evening cellulose, or evening inulin are represented as M-cellulose, M-inulin, E-cellulose or E-inulin, respectively.



(**a**) Number of bacteria significantly changed by M-inulin/all number of bacteria well-detected = 1/6. Number of bacteria significantly changed by E-inulin/all number of bacteria well-detected = 0/6. # *p* < 0.05, evaluated using the Mann–Whitney test with a two-stage linear step-up procedure of the Benjamini, Krieger, and Yekutieli test for multiple comparisons. (**b**) Number of bacteria significantly changed by M-inulin/all number of bacteria well-detected = 5/28. Number of bacteria significantly changed by E-inulin/all number of bacteria well-detected = 4/28. \$\$ *p* < 0.01, evaluated using the two-way ANOVA with Sidak post hoc test. ## *p* < 0.01, # *p* < 0.05, evaluated using the Mann–Whitney test with a two-stage linear step-up procedure of the Benjamini, Krieger, and Yekutieli test for multiple comparisons.

#### **4. Discussion**

In this study, inulin intake changed the composition and profile of the gut microbiota, increased SCFAs, and decreased the cecal pH (Figure 1, Figure 2 and Figure S2). SCFAs are important for health because they improve energy metabolism in the liver and muscles and immune function in the large intestine [9–11]. In addition, the effect of inulin on the microbiota was dependent on the timing of inulin intake. Therefore, we gave inulin to the mice in either the morning or evening. The microbiota was more affected by inulin feeding in the morning than in the evening (Figure 5 and Figure S4) because the fasting period was longer for the latter. There has been previous research on fasting time and dietary effects. Previous studies examining postprandial glucose metabolism have shown that breakfast, rather than dinner, can suppress postprandial hyperglycemia and that one of the primary factors is the difference in fasting time [51]. Additionally, in a previous study examining the circadian clock, a meal after a long fasting period strongly synchronized the peripheral clock [52,53]. Under a two-meals-per-day schedule in mice, the same amount of chow after 16 h of fasting could reset the *Per2* gene expression rhythm in the liver clock compared with the same amount of chow after 6 h of fasting; in the two-meal experiments presented here, we used exactly the same protocol. In the current experiment, there was no difference in the cecal pH or SCFAs measurements between morning and evening with the same fasting duration (Figure 6c,h). Considering actual human life, the fasting time until breakfast is generally the longest among the three meals. Thus, these results, along with those of the previous study [53], support that inulin intake in the morning is most effective at attenuating HFD-induced changes of the gut microbiota. However, since the gut microbiota is also related to the circadian clock, there may be a difference between morning and evening in the gut microbiota composition, regardless of the fasting time. In addition to the daily feeding model used in this study, a feeding model for equalizing fasting time has been considered [52,54,55]. By using these feeding models, the relationship between fasting time and the effects of foods may be clarified. Moreover, the feeding model of this study has too long a starvation period compared with actual human life. Therefore, a feeding model that mimics the actual human lifestyle of three meals a day, as reported by Kuroda et al., may be considered for future experiments [52].

In this study, we first regulated the access time to inulin-containing food under a two-meals-per-day schedule because mice access food in the morning rather than in the evening under ad-lib food conditions [54,56]. Under these feeding conditions, we found clear effects of inulin in the morning. Therefore, in ad-lib feeding conditions, functional food intake at an earlier time during the active period may be a considerable factor in microbiota changes. Next, we regulated the food volume under a two-meals-per-day schedule. Once again, inulin in the morning had a clear effect on the microbiota, clarifying the importance of inulin intake in the morning on the beta-diversity and profile of the microbiota. However, in these experimental conditions, we did not control feeding and/or digestive speed; therefore, volume- and speed-controlled feeding systems may be required to determine the effect of feeding time.

The first meal after a long fast, most often breakfast, resets the phase of peripheral clock [52,53]. We recently demonstrated that cellobiose, a water-soluble dietary fiber, produces SCFAs, allowing them to reset the peripheral clock [21]. Taken together, these results suggest that the intake of inulin-containing foods in the morning may help reset the peripheral clock through SCFAs production.

Previous studies have reported that inulin consumption increases *Bifidobacteria* and *Akkermancia muciniphila* and decreases gram-positive cocci in humans and mice [47–50]. In this study, the gram-positive cocci *Streptococcus* and *Staphylococcus* decreased, but the *A. muciniphila* was not significantly changed. We considered that the degree of polymerization of inulin is one of the reasons that the results of this study differ from previous studies. In the structure of inulin, fructose is a monomer linked by 2–60 molecules with β-glycosidic bonds. The inulin used in this study had 16 fructose bonds (a degree of polymerization of 16) [34]. It has been reported that the influence on the microbiota is different depending on the degree of polymerization of inulin [48]. Therefore, the results may have been different with other degrees of polymerization.

*Streptococcus* is known to produce lactic acid [57,58], and *Streptococcus mutans* increases in the intestines of type 2 diabetes patients and is induced by a high-calorie diet [59]. Furthermore, *Staphylococcus aureus* is increased in obese patients, and *Staphylococcus* has a positive correlation with energy intake [60,61]. The SCFAs produced by inulin feeding increase the concentration of GLP-1 in

the blood and promote insulin secretion [62]. Furthermore, SCFAs regulate insulin activity in adipose tissue through the GPR43 receptor [11]. Therefore, it has been suggested that inulin may be an effective food against diabetes. In fact, in rats and humans, inulin consumption inhibits blood glucose levels and lowers blood triglyceride levels and total cholesterol levels [35,63,64]. In this study, the species level was not fully detected, and blood glucose levels and triglycerides were not measured. If these factors were measured, we may have been able to clarify the relationship between the gut microbiota and glucose metabolism.

It has been reported that SCFAs produced by ingestion of water-soluble dietary fiber prevent fat accumulation in adipose tissue via GPR43 [11]. However, it has also been reported that water-soluble dietary fiber does not involve SCFAs and suppresses fatty acid accumulation itself. For example, water-soluble dietary fiber may form a highly viscous matrix in the small intestine, increase the viscosity of the small intestine, and then physically suppress fat absorption [65,66]. These reports should be considered when investigating the association between gut microbiota and lipid metabolism.

The analysis of the carbohydrate metabolism identified a significant association with the fructose and mannose metabolism in the morning inulin group under two meals per day but not under one meal per day. Inulin is a fructan polymerized with fructose. Therefore, it may be possible that fructose metabolism is more activated by inulin in the morning than in the evening. Furthermore, the production of SCFAs may be increased because fructose is metabolized in the morning. In addition, fructose metabolism may also be related to fasting time. PICRUSt is only a predictive tool. To determine accurate functional information of the related bacteria, metagenomic studies should be conducted. Additionally, the number of mice in each group should be increased to provide more accurate explanations regarding the microbiota and PICRUSt analysis.

#### **5. Conclusions**

In summary, inulin intake in the morning rather than in the evening affected the gut microbiota, promoted SCFAs production, and lowered the cecal pH. The difference between the morning and evening results was related to the fasting duration, suggesting that there may be a relationship between fasting duration and meal stimulation regarding control of the microbiota.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6643/11/11/2802/s1, Figure S1: the addition of cellulose does not affect the microbiota. Figure S2: inulin feeding in the morning changed the microbiota composition more than in the evening. Figure S3: morning inulin feeding changed the microbiota composition more than evening inulin feeding. Figure S4: morning inulin feeding changed the microbiota composition and the relative abundance of inferred functional profile more than evening inulin feeding under equivalent feeding conditions. Figure S5: the microbiota composition and the relative abundance of inferred functional profile was not significantly changed by inulin feeding, even under the condition of one meal per day.

**Author Contributions:** H.S., H.M. and S.S.; designed the research and analyzed the data. H.S. and S.S.; wrote the manuscript. H.S., H.M., A.W., Y.N., Y.L. and K.H.; performed the experiments.

**Funding:** This work was partially supported by the Council for Science, Technology, and Innovation, SIP, "Technologies for creating next-generation agriculture, forestry, and fisheries" (funding agency: Bio-oriented Technology Research Advancement Institution, NARO) (Shibata. S.) and Japan Society for the Promotion of Science (JSPS) KAKENHI (A and Houga) (Shibata. S).

**Conflicts of Interest:** We declare that there are no conflicts of interest related to this study.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*

### **Changes in Microbiota and Bacterial Protein Caseinolytic Peptidase B During Food Restriction in Mice: Relevance for the Onset and Perpetuation of Anorexia Nervosa**

**Manon Dominique 1,2,3, Romain Legrand 1, Marie Galmiche 1,2,3, Saïda Azhar 1, Camille Deroissart 1, Charlène Guérin 2,3, Jean-Luc do Rego 3,4, Fatima Leon 3,4, Séverine Nobis 3,4, Grégory Lambert 1, Nicolas Lucas <sup>1</sup> and Pierre Déchelotte 1,2,3,5,\***


Received: 9 September 2019; Accepted: 15 October 2019; Published: 18 October 2019

**Abstract:** Microbiota contributes to the regulation of eating behavior and might be implicated in the pathophysiology of anorexia nervosa. ClpB (Caseinolytic peptidase B) protein produced mainly by the *Enterobacteriaceae* family has been identified as a conformational mimetic of α-MSH, which could result in similar anorexigenic effects. The aim of this study was to highlight the role of the microbiome and the ClpB protein in deregulation and self-maintenance of anorexia pathology. Male C57Bl/6 mice were undergone to the ABA (Activity-Based Anorexia) protocol: after 5 days of acclimatization, both ABA and LFA (Limited Food Access) mice had progressively limited access to food until D17. At the end of protocol, the plasma ClpB concentration and *Enterobacteriaceae* DNA in colonic content were measured. As expected, dietary restriction induced lost weight in LFA and ABA mice. At D10, colonic permeability and plasma concentration of the ClpB protein were significantly increased in LFA and ABA mice vs. controls. At D17, plasma concentration of ClpB was increased in LFA and ABA mice and, it was correlated with proportion of *Enterobacteriaceae* in the faeces. These abnormally high ClpB concentrations and all associated factors, and therefore might contribute to the initiation and/or perpetuation of anorexia nervosa by interfering with satiety signaling.

**Keywords:** anorexia; food restriction; ClpB; microbiota; *Enterobacteriaceae*

#### **1. Introduction**

Eating Disorders (ED) are public health problems that have continued to worsen in recent years with a prevalence of 3.5% from 2000–2006 to 7.8% in 2013–2018 [1]. Among these disorders, anorexia nervosa (AN) is characterized by a difficulty in maintaining a minimum weight and an obsession with weight and body shape [2], the pathophysiology of which is multifactorial and remains partially debated [3].

Among the proposed mechanisms of AN, the role of the gut microbiota in regulating the physiology of AN is increasingly recognized [4,5]. Indeed, studies have shown that intestinal microbial composition is influenced directly by food, in the short and long term [6–8]. Conversely, behavior [9] and appetite [10] are modulated at least in part by several gut-microbiota derived signals, among which bacterial products (e.g., peptides, neurotransmitters) have been shown to influence peripheral and central mechanisms of satiety, reward [11,12] and anxiety [13]. Finally, microbiota composition is implicated in the regulation of body composition: dysbiosis has been reported both in obese individuals [14] and in patients with AN [15]. Moreover, increased *Escherichia coli*, a leading representative of *Enterobacteriaceae* in gut microbiota was also observed in anorexic patients [16]. Altogether, these data strongly suggest that dysfunction of the microbiota-intestine-brain axis in response to exogenous triggering factors might be a key factor in the onset and/or perpetuation of ED [10,17]. Communication between microbiota, gut and brain may rely on various microbiota-derived signals, such as proteins, peptides, monoamines, metabolites, or even gut-produced immunoglobulins gaining access to the brain or modulating afferent neuronal or hormonal regulations generated in the splanchnic area [17]. Among bacterial proteins, ClpB (Caseinolytic peptidase B), a heat shock protein produced by *Enterobacteriaceae* [18] including *E. coli* is of particular relevance to the control of satiety [19] since it holds in common a six amino acid discontinuous epitope sharing molecular mimicry with α-melanocyte-stimulating hormone (α-MSH), the main central neuropeptide signaling satiety in the hypothalamus [20,21]. In addition, other studies have shown that α-MSH could also be found at peripheral level [22]. Moreover, α-MSH could induce the activation of MC4R present on intestinal enteroendocrine L cells [22,23]. Through this specificity, ClpB could stimulate the secretion by enteroendocrine L cells of the satiating hormones GLP-1 or PYY and activate vagal and hormonal pathways leading to hypothalamic activation of the POMC neurons releasing α-MSH [10,11]. In accordance with a role of this protein in the physiological and pathological regulation of eating behavior, ClpB was found naturally in the plasma of healthy subjects and at a higher level in patients with eating disorders [24].

In addition to the direct effect of ClpB mentioned previously, the hypothesis that microbial proteins may also modulate eating behavior through the intestinal production of specific immunoglobulins (Ig) can be suggested. Indeed, previous reports have detected Ig which react with α-MSH, in the sera of both healthy individuals and rats [25]. The levels of these Ig correlate with psychological traits characteristic of eating disorders [25]. This suggests that α-MSH reactive Ig may interfere with melanocortin signaling in both normal and pathological conditions. Moreover, a recent study showed that the levels of α-MSH-reactive IgG, the binding of melanocortin 4 receptor (MC4R) and the cellular internalization rate of MC4R-expressing cells were all lower in obese subjects [26]. Inverse results were found in anorexic and bulimic patients [26]. Other studies also confirmed the implication of α-MSH reactive Ig in the physiological regulation of feeding and mood [27]. In patients with eating disorders, increasing ClpB plasma levels correlated with plasma levels of anti-ClpB and anti-α-MSH Ig [19]. These factors emphasize the physiological involvement of anti-α-MSH Ig in the regulation of food intake.

Thus, bacterial ClpB protein appears as a candidate for interfering with endogenous pathway of satiety regulation. To get further insights in its involvement during food restriction, we performed the present study in a well-established model of food restriction in rodents, the Activity-Based Anorexia (ABA) model, and evaluated the impact of food restriction on the plasma ClpB protein and its related Ig and on the proportion of *Enterobacteriaceae*.

#### **2. Materils and Methods**

#### *2.1. Animal Experimentation*

Animal experimentation procedures were approved by the Local Ethical Committee of Normandy (approval CENOMEXA n◦1112–05). Male C57Bl/6 mice (Janvier Labs, Genest-Saint-Isle, France), at 7 weeks old were kept in holding cages (four mice per cage) at environmental conditions 22 ◦C ± 3 ◦C

and relative humidity of 40 ± 20% on a 12 h light-dark cycle with lights on at 10:00 a.m. During acclimatization period, all mice were given *ad libitum* access to water and standard food (Kliba Nafag, Germany).

At D1 of the protocol, all mice were randomized individually into 3 groups: An *ad libitum* group (Control, *n* = 16), a limited-food access group (LFA, *n* = 16) and an activity-based anorexia group (ABA, *n* = 16). ABA mice were placed individually in cages with an activity wheel connected to Running Wheel ® software (Intellibio, Seichamps, France).

Food access was progressively limited in ABA and LFA groups from 6 h per day at D6, to 3 h at D9 and until the end of the experiment. Mice always had free access to water. Body weight, water and food intake were measured at 9:00 a.m. each day.

At D10, 8 mice of each group were chosen according to their weight and were anaesthetized by ketamine/xylazine (Imalgene® 1000, Murial/Xylazine Rompun 2%, Bayer) intraperitoneally and were euthanized by decapitation. Blood samples were taken from the mesenteric artery before decapitation. The hypothalamus was taken to perform qPCR to analyze the anorexigenic (POMC) and orexigenic (AgRP) neuronal populations. Intracolonic faeces were taken to perform qPCR to analyze the *Enterobacteriaceae* DNA. The plasma was recovered after centrifugation (3000× *g*, 20 min, 4 ◦C). Samples were taken and stored at −80 ◦C if their analysis was not done immediately.

At D17, the end of the experiment, remaining mice underwent the same procedures as D10.

#### *2.2. Permeability*

Colon permeability was assessed by measured FITC-dextran (4 kDa) (Sigma) by Ussing chambers. FITC-dextran (5 mg/mL) was placed on the mucosal side. After 3 h at 37 ◦C, medium from the serosal side was removed and stored at −80 ◦C. The fluorescence level of FITC-dextran (excitation at 485 nm, emission at 535 nm) was measured in a 96-well black plate with spectrometer Chameleon V (Hidex, Turku, Finland). Values were converted to concentration (mg/mL) using a concentration standard curve.

#### *2.3. ClpB Concentration*

The presence of the protein ClpB was measured by the technique of enzyme linked immunosorbent assay (ELISA) previously described by Breton et al., 2016 [9]. For this, two antibodies were used: rabbit polyclonal anti-ClpB (Delphi Genetics, Brussels, BEL) and a mouse monoclonal antibody anti-ClpB (Delphi Genetics, Brussels, BEL). The optical density was determined at 405 nm using a microplate reader Infinite F50 (Tecan Life Sciences, Switzerland). Each determination was performed in duplicate.

#### *2.4. ClpB and* α*-MSH Ig Assay*

Plasma levels of Ig reacting with ClpB or α-MSH were measured using enzyme-linked immunosorbent assay according to a published protocol [28]. For this, a concentration of 2 μg/mL of ClpB protein (Delphi Genetics, Brussels, BEL) or α-MSH peptides (Bachem, Budendorf, Swiss) were used to coat 96-well Maxisorp plates (Nunc, Rochester, NY, USA). Mice plasma samples were diluted at 1:200 in dissociative buffer (3 M NaCl and 1.5 M glycine buffer, pH 8.9) to determine the total Ig levels. Two antibodies were used for detection: Alkaline phosphatase (AP)-conjugated goat anti-mouse IgG or anti-mouse IgM (1:2000) (Jackson ImmunoResearch Laboratories, St. Thomas Place, Ely, UK). The optical density was determined at 405 nm using an Infinite F50 microplate reader (Tecan Life Sciences, Switzerland). Blank optical density values (without the addition of plasma samples) were subtracted from the sample optical density values. Each sample was done in duplicate.

#### *2.5. qPCR Assay for Faecal Enterobacteriaceae DNA*

*Enterobacteriaceae* DNA in faeces were extracted with the ZymoBIOMICS Kit according to the protocol given by the supplier (ZymoResearch, Irvine, CA, USA). After extraction, the total DNA was quantified using a NanoDrop spectrophotometer (ThermoScientific, Waltham, MA, USA). qPCR

was performed on 1 ng/μL of DNA and with Light Cycler®480 SYBR® Green I Master (Roche, Swiss). The primers for detection of *Enterobacteriaceae* were: 5 -TGTGCCCAGATGGGATTAGC-3 and 3 -TTAACCTTGCGGCCGTACTC-5 . The relative quantity of each DNA was calculated using standard curves normalized to a reference *16s DNA* gene.

#### *2.6. RT-qPCR Assay for Hypothalamus Neuronal Populations mRNA*

Hypothalamic total RNA was extracted within cold TRIZOL reagent (Invitrogen, Carlsbad, CA, USA). After extraction, the total RNA was quantified using a NanoDrop spectrophotometer (ThermoScientific, Waltham, MA, USA). cDNA was generated by reverse transcription with 1 μg of total RNA using M-MLV Reverse Transcriptase (200 U/μL) (ThermoFisher, Waltham, MA, USA). RT-qPCR was performed on all samples using a BioRad CFX96 Real Time PCR System (BioRad, Hercules, CA, USA) and SYBR Green Master Mix (Life Technologies, Carlsbad, CA, USA). The primers for detection of *pomc* were: 5 -CCTCCTGCTTCAGACCTCCA-3 and 5 -GGCTGTTCATCTCCGTTGC-3 ; for *agrp*, 5 -GCAGACCGAGCAGAAGAT-3 and 5 -CTGTTGTCCCAAGCAGGA-3 . The relative quantity of each mRNA was calculated from standard curves, normalized to a reference *gapdh* gene.

#### *2.7. Statistical Analysis*

Data are shown as means +/− standard error of means (SEM). Before statistical analysis the normality was evaluated by the Kolmogorov-Smirnov test. Then, statistical significance was calculated by the unpaired *t*-test, one-way ANOVA or two-way ANOVA, as appropriate. All statistical calculations were performed using Prism 6.0 software (GraphPad Software, Inc., San Diego, CA, USA) and *p* < 0.05 was considered as significant.

#### **3. Results**

#### *3.1. Body Weight and Food Intake*

During the adaptation phase (D1–D6), the animals had the weight between 20 g and25 g. From the beginning of the dietary restriction (D6), the animals started to lose weight. The ABA and LFA mice lost significant weight compared to the control (\*\* *p* < 0.01, D7, D8). This difference in weight loss continues until the end of the experiment (D17) (\*\*\* *p* < 0.001, D9 to D17). From D10 until the end of endurance, ABA mice lost significantly more weight than the LFA mice (\* *p* < 0.05, D10, D12; \*\* *p* < 0.01, D13 to D17) (Supplementary data, Figure S1A).

Food intake (Supplementary data, Figure S1B) of ABA mice increased between D4 and D6 as compared to LFA and the control mice (\*\*\* *p* < 0.001) (Supplementary data, Figure S1B), (\*\*\* *p* < 0.001, D6) (Supplementary data, Figure S1C). However, since the beginning of the limited access to food (D7), food intake decreased significantly in all groups (\*\*\* *p* < 0.001) (Supplementary data, Figure S1B,C). At D10, when access time to food is shortest, food intake was significantly reduced compared to D6 (before restriction) (\*\*\* *p* < 0.001) with a reduction of 34% in the LFA group and 58% in the ABA group (Supplementary data, Figure S1C).

From D11, restriction was even intensified for the ABA group compared to the LFA group (\* *p* < 0.05, D12; \*\* *p* < 0.01, D13; \*\*\* *p* < 0.001, D15) (Supplementary data, Figure S1B) until the end of this experiment (reduction of 16.4%) (Supplementary data, Figure S1C).

**Figure 1.** Intestinal permeability measure and ClpB concentration in plasma. The intestinal permeability was measured by an ELISA assay after FITC-dextran passage in the ussing chamber (**A**) at D10 and (**B**) at D17. The ClpB concentration was measured in plasma in pM by an ELISA assay (**C**) at D10 and (**D**) at D17. Data are means ± SEM. Unpaired Mann-Whitney test (**A**, *p* = 0.0541) or ne-way ANOVA test with Holm-Sidak's post-tests (**C**,**D**); \*\*\* *p* < 0.001, \*\* *p* < 0.01, \* *p* < 0.05.

#### *3.2. Wheel Activity*

Total wheel activity increased during the restriction phase (D6–D10) of ABA mice. (Supplementary data, Figure S2A,B) as compared to the adaptation phase (D2–D5) which resulted mainly from an increased activity during the dark phase. (Supplementary data, Figure S2C,D). From D9, wheel activity decreased during the dark phase (Supplementary data, Figure S2C,D), while it increased during the light phase (Supplementary data, Figure S2E,F) vs. D6–D10.

#### *3.3. Intestinal Permeability, ClpB and Immunoglobulins Plasma Levels*

Intestinal permeability assessed in vitro by FITC-dextran flux increased at D10 (Figure 1A) while no difference was observed at D17 (Figure 1B).

ClpB protein concentration in plasma increased significantly at D10 (Figure 1C) and D17 (Figure 1D) in ABA and LFA groups vs. controls.

The plasma levels of anti-α-MSH IgG were increased at D10 and D17 in LFA vs. controls and the ABA group (Figure 2A,B). IgM anti-α-MSH levels were increased at D10 and D17 in LFA and ABA vs. controls (Figure 2E,F). The anti-ClpB IgG were increased at D17, but not at D10 in LFA vs. controls and ABA group (Figure 2C,D). The anti-ClpB IgM were increased at D10 but not at D17 in LFA vs. controls and vs. ABA group at D10 (Figure 2G,H).

**Figure 2.** Impact of food restriction and physical activity on modulation of anti-α-MSH and anti-ClpB IgG and IgM. Anti-α-MSH and anti-ClpB IgG antibodies (%) (**A**,**C**) at D10 and (**B**,**D**) at D17 were measured in plasma. Anti-α-MSH and anti-ClpB IgM were measured in the same way at (**E**,**G**) at D10 and (**F**,**H**) at D17 in plasma. Data are means ± SEM. Unpaired Mann-Whitney test (**D**,**F**) or unpaired *t*-test (**A**,**B**,**E**,**G**,**H**); \*\* *p* < 0.01, \* *p* < 0.05, \$ *p* < 0.10.

#### *3.4. Faecal Enterobacteriaceae DNA*

Relative quantitative amount of *Enterobacteriaceae* DNA in faeces was not different at D10 (Figure 3A) but increased at D17 (Figure 3B). This increased in *Enterobacteriaceae* DNA is positively correlated with the ClpB plasma concentration (Figure 3C).

**Figure 3.** *Enterobacteriaceae* DNA in faeces and correlation with ClpB Plasma concentration. Relative quantitative expression of *Enterobacteriaceae* DNA in faeces by qPCR (**A**) at D10 and (**B**) at D17. The relative expression was calculated with a 1 ng/μL *Enterobacteriaceae* concentration normalized by *16sDNA* gene. (**C**) Correlation between *Enterobacteriaceae* DNA in faeces and ClpB plasma concentration (pM). Data are means ± SEM. Unpaired Mann-Whitney test (**A**,**B**) or Pearson correlation (**C**); \*\* *p* < 0.01, \* *p* < 0.05.

#### *3.5. Hypothalamic Neuropeptides*

Hypothalamic POMC mRNA relative expression increased in LFA and ABA groups at D10 (Figure 4A) and D17 (Figure 4B). AgRP mRNA relative expression was not altered at either D10 (Figure 4C) or D17 (Figure 4D).

**Figure 4.** Impact of food restriction on neuronal population gene expression. Relative quantitative expression of (**A**,**B**) POMC and (**C**,**D**) AgRP mRNA in the hypothalamus by qPCR. The relative abundance of mRNA was calculated as the ratio of the normalized level (SQ of gene of interest mRNA/SQ of GAPDH mRNA). Data are means ± SEM. Unpaired *t*-test (**A**,**B**); \*\* *p* < 0.01, \* *p* < 0.05.

#### **4. Discussion**

In this study, we highlighted that an increase of ClpB plasma concentration correlated with the relative amount of *Enterobacteriaceae* in faeces of food restricted mice. As expected in the ABA model, mice significantly lost weight which was amplified by the wheel activity of ABA mice [29,30]. Reduced food intake in ABA was not a consequence of wheel activity, since activity ceased when food was again available, which rather suggests that the hyperactivity was a consequence of food restriction.

Several studies have previously reported an increase permeability in the colon of ABA mice [31,32] which suggests that a dysfunction of the intestinal barrier may occur during anorexia nervosa. In the present study, the increased intestinal permeability observed at D10 was associated with an increase in the plasma concentration of the ClpB protein. Accordingly, a previous study from our group [33] reported alterations of the colonic mucosa proteome in ABA mice, suggesting that a decreased energy supply to the colonic mucosa may compromise its functional integrity metabolism [33].

No difference in intestinal permeability or ClpB levels were observed between the LFA and ABA groups. This suggests that physical activity alone has no significant effect on colonic barrier function in this model, and that food restriction induces the increase of ClpB protein plasma level. This increased ClpB may result from an increased transcellular passage of this protein across the enterocytes. In fact, the enterocyte endocytosis of intact proteins is a well-established process. Milk proteins such as β-lactoglobulin (18.36 kDa) and α-lactalbumin (14.2 kDa) can cross the enterocytes via a non-specific liquid phase endocytosis mechanism and reach the basolateral side by a transcytosis mechanism [34]. Even larger proteins such as the 44 kDa glycoprotein Horseradish Peroxidase (HRP) can enter the intestinal absorptive cells by apical endocytosis [35]. Thus, endocytosis of the whole 96 kDa ClpB looks plausible.

Alternatively, paracellular passage may be possible for lower molecular weight proteins or fragments. An increased paracellular passage may be allowed by a degradation of the intercellular tight junction proteins network (e.g., occludin, claudin-1) [36], as already reported in inflammatory bowel diseases [37,38], irritable bowel syndrome [39,40], obesity [41] and malnutrition states [42] including the ABA model [31], and several intestinal diseases [43]. This paracellular pathway may be of relevance for the fragments of ClpB. Indeed, the ClpB protein has a capacity to fragment naturally as shown in vitro (Mogk et al., 1999), and smaller fragments may access the basolateral space before finally reaching the plasma compartment. This hypothesis is consistent with the increased ex vivo colonic passage of the FITC Dextran molecule (4 kDa) observed in food restricted mice. Our home-made ELISA test probably identified both the whole ClpB protein and several of its fragments.

In the present study, we also observed that food restriction induced changes in the plasma levels of anti-α-MSH IgG and IgM. The most consistent finding was an increase of anti-α-MSH IgG and IgM in the LFA group at D10 and D17. This immune activation may result from the activation of the hypothalamo-pituitary axis with the increased release of CRF and related-peptides such as α-MSH [44], and consequently an increase in the corresponding Ig. Our results suggest a higher increase of anti-α-MSH Ig as compared to anti-ClpB Ig. This may reflect the fact that the primary antibodies used for the ELISA assay may recognize other epitopes in the α-MSH structure in addition to those in common with ClpB. Furthermore, this increase could be a consequence of food restriction, with or without activity, which is stressful for mice. A previous study showed that repeated exposure of rats to mild stress induced by food restriction and repeated blood sampling increased the levels and affinity of α-MSH reactive IgG Ig [27]; passive transfer of these Ig purified from the blood of stressed to naïve animals induced acute food intake and suppressed anxiety. This suggests that the production of these Ig might be an adaptive response to stress aiming to counteract its effects by blunting the satiating effect of α-MSH. The moderate increase of anti-α-MSH Ig in the ABA group raises the question of a possible immunosuppressive effect caused by intense physical activity, which remains debated [45]. Although other studies need to be done, these results confirm the hypothesis proposed by Fetissov et al. that the anorexia physiopathology performed from altered signaling between the gut microbiota, the immune system and the neuropeptides involved in feeding behavior regulation [46].

In our study, colonic content analysis showed an increase proportion of *Enterobacteriaceae* DNA in the ABA group at D17 compared to controls, with intermediate values in the LFA group. This is in accordance with the increased ClpB plasma level at D17, since *Enterobacteriaceae* are known to produce the ClpB protein [18,24]. Already at D10 an increased production and/or release of ClpB by *Enterobacteriaceae* may have occurred before a significant growth of this family. Accordingly, the *Enterobacteriaceae* DNA correlated with ClpB plasma concentration across groups. Moreover, the significant correlation between the ClpB protein plasma levels and the relative amount of *Enterobacteriaceae* seen only in the anorexic mice suggests that dietary restriction impairs microbiota composition. The increased *Enterobacteriaceae* and ClpB protein production is in accordance with previous papers reporting increased *Enterobacteriaceae* in patients with anorexia and in malnourished animals [15]. Under critical dietary restriction conditions, an increased production of ClpB may also be an adaptive process to support the survival of the microorganism since ClpB is a chaperone protein [47]. Accordingly, Breton et al. reported that the production of ClpB by in vitro *E. coli* was increased in the stationary growth phase, after the disposal of added nutrients during the exponential phase [11]. The food restriction led to an increase in *Enterobacteriaceae* population, combined with an increase in both ClpB production and colonic permeability, together these factors have led to abnormally high plasma levels of ClpB.

The increased plasma concentration of ClpB or its fragments allows a direct central effect at the hypothalamic level of stimulating the POMC-related satiating pathways [11] which in turn contribute to either the onset or the perpetuation of anorexia and hyperactivity which could be explain by a satiating and anxiogenic effects of ClpB mimicking α-MSH [19].

It is important to emphasize that this biological approach of anorexia nervosa does not come in contradiction with the well-established triggering role of psychological stress. Indeed, stress might reduce food intake at the hypothalamic level or via the mesocorticolimbic system, but also at the peripheral level by increasing intestinal permeability [48,49] and altering microbiota virulence, proliferation and release of pro-inflammatory and anorexigenic signals acting on neuronal afferents [50, 51]. Using the data from this article and others, we can propose anintegrative perspective (Figure 5) which links dietary restriction, stress, microbiota-gut-brain axis dysregulation (including increased ClpB signaling) and the on-going self-maintenance of anorexia nervosa. When fully confirmed, this approach may open innovative therapeutic perspectives via modulation of gut microbiota by different nutritional, microbial or pharmacological approaches [52,53].

**Figure 5.** Vicious circle of the physiopathology of anorexia nervosa. Anorexia is characterized by psychological disorders (deformation of the self-image, obsessive fear of gaining weight) which are the cause of a restriction of food intake. This limited dietary intake leads to dysbiosis, characterized by an increase in *Enterobacteriaceae* within the microbiota. This increase generates an increased production of the ClpB protein, resulting from the prolongation of the stationary growth phase of these bacteria. In parallel, dietary restriction also causes an increase in intestinal permeability, which explains the increased passage of this protein through the intestinal mucosa. This protein is then found in the bloodstream with the other satietogenic peptides (GLP-1, PYY) released via the activation of the MC4R receptor present on the L cells. The mechanism of passage of this protein through the mucosa remains unknown, but hypotheses suggest that it may pass through the mucosa in fragments or through a mechanism of endocytosis. Finally, because of its anorectic action, the ClpB protein can activate anorexigenic neuronal populations such as POMC, whose response will lead to an increase in satiety. As well, the vicious cycle of the physiology of anorexia nervosa will can continue...

#### **5. Conclusions**

In conclusion, we have shown here that bacterial ClpB plasma levels increase during dietary restriction in mice, regardless of physical activity, and correlates with amount of *Enterobacteriaceae* in feces. This brings additional arguments for the role of the gut microbiota in the mechanisms of eating disorders and so, suggests its impact in the perpetuation and self-maintenance of the anorexia. These data suggest that nutritional or probiotic interventions aiming to restore gut microbiota may be useful in the therapeutic strategy of eating disorders.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6643/11/10/2514/ s1, Figure S1: Food restriction model confirmation–Body weight and Food intake, Figure S2: ABA model confirmation–Wheel activity.

**Author Contributions:** Conceptualization, R.L., N.L., G.L., P.D.; methodology, R.L., N.L.; formal analysis, M.D.; investigation, M.D., M.G., S.A., C.D., C.G., F.L., S.N.; resources, J.-L.d.R., P.D., G.L.; writing—original draft preparation, M.D. and P.D.; writing—review & editing, R.L., N.L., G.L., P.D.; visualization, M.D.; supervision, P.D.; project administration, P.D.; funding acquisition, P.D. and G.L.

#### **Funding:** This research received no external funding

**Acknowledgments:** Supported by European Union and Normandie Regional Council. Europe gets involved in Normandie with European Regional Development Fund (ERDF). The support of the TargEDys SA company and of the Ministry of Industry and Technology for the funding of M.D.'s PhD CIFRE thesis contract is acknowledged. The authors would like to thank Antonia Trower, Marketing Manager at TargEDys SA, for her help in manuscript proofreading.

**Conflicts of Interest:** M.D. and M.G. are PhD employee of TargEDys SA; N.L., R.L., S.A., C.D. were employees of TargEDys SA. P.D. is consultant for TargEDys SA, shareholder and cofounder of the company. G.L. is the CEO of TargEDys SA. Other co-authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **New and Preliminary Evidence on Altered Oral and Gut Microbiota in Individuals with Autism Spectrum Disorder (ASD): Implications for ASD Diagnosis and Subtyping Based on Microbial Biomarkers**

**Xuejun Kong 1,2,\*,**†**, Jun Liu 1,2,**†**, Murat Cetinbas 3, Ruslan Sadreyev 3, Madelyn Koh 1, Hui Huang 4, Adetaye Adeseye 4, Puhan He 4, Junli Zhu 5, Hugh Russell 6, Clara Hobbie 1, Kevi Liu <sup>1</sup> and Andrew B. Onderdonk <sup>7</sup>**


Received: 3 August 2019; Accepted: 23 August 2019; Published: 6 September 2019

**Abstract:** Autism Spectrum Disorder (ASD) is a complex neurological and developmental disorder characterized by behavioral and social impairments as well as multiple co-occurring conditions, such as gastrointestinal abnormalities, dental/periodontal diseases, and allergies. The etiology of ASD likely involves interaction between genetic and environmental factors. Recent studies suggest that oral and gut microbiome play important roles in the pathogenesis of inflammation, immune dysfunction, and disruption of the gut–brain axis, which may contribute to ASD pathophysiology. The majority of previous studies used unrelated neurotypical individuals as controls, and they focused on the gut microbiome, with little attention paid to the oral flora. In this pilot study, we used a first degree-relative matched design combined with high fidelity 16S rRNA (ribosomal RNA) gene amplicon sequencing in order to characterize the oral and gut microbiotas of patients with ASD compared to neurotypical individuals, and explored the utility of microbiome markers for ASD diagnosis and subtyping of clinical comorbid conditions. Additionally, we aimed to develop microbiome biomarkers to monitor responses to a subsequent clinical trial using probiotics supplementation. We identified distinct features of gut and salivary microbiota that differed between ASD patients and neurotypical controls. We next explored the utility of some differentially enriched markers for ASD diagnosis and examined the association between the oral and gut microbiomes using network analysis. Due to the tremendous clinical heterogeneity of the ASD population, we explored the relationship between microbiome and clinical indices as an attempt to extract microbiome signatures assocociated with clinical subtypes, including allergies, abdominal pain, and abnormal dietary habits. The diagnosis of ASD currently relies on psychological testing with potentially high subjectivity. Given the emerging role that the oral and gut microbiome plays in systemic diseases, our study will provide preliminary evidence for developing microbial markers that can be used to diagnose or guide treatment of ASD and comorbid conditions. These preliminary results also serve as a starting point to test whether altering the oral and gut microbiome could improve co-morbid conditions in patients with ASD and further modify the core symptoms of ASD.

**Keywords:** autism spectrum disorders; gut microbiota; oral microbiota; dysbiosis; co-occurring conditions; allergy; abdominal pain; biomarker discovery

#### **1. Introduction**

Autism Spectrum Disorder (ASD) is a complex neurological and developmental disorder with a rapidly increasing prevalence on a global scale [1]. The etiology of ASD likely involves an interplay between genetic and environmental factors, as well as both systemic inflammation and inflammation of the central nervous system (CNS) [2–4]. Recent studies suggest that microbiome dysregulation plays an important role in the pathogenesis of inflammation [5–8], which may contribute to the manifestation of ASD symptoms [9–12]. Evidence from animal studies supports a link between microbiome dysregulation, inflammation in the body, and development of ASD [13,14]. Patients with autism often have difficulties maintaining a balanced diet, due to multiple factors such as highly selective food preference, organic gastrointestinal (GI) diseases, and oral motor difficulties, and they show high rates of gut dysbiosis compared to neurotypical individuals [9,12]. Notably, some studies demonstrated a correlation between the severity of GI dysfunction and the severity of behavioral symptoms [15]. Gut dysbiosis may affect the CNS via the vagus nerve, microbial metabolites and neuroinflammation [16–18].

While most studies agree that the microbiome composition is different between autistic and neurotypical populations, these studies have yielded inconsistent results as to the nature or extent of these GI bacterial community differences [12,19]. Environmental factors are the dominant determinants for gut microbiome composition [20–22], yet most previous studies using age and sex matched controls have not adequately controlled for environmental influences [12,23,24]. In addition, compared to the gut, the oral microbiome is understudied, despite dental plaque and saliva samples being easier to obtain than stool samples. Alterations of the oral microbiota are associated with not only periodontal diseases [25], but also the upper GI tract flora [26], systemic diseases such as Rheumatoid Arthritis [27] and neurological conditions such as Alzheimer's disease [28]. Epidemiological studies have demonstrated a higher prevalence of oral health issues among patients with ASD, as compared to neurotypical individuals [29]. Only two studies to date have explored differences in oral microbiota between children with autism and controls [30,31]. Results from these studies have low degrees of concordance, likely due to the different sequencing methodologies and study designs.

Here, we have designed a pilot study to investigate the oral and gut microbiome simultaneously in patients with ASD and their first-degree family members. This would control for genetic and lifestyle factors while investigating the existence of ASD-microbiome signatures and whether these signatures hold any diagnostic value. Furthermore, to explore the poorly understood oral microbiome, we have directly compared oral and gut microbiome to explore their relationship in ASD and their association with systemic clinical indices. These questions are important to address in order to detail the roles of the human microbiome in ASD, and its utility in guiding diagnosis of ASD, clinical subtypes, and potential targeted interventions.

Given the multitude of factors that influence microbiome-host interactions, a secondary goal of the study attempts to characterize the potential relationships between the gut and oral microbiome and relevant clinical indices, including allergy, abdominal pain and dietary habits. Previously, Plaza-Diaz investigated gut microbiome in ASD patients with or without mental regression and found microbiome signatures associated with different psychiatric subtypes [32]. However, the association between medical subtypes and microbiome has been poorly explored in ASD patients.

Research on high impact diseases such as Rheumatoid Arthritis has revealed fascinating associations between oral and gut microbiomes [27]. Our study will serve as a starting point to address the complex interplay between the oral microbiome and the gut microbiome in the phenotypic presentation and pathophysiology of ASD. We believe that this study will open new horizons and opportunities in disease investigation and management. As a pre-probiotics clinical trial pilot project, we hope that this study and its continuation will provide insight for whether this new methodology with combined oral and fecal data can be used to (1) screen, diagnose, and determine subtypes of ASD, (2) stratify patients who may respond to probiotics therapy, (3) provide guidance on treatment strategies and develop targeted probiotic formulation, and (4) help to monitor treatment efficacy.

#### **2. Materials and Methods**

#### *2.1. Study Participants*

We recruited 20 patients diagnosed with ASD (autism spectrum disorder) and compared them with 19 family members (parent or sibling) as neurotypical controls. Patients had been diagnosed with ASD according to DSM-5 (Diagnostic and Statistical Manual of Mental Disorders) criteria [33]. Individuals with ASD between 7–25 years old with a disease duration of at least 6 weeks were enlisted. Exclusion criteria for all subjects included known genetic conditions, clinically evident serious infections or inflammatory conditions, history of cancer, severe dental/periodontal diseases or possession of dental braces. Subjects who had received probiotic treatment were asked to stop treatment at least one week prior to sample collection and subjects were excluded if they had taken antibiotics in the preceding month. Neurotypical controls had to meet the following criteria: biological sibling or biological parent of autistic subjects with IQ equal to or greater than 80 who do not have a diagnosis of ASD, attention deficit hyperactivity disorder, other intellectual developmental disorders, or psychiatric conditions. For recruitment of control subjects, siblings of the same gender and comparable age (+/− 5 years apart) received the highest priority, but an opposite-gender sibling was recruited for a control as needed. If the subject with ASD had no siblings, a parent acting as primary caretaker was recruited. Demographics and characteristics of study subjects are available in Table S1 and summarized in Table 1. Visual dental inspections were performed to determine oral health status for all subjects. Lifestyle questionnaires were distributed to assess factors that could affect microbiome status and create a GI clinical indices (GSI) score (Table 1, Table S1) [34,35].


**Table 1.** Characteristics of study participants and microbiome lifestyle factors.

*Autistic Neurotypical* Allergies (n) Yes 60% (12) 37% (7) No 40% (8) 63% (12) Drink alcohol (n) Yes 0% (0) 11% (2) No 95% (19) 84% (16) n/a 5% (1) 5% (1) Recreational drugs (n) Yes 0% (0) 0% (0) No 95% (19) 19% (18) n/a 5% (1) 5% (1) Tobacco products (n) Yes 0% (0) 0% (0) No 95% (19) 95% (18) n/a 5% (1) 5% (1) First 6 months of life Breast Fed 70% (14) 74% (14) Bottle Fed 15% (3) 5% (1) Both 25% (5) 16% (3) n/a 5% (1) 5% (1) Picky Eater Yes 20% (4) 11% (2) No 80% (16) 84% (16) n/a 0% (0) 5% (1) Servings of vegetables and fruits per day (n) Less than three 65% (13) 74% (14) Three 30% (6) 21% (4) More than three 5% (1) 5% (1)

**Table 1.** *Cont.*

ASD patients were recruited from clinics at Massachusetts General Hospital (MGH), Beth Israel Deaconess Medical Center, community ASD education events, and charity ASD programs in Boston. The study was approved by institutional review board of MGH (Boston, MA, USA, IRB protocol number: 2017P000573). Informed consents were obtained from subjects or the legal guardians of the subjects. All methods were performed in accordance with the relevant guidelines and regulations.

#### *2.2. Sample Handling and Collection*

To obtain oral microbiome samples, participants were asked to produce 1–3 mL of saliva after refraining from eating, drinking and oral hygiene practice for 1 h. Samples were collected with sterile DNA- and RNA-free 15 mL Falcon tubes and immediately frozen at −80 ◦C. De-identified and coded samples were shipped to Precidiag Inc. (Natick, MA, USA) for DNA extraction and sequencing on dry ice. Stool samples were collected by the participants at home under the supervision of trained parents with a HR-Easy Stool Collection Kit (Precidiag, Inc.) and stored at room temperature, followed by de-identification and shipment to a Precidiag CLIA-certified laboratory for DNA extraction and sequencing analysis. The HR-Easy Stool Collection Kit provides a superior method for collection, storage and stabilizing stool samples for microbiome study at ambient temperature for up to a month with minimal alterations when compared with freshly-collected samples (Yu et al., manuscript in preparation). Microbial DNA was then extracted using a HR-Easy Fecal DNA Kit (Precidiag, Inc.) according to the manufacturer's instructions and DNA samples were carefully quantified with a nanodrop spectrophotometer. A260/A280 ratios were also measured to confirm high-purity DNA yield. DNA samples were frozen at −20 ◦C until use.

#### *2.3. 16S rRNA Gene Amplicon Sequencing*

Microbial 16S rRNA V3-V4 genomic regions from total oral and gut DNA samples were amplified with the following primers 341F: 5 AATGATACGGCGACCACCGAGATCTA-CACTCTTTCCCTAC ACGACGCTCTTCCGATCTCCTACGGGAGGCAGCAGCCTACGGGNBGCASCAG3 and 805R: 5 CAAGCAGAAGACGGCATACGAGATNNNNNNG-TGACTGGAGTTCAGACGTGTGCTCTTT CCGATCTGACTACNVGGGTATCTAATCC3 via polymerase chain reaction (PCR) (95 ◦C for 2 min, followed by 25 cycles at 95 ◦C for 30 s, 55 ◦C for 30 s, and 72 ◦C for 30 s, and a final extension at 72 ◦C for 5 min). PCR products were purified and analyzed using a Bioanalyzer DNA kit, followed by quantification with real-time PCR. Serially diluted PhiX control library (Illumina, San Diego, CA, USA) was included as a standard. DNA libraries were pooled and sequenced on an Illumina MiSeq next-generation sequencing system (Illumina; CA) using a V2 2 × 250 bp paired-end protocol with overlapping reads.

Of note, we included strict quality control processes involving microbial DNA extraction, 16S rRNA gene amplicon amplification, and amplicon sequencing with a set of controls that enabled us to evaluate the potential introduction of contaminants or off-target amplification. Non-template controls (extraction chemistries) were included in the microbial DNA extraction process and the resulting material was subsequently used for PCR amplification. Additionally, at the step of amplification, another set of non-template controls (PCR-mix) was included to evaluate the potential introduction of contamination at this step. Similarly, a positive control comprised of known and previously characterized microbial DNA was included at this step to evaluate the efficiency of the amplification process. Before samples were pooled together, sequencing controls were evaluated, and samples were rejected if the presence of amplicons in any of the non-template controls or the absence of amplicons in the positive control was detected. In the present study, no amplicons were observed in the non-template controls and a negligible number of raw reads were recovered after sequencing.

#### *2.4. Sequencing Data Processing*

Sequencing data were processed and analyzed with a QIIME software package v. 2018.2.0 [36]. The sequencing reads with a low quality score (average Q < 25) were truncated to 240 bp, followed by filtering using the deblur algorithm with default settings [37]. The remaining high-quality reads were aligned with the reference library using mafft [38]. Next, the aligned reads were masked to remove highly variable positions, and a phylogenetic tree was generated from the masked alignment using the FastTree method [39]. Taxonomy assignment was performed using the feature-classifier method and naïve Bayes classifier trained on the Greengenes 13\_8 99% operational taxonomic units (OTUs) (Table S2).

#### *2.5. Biostatistical Analysis*

#### 2.5.1. Variables Measured

The main variables are the compositions of oral and gut microbiome, and quantities of microbes on genus and phylum level within each sample (OTUs). Other variables include patients' demographic information, baseline medical conditions, lifestyle factors and clinical indices.

#### 2.5.2. Alpha and Beta Diversity

Alpha diversity was calculated on the basis of the gene profile for each sample based on the Shannon index, Faith's index, and Simpson's evenness index [40–42]. Beta diversity was calculated on the unweighted and weighted UniFrac distances, Jaccard and the Bray–Curtis dissimilarity [43,44]. Alpha and beta-diversity estimates were computed using QIIME2 [36]. Alpha and beta diversity metrics and Principal Component Analysis plots based on the Jaccard distance were generated using default QIIME2 plugins [36,43,45–47].

Kruskal–Wallis tests were used to compare alpha diversity between ASD patients and controls for oral or gut microbiome respectively. A cut off false discovery rate (FDR) of 0.05 based on the Benjamini–Hochberg (BH) method was applied [48]. Comparison of beta diversity indices were calculated by Permutational multivariate analysis of variance (PERMANOVA).

#### 2.5.3. Statistical Analyses of Differentially Enriched Microbiome Taxa

Significant differences in the relative abundance of microbial genera and phyla between individuals with ASD and controls were identified by Kruskal–Wallis tests and BH adjustment for multiple comparisons. In addition, we performed a paired Wilcoxon signed-rank test on the relative abundances with BH adjustment. Furthermore, we explored differential bacteria enrichment on all taxonomy levels using the ANCOM (Analysis of Composition of Microbiomes) method, an algorithm that accounts for compositional constraints to reduce false discoveries in detecting differentially abundant taxa at an ecosystem level, while maintaining high statistical power [49]. An FDR cutoff of 0.2 was applied for taxa-level comparison [50].

#### *2.6. Microbiome Biomarker Discovery*

In order to measure whether the relative abundance of gut and oral microbial taxa and the dysbiosis markers could classify ASD and control groups correctly, we created a receiver operator characteristics (ROC) curve using Prism GraphPad (version 7.00 for Mac, GraphPad Software, La Jolla, San Diego, CA, USA, www.graphpad.com). Statistical significance of areas under the curves (AUCs) for dysbiosis markers were performed with the default plugin of Prism GraphPad.

#### *2.7. Microbiome Network Analysis*

In order to assess the taxonomic relatedness/association within the gut and oral microbiota as well as between oral and gut microbiota, we performed correlation-based network analysis using the SparCC (Sparse Correlations for Compositional data) method [51,52]. We performed SparCC for microbiome data on phylum and genus level from all subjects, as well as within ASD and control groups, respectively (Correlation coefficient cut-off = 0.3).

#### *2.8. Influence of Clinical and Lifestyle Factors*

Kruskal–Wallis tests with BH adjustment for FDR were used to assess differential abundance of dysbiosis markers and bacterial taxa (phylum and genus level) between binary clinical classifiers (i.e., presence or absence of allergy, constipation and abdominal pain) with a FDR cut off of 0.2. Relevant clinical indices were treated as binary even though some data were collected as ordinal (e.g., GSI scores). Analysis was further stratified by ASD and control groups. Genus level analysis was performed with genera that have a relative abundance of at least 0.5%. We compared the dietary habits between ASD patients and neurotypical controls based on numerical scores from baseline survey questions. The responses for each question were recorded on a numerical scale from 0 to 4, where a larger score indicated that the subject exhibited the behavior with greater prominence. We next assessed the correlation between eating habit scores, allergy/autoimmunity scores, GSI total score, and key ASD gut microbiome markers in patients with ASD. We used the Spearman's correlation and an FDR cutoff of 0.05.

#### *2.9. Softwares Used*

QIIME software package v. 2018.2.0 [36], RStudio (RStudio Team, 2017), R (R Core Team, 2017) and Prism GraphPad version 7.00 for Mac (GraphPad Software, La Jolla, CA, USA) were used for statistical testing and graph generation. Adobe Illustrator CC was used for figure editing.

#### **3. Results**

To characterize the gut and oral microbiota associated with autism, we recruited 20 autistic subjects and 19 controls (Table 1). Of the controls, 8 were neurotypical biological parents and 11 were neurotypical biological siblings. Demographic information is summarized in Table 1. One family had 1 parental control with 2 ASD children. Overall, there were significant inter-subject and inter-pair variabilities in microbiota composition (Figure 1A,B, Figures S1 and S2).

**Figure 1.** Bar plots of bacterial phylum-level relative abundances of the salivary (**A**) and gut (**B**) microbiomes. Each bar represents one subject. (**C**) Salivary microbiome class-level heatmap expression profile. (**D**) Gut microbiome class-level heatmap expression profile.

#### *3.1. Autistic Subjects Harbor an Altered Oral Microbiota Compared to First Degree-Family Member Controls*

Consistent with previous studies, analysis of alpha diversity calculated by the Shannon index revealed no significant differences between autistic and neurotypical subjects' salivary microbiota (Figure S2, Table S3). A heatmap (Figure 1C) visiually demonstrates that the beta diversity calculated on the unweighted, weighted UniFrac distances and the Bray–Curtis dissimilarity revealed no significant difference between the ASD and control groups for oral flora (Figure 2A, Figure S3, Table S4, PERMANOVA). The major phyla that contributed to the oral microbiome in ASD and control groups are summarized in Figure 2C. On the genus level, the ASD and control groups share 9 out of 10 most abundant genera, including *Prevotella*, *Fusobacterium*, *Rothia*, *Haemophilus*, *Streptococcus*, *Neisseria*, *Veillonella*, and an unknown genus in the *Neisseriaceae* family.

**Figure 2.** PCA of bacterial beta diversity of saliva (**A**) and gut (**B**) microbiomes based on the Bray–Curtis dissimilarity for ASD and neurotypical subjects. ASD and neurotypical subjects are colored in blue and red, respectively. (**C**) The major contributing phyla of gut and oral microbiome, in ASD and control subjects. The values used to compose the figures represent group mean relative abundances. (**D**,**E**) Box plots depicting relative abundances of the most differentially abundant salivary or gut bacterial phyla between patients with ASD and control subjects. Single asterisk indicates *p* < 0.1 with adjusted FDR > 0.2; double asterisk indicates *p* < 0.05 with adjusted FDR > 0.2, triple asterisk indicates *p* < 0.05 and adjusted FDR < 0.2, Kruskal–Wallis test.

We found differential enrichment of bacterial taxa in the oral microbiota of autistic individuals compared to the controls. On the phylum level, ASD patients showed a trend of lower relative abundance of *TM7* bacteria (Figure 2D, Figure S4). In total, 6 genera showed altered relative abundance between the two groups (Kruskal–Wallis test, *p* < 0.05, Figure 3A, Figure 4B, Table S5). In particular, the relative abundance of an unspecified genus in the class of *Bacilli* was statistically significant after adjusting for the false discovery rate (FDR) (Figure 5B, Table S5).

#### *3.2. Autistic Subjects Harbor an Altered Bacterial Gut Microbiota Compared to First Degree-Family Member Controls*

Consistent with previous studies, the analysis of gut alpha and beta diversity as well as principal component analysis (PCA) revealed no significant differences between autistic and neurotypical subjects (Figure 2B, PERMANOVA, Table S4, Figure S3), as visualized by a heatmap (Figure 1D). On the phylum level, *Firmicutes*, *Bacteroidetes* and *Proteobacteria* are the most abundant gut phyla in both ASD patients and control subjects, comprising more than 90% of all operational taxonomic units (OTUs) (Figure 2C). On the genus level, ASD and control groups share 9 out of 10 most abundant genera, including *Bifidobacterium*, *Blautia*, *Prevotella*, *Bacteroides*, *Faecalibacterium*, and unknown genera in *Ruminococcaceae* family, *Lachnospiraceae* family, *Enterobacteriaceae* family and *Clostridiales* order.

**Figure 3.** (**A**,**B**) Box plot representations of the relative abundances of differentially abundant salivary or gut bacterial genera in patients with Autism Spectrum Disorder (ASD) and control subjects. (**C**) Box plots representation of gut phylum-level dysbiosis marker *Firmicutes*/*Bacteroidetes* ratio, in patients with ASD and control subjects. ASD and neurotypical subjects are colored in blue and red, respectively. Single asterisk indicates *p* < 0.1 with adjusted FDR > 0.2; double asterisk indicates *p* < 0.05 with adjusted FDR > 0.2, triple asterisk indicates *p* < 0.05 and adjusted FDR < 0.2, Kruskal–Wallis test. (**D**) receiver operator characteristics (ROC) curve of the 3 differentially abundant gut or oral genera and dysbiosis markers that have the highest area under the curve (AUC), and *p* < 0.05 based on two-sided Z-test for ROC.

**Figure 4.** (**A**,**B**) Overlap of differentially abundant gut or salivary genera based on Kruskal–Wallis test and paired Wilcoxon test. Results are for taxa with unadjusted *p* < 0.05. (**C**,**D**) Paired-test representation of the relative abundances of top most differentially abundant salivary bacterial genera between ASD patient–family member control pairs. (**E**,**F**) Paired-test representation of the relative abundances of top most differentially abundant gut bacterial genera between ASD patient–family member control pairs. Single asterisk indicates *p* < 0.1 with adjusted FDR > 0.2; double asterisk indicates *p* < 0.05 with adjusted FDR > 0.2, triple asterisk indicates *p* < 0.05 and adjusted FDR < 0.2, Wilcoxon's paired test.

Further analysis of the dysbiosis markers revealed differences in the gut microbiota of subjects with autism and their family member controls. Several phylum level markers showed statistically significant changes between ASD and control, including *Firmicutes*/*Bacteroidetes* ratio (Figure 3C) likely driven by *Bacteroidetes* (Figure 2E, Figure S4). The phylum *Proteobacteria* is associated with metabolic syndrome and inflammatory bowel disease (IBD), and normally makes up less than 10% of the gut microbiome in healthy individuals [53]. Among the six subjects with significant *Proteobacteria* overgrowth (with

relative abundance values greater than 30%), 4 were ASD patients (Table S1). On the genus level, 6 taxa showed trends of altered abundance between the two groups, including *Paraprevotella*, *Granulicatella, Butyricimonas*, *cc\_115*, *Peptoniphilus* and *Eubacterium* (Figure 5A, Table S5).

' \* XWDQGRUDOP LFURELRP HSK\OXP OHYHO 6SDU&&FRRFFXUUHQFHQHWZ RUN&RQWUROV&RUU!

(\* XWDQGRUDOP LFURELRP HSK\OXP OHYHO 6SDU&&FRRFFXUUHQFHQHWZ RUN\$ 6' &RUU!

**Figure 5.** (**A**) Phylum-level heat map expression profiles of gut and oral microbiomes in ASD patients. (**B**) PCA of bacterial beta diversity based on Bray–Curtis dissimilarity for saliva and gut (all subjects are represented). Saliva and gut microbiome are colored in yellow and green, respectively. (**C**–**E**) Gut and oral microbiome phylum level co-occurrence network using the Sparse Correlations for Compositional data (SparCC) method with a correlation cut-off >0.3 ((**C**) all subjects, (**D**) control only, (**E**) ASD only). Each node represents a saliva (Sl) or stool (St) phylum, and saliva and stool microbiomes are colored in yellow and green, respectively. The dotted red circle highlights a co-occurrence cluster with the greatest inter-nodal correlations.

#### *3.3. Gut and Saliva Biomarkers Can Classify ASD and Control Groups*

In order to measure how correctly the relative abundance of gut and oral microbial taxa and the dysbiosis markers could classify two groups of samples, we created a receiver operator characteristics (ROC) curve, which is a common methodology used to evaluate classification performance of potential biomarkers (Figure 3D). The performance of a potential classifier (binary) can be evaluated by measuring the area under the curve (AUC), which represents true versus false positive rates. An AUC value of 0.5 corresponds to random classification and a value of 1.0 corresponds to perfect classification. Taking all gut and saliva genera as well as gut dysbiosis markers that showed statistically significant differential expression (Kruskal–Wallis tests) from previous analyses, two genera (gut *Butyricimonas*, saliva *Parvimonas*, Figure 3A,B) and the well-recognized dysbiosis marker gut *Firmicutes*/*Bacteroidetes* ratio (Figure 3C), all showed the highest AUC values (up to 0.724) with *p* value < 0.05 (Figure 3D, Table S8).

#### *3.4. Results of Paired Analysis Overlap Partially with Group Analysis*

Due to the nature of paired study design, we also performed paired a Wilcoxon signed-rank test on the relative abundance of the OTUs, in addition to Kruskal–Wallis tests, by subject groups (ASD vs. control). Those with significant Wilcoxon's *p* values had partial overlap with results from grouped Kruskal–Wallis tests (Figure 4A,B). However, after adjustment for multiple comparison, FDRs from paired analyses were not statistically significant (Table S6). Examples of gut and oral genera that showed the most significant pairwise changes are recorded in Figure 4C–F. Due to high inter-individual variabilities, subsequent analysis consisted of group-wise approaches.

In addition to Kruskal–Wallis tests with FDR adjustment, we explored differential bacteria enrichment on all taxonomy levels using the more conservative ANCOM method [49]. This method did not reveal statistically significant differences in the enrichment patterns detected by the Kruskal–Wallis test (Table S7).

#### *3.5. Exploring the Relationship between Gut–Oral Microbiome and Their Co-Occurrence Network*

Since the current project characterized gut and oral microbiota samples from the same subjects, we explored the relationship between gut and oral microbiota within individuals. Consistent with previous publications, we found that the gut and oral microbiome are distinct, based on beta diversity indices and PCA (Figure 5B, PERMANOVA). This can be seen through heatmap clustering (Figure 5A) as well as the OTU level ANCOM analysis (Figure S5, Table S7).

In order to assess the taxonomic association within the gut and oral microbiota as well as between oral and gut microbiota in a non-biased manner, we performed correlation-based network analysis using the Sparse Correlations for Compositional data (SparCC) method [51,52] (Figure 5C). This method is capable of estimating correlation values from compositional data and has been validated as a superior analysis technique than Pearson's correlation methods for compositional data such as 16S rRNA gene amplicon sequencing [51]. The goal of this analysis is to infer any potential synergistic relationships between bacterial taxa within a community and between communities. We also hoped to detect GI dysbiosis purely using salivary microbial markers because, due to high prevalence of constipation in the ASD population, it is much easier to obtain saliva samples than stool samples. The salivary microbiome could then serve as a diagnostic window into the GI environment of the ASD patients. Previously, network correlation analysis has yielded important insights regarding bacterial community structures related to enterotypes [54].

Overall, the oral microbiome exhibits a denser co-occurrence network compared to the gut, both at the phylum and genus level (Figure 5C–E). The same trend holds true when analyzing ASD subjects and control subjects separately (Figure 5D,E). Within the salivary co-occurrence network at the phylum level, the highest correlations are observed in a cluster consisting of *Actinobacteria*, *Proteobacteria*, *Firmicutes* and *Bacteroidetes* (Figure 5C, dotted circle), especially between *Firmicutes* and *Actinobacteria* (Figure S6). Importantly, some gut and oral phylum show positive inter-community co-occurrence. There is a positive correlation between saliva *Verrucomicrobia* and gut *Actinobacteria* (Figure 5C). In the ASD population but not the controls, gut *Firmicutes*, which is a known dysbiosis marker, showed positive correlation with saliva level of *Chloroflexi* (Figure 5D,E). We then computed the co-occurrence network on the genus level using bacteria genera that make up at least 0.5% of all OTUs. The genus-level co-occurrence density was notably higher compared to phylum level (Figure S6B,C), as many genera demonstrated intra-community co-occurring relationships. In terms of inter-community co-occurrence, several gut genera, including *Bifidobacteria*, *Dialister*, *Escherichia*, *SMB53* and an unspecified genus in *Enterobacteriaceae* all exhibited positive correlation with salivary genera in the control subjects, whereas only *Escherichia* and an unspecified genus of *Clostridiales* showed co-occurrence with saliva genera in ASD patients (Figure S7).

Alpha diversity has been conventionally used as an index for dysbiosis, as low alpha diversity indicates diminished community richness and potentially diminished resilience to disturbances. Alpha diversity shows a positive correlation between the gut and oral microbiota, although it is not statistically significant (Table S3, Figure S8).

#### *3.6. Microbiome Signatures in Clinical Subtypes*

Due to the tremendous clinical heterogeneity of the ASD population, we explored the relationship between microbiome and clinical indices as an attempt to extract microbiome signatures assocociated with clinical subtypes. We focused on three major medical comorbidities that have previously reported associations with microbiome, including allergy, GI disturbances and poor diet.

#### *3.7. Allergies*

We first investigated whether phylum level dysbiosis markers (including gut *Proteobacteria*, *Firmicutes*, and *Bacteroidetes*, and oral *SR1* and *Synergistetes*) may be associated with disease states. Among all clinical indices assessed, the incidences of allergy were notably higher in the ASD group (7/19 vs. 11/18, Chi-square test, *p* value < 0.05). The relative abundance of oral *SR1* is significantly lower in ASD patients who also have allergies, in comparison to ASD patients without allergies, but this trend is not present in control subjects (Kruskal–Wallis, Figure 6A). Subjects with allergies also showed increased relative abundance of gut *Proteobacteria*, a phylum previously associated with autoimmune conditions (Kruskal–Wallis, Figure 6B). These differences are detected only in ASD patients and not in controls (Figure 6B). All ASD subjects who had significant gut *Proteobacteria* overgrowth (>30%) also suffered from allergies (4/4), whereas none of the 2 control subjects with *Proteobacteria* overgrowth did (0/2).

We next performed genus level correlation analysis of the oral and gut bacterial relative abundances against allergy status, using bacteria genera that make up at least 0.5% of all OTUs. No salivary or gut genus was significantly and differentially enriched by allergy status, after stratifying by ASD and control group (Table S9).

**Figure 6.** Box plot representation of the relative abundances of oral (**A**–**A"**) and gut (**B**–**B"**) bacterial phyla correlating with the allergy status of the subjects enrolled in this study. (**A**) Oral SR1 relative abundance in all subjects with no allergy and those with allergy; (**A'**) Oral SR1 relative abundance in ASD subjects with no allergy and patients with allergy; (**A"**) Oral SR1 relative abundance in neurotypical subjects with no allergy and neurotypical subjects with allergy. (**B**) Gut *Proteobacteria* relative abundance in all subjects with no allergy and those with allergy; (**B'**) Gut *Proteobacteria* relative abundance in ASD subjects with no allergy and patients with allergy; (**B"**) Gut *Proteobacteria* relative abundance in neurotypical subjects with no allergy and neurotypical subjects with allergy. Box plot representation of the gut alpha diversity (Shannon index) that correlated with the allergy status of the subjects enrolled in this study. (**C**) Gut alpha diversity in all subjects with no constipation and those with constipation; (**C'**) Gut alpha diversity in ASD subjects with no constipation and patients with constipation; (**C"**) Gut alpha diversity in neurotypical subjects with no constipation and neurotypical subjects with constipation. Single asterisk indicates *p* < 0.1 with adjusted FDR > 0.2; double asterisk indicates *p* < 0.05 with adjusted FDR > 0.2, triple asterisk indicates *p* < 0.05 and adjusted FDR < 0.2, Kruskal–Wallis test.

#### *3.8. GI Disturbances*

Patients with autism suffer from many co-occurring GI conditions [55]. Previous studies found that gut microbiome is associated with and may play important roles in GI symptoms such as constipation and abdominal pain [23]. We performed genus level correlation analysis of the gut bacterial relative abundances by constipation and abdominal pain status, using gut genera that make up at least 0.5% of all OTUs. *Roseburia* and *Bacteroides* were differentially enriched in subjects without abdominal pain (Figure 7A, Kruskal–Wallis, Table S10), and this difference in *Roseburia* remained statistically significant after FDR adjustment (pain 2.7% vs. no pain 5.7%). After stratifying by ASD and control subjects, ASD patients without abdominal pain had significantly higher levels of *Bacteroides*, as compared to ASD patients with abdominal pain, whereas control subjects without abdominal pain had lower levels of *Bacteroides*, as compared to control subjects with abdominal pain (Figure 7A, Kruskal–Wallis, Table S10).

**Figure 7.** Bar plot representation of the relative abundances of gut (**A**–**A'**) and oral (**B**–**B'**) bacterial genera correlating with the abdominal status of the subjects enrolled in this study. (**A**) The most differentially abundant gut genera in all subjects with no abdominal pain and those with abdominal pain; (**A'**) The most differentially abundant gut genera in ASD patients with no abdominal pain and patients with abdominal pain. (**B**) The most differentially abundant oral genera in all subjects with no abdominal pain and those with abdominal pain; (**B'**) The most differentially abundant oral genera in ASD patients with no abdominal pain and patients with abdominal pain. Single asterisk indicates *p* < 0.1; double asterisk indicates *p* < 0.05, Kruskal–Wallis test.

Given the concordance between the oral microbiome and upper GI microbiome [26], it is possible that the oral microbiome may be associated with upper GI health and contribute to abdominal pain. We explored phylum and genus levels correlation analysis of the oral bacterial relative abundances between subjects with or without abdominal pain. No oral phylum showed differential enrichment, but several oral genera are differentially enriched based on abdominal pain status, including *Porphyromonas*, *Megasphaera*, *Haemophilus* (Figure 7B, Kruskal–Wallis test, Table S10). Remarkably, *Porphyromonas* is significantly less abundant in subjects without abdominal pain after FDR adjustment (pain 0.7% vs. no pain 2.2%). When stratifying based on ASD status, ASD patients with abdominal

pain showed a higher trend of *Actinomyces*, as compared to ASD patients without abdominal pain (Figure 7B, Kruskal–Wallis test, Table S10).

The gut alpha diversity showed no difference between the constipated and non-constipated group (Figure 6C). When stratifying patients with ASD from the control group, there was an increased trend of gut alpha diversity in constipated ASD patients but not in constipated controls (Figure 6C',C"), consistent with a previous study showing increased gut alpha diversity in functional constipation patients [56].

#### *3.9. Dietary Habits and Gut Microbiome Markers*

Previous studies indicate dietary challenges in ASD patients, but the association between altered dietary patterns with gut dysbiosis has not been explored in ASD patients. We found that ASD patients exhibit a statistically more restricted diet, while finding it more difficult to accept certain foods and try new foods (Mann–Whitney U test, Figure 8A–C). However, no significant differences were found between groups in respect to the amount, rate, interest, environment, or multitasking habits while eating.

**Figure 8.** Box plot representation of abnormal dietary habit severity scores in ASD and control subjects. (**A**) Unwilling to try new foods. (**B**) Diet lacks variety. (**C**) Refuse to eat certain foods. Single asterisk indicates *p* < 0.1; double asterisk indicates *p* < 0.05, Mann–Whitney U test. (**D**) Spearman's correlation matrix between habit scores, allergy/autoimmunity scores, gastrointestinal severity indices (GSI) total score, and selected ASD gut microbiome markers in patients with ASD (results with FDR < 0.05 were shown).

We next assessed correlation between eating habit scores, allergy/autoimmunity scores, GSI total score, and key ASD gut microbiome markers in patients with ASD. Examined gut microbiome markers include Shannon alpha diversity index, gut *Firmicutes*/*Bacteroidetes* ratio and relative abundances of gut butyricimonas, paraprevotella, granulicatella, eubacterium, and cc\_115 genera which showed significant difference between ASD and control groups based on previous grouped or paired analysis (Figure 4). Most notably, we found that ASD individuals uniquely display correlations between gut butyricimonas relative abundance, eating habit total score, and allergy/immune functions (Figure 8D). *Firmicutes*/*Bacteroidetes* ratio is negatively correlated with allergy/immune function while the same trends are not observed in neurotypical controls. Assessed variables lacking significant correlations with gut microbiome markers are not shown.

#### **4. Discussion**

In this cross-sectional study, we conducted a comparative analysis between the gut and oral microbiota of ASD children and that of healthy, first-degree relative co-inhabitant controls. Our study is the first to use a first-degree relative matched subject design combined with high fidelity next generation sequencing technology to investigate the microbiome of ASD individuals. We believe that this study design better controls for variations in genetic background and environmental factors, and therefore has better specificity for detecting ASD-related microbial signatures [23,24]. This paired control scheme has been increasingly used in microbiome studies for diseases that have strong genetic and environmental contributing factors, such as IBD [57].

Our analysis detected differences between ASD and control subjects in both their gut and oral microbiomes. We identified an unspecified oral *Bacilli* genus, the relative abundance of which is significantly different between the ASD and control groups (FDR < 0.05), which has not been described by previous reports [30,31]. Parallel to this observation, amounts of bacteria in the class *Bacilli* were significantly higher in the gut of ASD individuals compared to controls (0.7% vs. 0.4%, Kruskal–Wallis test, *p* < 0.05), consistent with findings of Adams et al. [15]. Previous studies of the gut microbiome have revealed significant increases in facultative anaerobic commensal bacteria belonging to the class *Bacilli* seen in individuals with IBD, supporting a potential connection between *Bacilli* and gut inflammation [58]. It is unknown whether the simultaneous upregulation of *Bacilli* species in the mouth and the gut environment of ASD patients represents any common causal environmental factor (such as diet), or whether overgrowth of *Bacilli* in the mouth could lead to overgrowth of *Bacilli* in the gut. Answers to these questions would help elucidate further the interactions between gut and mouth microbiomes, as well as provide insight into potential ASD pathology.

Consistent with prior reports, ASD patients demonstrated a significantly higher gut *Firmicutes*/ *Bacteroidetes* ratio [59,60], which is a measure associated with inflammatory conditions such as IBD [61,62]. Overgrowth of *Proteobacteria* has been associated with diarrheal diseases, metabolic syndrome and IBD [53], and 4 out of the 6 subjects who exhibited significant *Proteobacteria* overgrowth were ASD patients. *Proteobacteria* overgrowth observed in our study is unlikely due to confounding factors: none of the six subjects were under 5 years-old (age range: 15–45), and none had used antibiotics in the past month. We also explored other putative combined phylum level relative abundance or ratios as dysbiosis markers, which all appear to be abnormal in patients with ASD.

#### *4.1. Microbial Signatures Can Serve as Potential Diagnostic Markers for ASD*

Although oral and gut microbiomes are distinct, we showed that analysis of both can be combined to classify ASD subjects from controls. Among the dysbiosis markers and differentially expressed taxa in the present study, three promising candidates stood out from our analysis: gut *Butyricimonas*, saliva *Parvimonas*, and gut *Firmicutes*/*Bacteroidetes* ratio. In support of our findings is the work done by Kang et al. (2013) which also reported decreased *Butyricimonas* in the gut of ASD patients as compared to controls [24]. *Butyricimonas* is prevalent in healthy individuals and produces butyrate, which has been shown to improve gut health [63]. In addition, recent work on multiple sclerosis suggests that it may play an important role in immune tolerance and prevention against disease pathogenesis and progression [64,65]. *Butyricimonas* had negative correlations with gene expression implicated in cytokine signalling molecules IFN and IL-2, and activation of receptors PPAR and RXR [64]. Given the important association between autoimmune conditions and ASD, it will be important to further explore the role of *Butyricimonas* in the pathogenesis and autoimmune manifestation of ASD patients. Another

study reported thedepletion of oral *Parvimonas* in IBD patients, although this has not been reported in ASD patients [66,67].

Currently, ASD diagnosis is guided by criteria in the DSM-5, which are based solely on clinical symptoms without any objective laboratory measures. Utilizing a combination of gut and oral microbiome signatures could improve the diagnosis and screening process of ASD individuals. This could also identify subclinical or clinical subgroups of ASD patients with potential GI involvement, autoimmunity, or inflammation. Future studies should explore whether these microbiome markers can predict a patient's response to treatment. This would be particularly useful to guide treatment with probiotics or drug options during probiotics therapy and anti-inflammatory interventions, as it could individualize treatment and improve outcomes for patients with ASD.

#### *4.2. Gut and Oral Co-Occurrence Network Reveal Possible Connections between Distinct Microbial Communities*

Our study is the first to co-analyze stool and oral microbiota in patients with ASD. We explored methodologies to investigate the relationship between the oral and gut microbiomes using unbiased approaches. Our analysis revealed novel co-occurrence networks within and between microbial communities that may hold diagnostic significance for ASD. Given how environmental factors (such as diet) can facilitate competitive and cooperative relationships between microbial groups [68], it is possible that such effects can span across distant communities along the digestive tract. The SparCC co-occurrence network analysis revealed an overall denser correlation network of the saliva microbiome compared to the gut. It is known that inter-individual variability of gut microbiota is higher compared with that of salivary microbiota [26], which may explain this observed difference.

Interestingly, some gut and oral taxa show evidence of co-occurrence despite the distal separation. For example, gut *Firmicutes* and saliva *Chloroflexi* showed strong correlation in the ASD population. From a diagnostic perspective, it would be pertinent to explore whether oral *Chloroflexi* can serve as a read-out for the status of gut *Firmicutes* in patients with ASD, thereby using oral microbiome as a more convenient tool to assess dysbiosis of the gut when stool samples are not readily available. More significant oral–gut co-occurrence clusters were observed at the genus level. The oral microbiome may help predict the levels of *Bifidobacteria*, *Escherichia* and *Clostridiales* genera in the gut, which all showed positive correlations with oral genera and are likely correlated with GI and/or ASD pathophysiology [12,69,70].

#### *4.3. Clinical Correlates of ASD Microbiome*

Despite the recognized importance of the gut microbiota in health and disease, our study is the one of the few designed to investigate the relationship between the human microbiota and medical comorbidities of ASD patients. Previously, Plaza-Diaz investigated gut microbiome in ASD patients with or without mental regression and found microbiome signatures associated with different psychiatric subtypes [32]. We analyzed gut alpha diversity, as well oral and gut phylum and genus levels of relative abundance in the context of three common co-occurring medical conditions affecting the ASD individuals: allergies, abdominal pain and poor dietary habits.

We found ASD patients tend to have more unhealthy and restricted dietary habits. This is consistent with previous studies, showing that up to 79% of children with ASD suffer from feeding-related difficulties or nutritional challenges [71] and strong preference for nutrient-poor foods [72]. Given the correlation between severity of poor dietary habits and relative abundances of gut microbiome biomarkers, it is conceivable that the unhealthy dietary habits may be driving gut dysbiosis [73].

Second, we detected a significantly higher prevalence of allergies in ASD patients with than those without. Gut *Proteobacteria* overgrowth is also over-represented in ASD patients and its relative abundance is positively correlated with allergy status. Overgrowth of *Proteobacteria* has been implicated in autoimmune disorders such as IBD [74]. This is opposite to the trend of *Bacteroidetes*, a marker for healthy flora. We also report a negative association between oral *SR1* numbers and allergy status, but this association is only present in ASD patients and not healthy controls.

Little is known about the connection between allergies and autism. In a recent, large population-based, cross-sectional study of data provided by the National Health Interview Survey (NHIS) from 1997–2016, Xu et al. found that children with ASD were more likely to have a food allergy (11.25% versus 4.25%), respiratory allergy (18.73% versus 12.08%), and skin allergy (16.81% versus 9.84%) than neurotypical children. Further, the odds ratio of ASD among children with a food allergy is nearly triple the ratio of ASD among those without a food allergy [75].

The "bi-directional" association between allergies and ASD raises the following questions: (1) whether these dysbiosis markers are simply associated with allergy or whether an abnormal microbiome is involved in the pathogenesis of allergy in ASD, (2) if a pathogenic mechanism could be established, whether ASD patients are more vulnerable to it than neurotypical individuals, and (3) whether there are common underlying mechanisms, potentially involving the dysregulation of the immune system and gut and oral microbiota, that could induce the development of both allergy and ASD. Future studies using animal models, immunology markers, genomics and metabolomics approaches are needed to elucidate the mechanisms of possible causal relationships.

In analyzing the relationship between microbiota and GI pathology, we found significantly higher levels of gut *Roseburia* in subjects without abdominal pain. The genus *Roseburia* consists of obligate Gram-positive anaerobic commensal bacteria that affect one's health in many ways. These bacteria produce short-chain fatty acids such as butyrate, affect colonic motility, maintain the immune response, and contribute anti-inflammatory factors to their environments [76]. Although previous studies have linked *Roseburia* abundance to some disease states such as irritable bowel syndrome and IBD [77], certain species in the genus likely play a positive role in GI health. One recent study found that treatment with the *Roseburia hominis* bacterium provided protection against dextran sulfate sodium (DSS)-induced colitis due to its immunomodulatory properties [78].

Interestingly, the oral genus *Porphyromonas* is significantly more abundant in subjects with abdominal pain. Many members of this genus have been associated with periodontal diseases [79]. The most well-characterized species, *Porphyromonas gingivalis*, has been linked to systemic diseases including upper GI tract inflammation and cancer due to upregulation of systemic cytokine release [80]. Further investigations should consider the mechanistic roles these genera could play in abdominal pain, and whether these gut and oral genera can serve as markers for the diagnosis and treatment monitoring of abdominal symptoms in patients with ASD.

The correlation between abdominal pain status and differential expression of bacterial genera differs between the ASD and control groups. Previously, Strati et al. found that constipation status is correlated with different amounts of bacterial taxa depending on whether an individual has ASD or not [23]. Notably, *Bacteroides* is one genus that shows the most prominent differential patterns: whereas *Bacteroides* appears to be protective against abdominal pain in ASD patients (higher levels are associated with no abdominal pain), the association is the opposite in controls. *Bacteroides* genus harbor species that can have either positive or negative effects on GI health. Some *Bacteroides* species synthesize lipopolysaccharide, an important bacterial virulence factor, and can cause diseases such as GI infection and septicemia in children. Many other *Bacteroides* species can be healthy commensals [81]. A recent meta-analysis concluded that a lower level of *Bacteroides* in the gut microbiota is associated with IBD [82], and functional analysis showed that *Bacteroides* expresses polysaccharide A, which can induce regulatory T-cell growth and cytokine expression to protect against colitis [83]. It is possible that ASD patients may be more prone to positive effects of *Bacteroides* than control subjects, potentially through the action of bacterial metabolites and the gut–brain axis [13]. This is supported by a mouse study which found that administration of *Bacteroides fragilis* corrects gut permeability, alters microbial composition and ameliorates ASD-related defects [13]. The ASD mice also display an altered serum metabolomic profile, and *B. fragilis* modulates levels of several metabolites. Further species-level analysis with higher 16S rRNA gene amplicon sequencing resolution and functional studies could

elucidate the roles of different Bacteroides species on abdominal pain in ASD subjects. Future studies should also investigate the relationships between abdominal pain, *Bacteroides* abundance, and the severity of ASD symptoms.

#### **5. Conclusions**

In conclusion, our study is the first to use a first degree-relative matched design combined with high fidelity 16S rRNA gene amplicon sequencing technology to characterize the microbiome of patients with ASD compared to neurotypical individuals. To our knowledge, this study is the first to co-analyze the oral and gut microbiomes in patients with ASD, as well as explore the relationship between the two microbial communities and clinical indices. This study identified distinct features of gut and salivary microbiota that differ between individuals with and without an ASD diagnosis. The diagnosis of ASD currently relies on psychological testing with potential high subjectivity and inconsistencies. We suggest improvement of current diagnostic approaches based on gut and oral microbial signatures and co-occurrence networks. Given the emerging role that the human microbiome plays in systemic diseases, we hope that these analyses will provide clues for developing microbial markers for diagnosing ASD and comorbid conditions, and to guide treatment. In particular, ASD patients have disproportional gastrointestinal symptoms compared to neurotypical individuals. Therefore, developing "gut microbiome markers" is particularly important for monitoring GI health or guiding interventions of the gut. For example, these preliminary results can serve as a starting point to test whether changing the microbiome (e.g., with probiotics) would improve co-morbid conditions in patients with ASD and further modify the core and GI symptoms of ASD.

The explorations of causal relationships between microbiomes, ASD status and co-morbidities await future investigations. Further research could explore metabolomics profiles to characterize microbiome-related inflammatory factors and metabolites in the oral and gut cavity such as interleukins and short-chain fatty acids. Other areas of future study should include exploring the role of microbiota in inflammatory conditions such as allergy and autoimmunity, investigating their genetic and/or epigenetic linkage, researching mechanism of the gut–brain axis and relevant neural circuits, and ultimately inquiring more about the pathogenesis of ASD. These indices and studies will improve the algorithm for ASD screening, diagnosis, and treatment monitoring in the future.

Limitations of the current study include: (1) The use of both sibling and parental controls, where age could contribute to the large inter-individual variability. Future studies should focus on only age-matched sibling controls, if possible. (2) The small sample size, which likely contributed to high FDR in the majority of our analyses and the difficulty in distinguishing true differences from noise. Verification of our findings with a larger cohort is required. The current study was not sufficiently powered for detecting clinically relevant biomarkers. However, with the methodologies in hand, we will be able to expand the study to develop clinically biomarkers in the future. That being said, even with the relatively small sample size, we were able to find biomarkers that have withstood rigorous statistical testing and adjustment. (3) Our genus level differential expression patterns showed discrepancies from previous reports that used neurotypical controls [12,23], but this likely reflects the differences in study design [12]. For example, we did not detect changes in *Prevotella*, *Bacteroides*, *Clostridium* cluster I/II, or *Lactobacillus*, which have been reported by some studies to be differentially expressed between ASD and control groups [12], but previous studies using sibling-matched designs also did not detect these differences [84–86].

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6643/11/9/2128/s1. Figure S1. Saliva and gut microbiome OTU-level relative abundance, all subjects. Figure S2. Box plots of saliva and stool alpha diversity by groups (saliva\_ASD, saliva\_control, stool\_ASD, stool\_control), by ASD-control pairs, and by individual subjects. Figure S3. PCA beta diversity plots based on Bray–Curtis, Unweighted Unifrac, and Weighted Unifrac dissimilarity index. Figure S4. Differential abundances of saliva and gut microbiome in ASD and control subjects at phylum level, shown as fold changes in mean relative abundances. Figure S5. ANCOM volcano plots of saliva vs. gut microbiome OTU level differential expression in control and ASD subjects, respectively. Figure S6. Pearson's correlations of salivary Actinobacteria, Bacteroidetes, Firmicutes

and Proteobacteria co-occurrence cluster identified from SparCC co-occurrence analysis. Figure S7. Gut and oral microbiome genus level SparCC co-occurrence network (Corr > 0.3 and Corr > 0.7). Figure S8. Correlation between gut and oral alpha diversity (Shannon index). Table S1. Metadata, subject characteristics and clinical indices. Table S2. Salivary and gut phylum and genus level OTU table. (XLS, 195KB). Table S3. Salivary and gut alpha Diversity raw data and statistics (Kruskal–Wallis test). Table S4. Statistics of beta diversity indices (PERMANOVA). Table S5. Stool and saliva phylum and genus level differential relative abundances (group means, *p* values, FDR adjustment with BH method, Kruskal–Wallis test). Table S6. Saliva and gut genus level paired Wilcoxon test, ASD vs. Control subjects (Wilcoxon's F statistic, *p* values, FDR adjustment with BH method, Paired Wilcoxon's test). Table S7: Differential abundances by ANCOM (ASD: saliva vs. stool, Control: saliva vs. stool, Saliva: ASD vs. Control, Stool: ASD vs. control). Table S8: ROC curve AUC and test statistics for significant biomarkers that distinguish ASD and control groups. Table S9: Saliva and stool genus level relative abundances based on allergy status using genera with at least 0.5% mean relative abundances (group means, *p* values, FDR adjustment with BH method, Kruskal–Wallis test). Table S10: Saliva and stool genus level relative abundances based on abdominal pain status using genera with at least 0.5% mean relative abundances (group means, *p* values, FDR adjustment with BH method, Kruskal–Wallis test).

**Author Contributions:** Author Contributions: X.K. devised the project and the main conceptual ideas. X.K. and J.L. worked out almost all of the technical details, and X.K., J.L., M.K., H.H., A.A., P.H. and J.Z. performed data collections and measurements. J.L., R.S., and M.C. conceptualized and performed bioinformatics analyses and statistical analyses, with assistance from H.R. and input from X.K. X.K. and J.L. wrote the manuscript, with assistance from M.K.,K.L, and input from A.B.O. J.L. drafted the figures with input from R.S., M.C., and A.B.O., and K.L. X.K., J.L., C.H. and M.K. participated in manuscript editing.

**Funding:** This study was funded by Massachusetts General Hospital grant number 230361 at the Athinoula A. Martinos Center for Biomedical Imaging, Boston, MA.

**Acknowledgments:** This work was conducted with support from Harvard Catalyst|The Harvard Clinical and Translational Science Center (National Center for Advancing Translational Sciences, National Institutes of Health Award UL 1TR002541) and financial contributions from Harvard University and its affiliated academic healthcare centers. We thank Georg Gerber for extensive feedback on methodology, Dong Kong and Zuoteng Yu for assistance with manuscript preparation. We thank Bruce Paster and Hattice Hasturk for advice on study design.

**Conflicts of Interest:** Author Hugh Russell was employed by company Precidiag INC. Xuejun Kong served as a short term consultant for Precidiag more than two years ago. All other authors declare no competing interests. The funding agencies and Precidiag INC have no role in the study design, implementation, the interpretation of the results.

#### **Nomenclature**


#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### **E**ff**ects of Regular Kefir Consumption on Gut Microbiota in Patients with Metabolic Syndrome: A Parallel-Group, Randomized, Controlled Study**

**Ezgi BELLIKCI-KOYU 1,2, Banu Pınar SARER-YUREKLI 3, Yakut AKYON 4, Fadime AYDIN-KOSE 5, Cem KARAGOZLU 6, Ahmet Gokhan OZGEN 3, Annika BRINKMANN 7, Andreas NITSCHE 7, Koray ERGUNAY 4, Engin YILMAZ <sup>8</sup> and Zehra BUYUKTUNCER 1,\***


Received: 13 August 2019; Accepted: 22 August 2019; Published: 4 September 2019

**Abstract:** Several health-promoting effects of kefir have been suggested, however, there is limited evidence for its potential effect on gut microbiota in metabolic syndrome This study aimed to investigate the effects of regular kefir consumption on gut microbiota composition, and their relation with the components of metabolic syndrome. In a parallel-group, randomized, controlled clinical trial setting, patients with metabolic syndrome were randomized to receive 180 mL/day kefir (*n* = 12) or unfermented milk (*n* = 10) for 12 weeks. Anthropometrical measurements, blood samples, blood pressure measurements, and fecal samples were taken at the beginning and end of the study. Fasting insulin, HOMA-IR, TNF-α, IFN-γ, and systolic and diastolic blood pressure showed a significant decrease by the intervention of kefir (*p* ≤ 0.05, for each). However, no significant difference was obtained between the kefir and unfermented milk groups (*p* > 0.05 for each). Gut microbiota analysis showed that regular kefir consumption resulted in a significant increase only in the relative abundance of *Actinobacteria* (*p* = 0.023). No significant change in the relative abundance of *Bacteroidetes, Proteobacteria or Verrucomicrobia* by kefir consumption was obtained. Furthermore, the changes in the relative abundance of sub-phylum bacterial populations did not differ significantly between the groups (*p* > 0.05, for each). Kefir supplementation had favorable effects on some of the metabolic syndrome parameters, however, further investigation is needed to understand its effect on gut microbiota composition.

**Keywords:** kefir; gut microbiota; metabolic syndrome

#### **1. Introduction**

Metabolic syndrome (MetS) is a pathologic condition that includes abdominal obesity, insulin resistance, dyslipidemia, and arterial hypertension [1]. Each component of MetS is known as a risk factor for the development of type 2 diabetes and cardiovascular diseases. It was found that the risk of type 2 diabetes was five times, the risk of cardiovascular disease was two times, and the risk of death

was one-half times higher in individuals with MetS compared to those without the syndrome. Due to its high prevalence and related health problems, the MetS is currently considered as a significant public health problem [1,2].

MetS has a multi-factorial etiology comprising complex interactions between genetic predispositions and environmental factors including diet, physical activity, and other lifestyle factors [3,4]. Since Turnbaugh et al. showed the linked between gut microbiota and obesity, there has been growing evidence that suggests a causal relationship between gut microbiota and the components of MetS [5]. Primary, the low-grade chronic inflammation state in MetS has been explained by the metabolic endotoxemia that was a result of gut dysbiosis [6,7]. Most of the animal and human studies have reported that obesity and insulin resistance are associated with an altered ratio of *Firmicutes* and *Bacteroidetes* [8,9]. In addition to the effects on immune function, the gut microbiota also exerts its role through the influence on host energy metabolism and gut barrier integrity [10,11]. Therefore, the gut microbiota has been suggested as a potential target to modify the risk factors that contribute to conditions of MetS.

The modification of diet using prebiotics and probiotics has been suggested as a useful strategy to improve metabolic health via the modulation of gut microbiota. Although the effects of probiotic and prebiotic supplementation on metabolic health have been examined in previous studies, the results are inconsistent due to the choice of probiotic strain, formulation of the probiotic, outcome of interest, and duration of the intervention [12–14]. Furthermore, ingestion of probiotics through traditional fermented foods has not been widely examined in terms of their efficiency on MetS components. Kefir is a fermented milk product, traditionally produced with kefir grains that have a specific combination of bacteria and yeasts [15,16]. Microbial composition of kefir varies depending upon the type of kefir grains, the type and composition of milk, culture medium, fermentation period and temperature, and also storage conditions [17]. *Lactobacillus*, *Lactococcus, Streptococcus Leuconostoc,* and acetic acid bacteria are the most common bacteria; and *Saccharomyces*, *Kluyveromyces,* and *Candida* species are mostly found yeasts in kefir [18]. Animal studies have suggested that kefir has anticarcinogenic, antimicrobial, anti-inflammatory activities, and thus may ameliorate MetS components [19–23]. However, there is still limited clinical evidence for its potential effects on MetS patients. To our knowledge, especially, the effects of kefir on MetS components via the modulation of gut microbiota have not been examined widely in clinical settings. To address the research gap, this study aimed to investigate the effects of daily kefir consumption on gut microbiota composition and their relation with the components of metabolic syndrome in adults with MetS.

#### **2. Materials and Methods**

#### *2.1. Subjects*

Subjects with MetS, aged 18–65 years, were recruited from the outpatient clinic of the Department of Endocrinology and Metabolism at the Ege University, Izmir, Turkey. MetS was diagnosed using the IDF-2005 guidelines [24]. The eligibility of a subject was confirmed following a physical examination by the research endocrinologist and a nutritional assessment by the research dietitian in the screening period. Adults were excluded if they (1) were using antibiotics in the past 1 month or during the intervention period, (2) were using dietary supplements (probiotic, prebiotic, or symbiotic) during the past three months or during the intervention period, (3) were pregnant or lactating, (4) had severe liver, kidney, heart, or immune deficiency, (5) had chronic gastrointestinal system diseases, type 1 diabetes or cancer, (6) had allergy to the dairy products or lactose intolerance, (7) were currently taking prescribed drugs that can modulate lipid profile or glycaemic control, and (8) did not comply with the consumption of test drinks.

The compliance was assessed by interviewing the participants and reviewing the record of their consumption in each visit. Non-compliance was defined as consuming < 80% of the scheduled serving during the study period.

The study protocol was approved by the Ethics Committee of Clinical Research at Ege University Faculty of Medicine (15-2.1/14) and registered at clinicaltrials.gov (NCT03966846). All procedures were performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants.

#### *2.2. Study Design*

A parallel-group, randomized, controlled clinical trial was performed. A total of 40 eligible participants were randomized, and 20 participants in each group were allocated to intervention. Five participants in the kefir group and four participants in the unfermented milk group did not receive the allocated intervention due to medical conditions and not providing the fecal samples, and also three participants in the kefir group and six participants in the unfermented milk group discontinued the intervention due to taking antibiotics and declined consent. Therefore, the study was completed with 22 participants and an allocation ratio of 55%. The recruitment and follow up of participants were conducted between March 2015–July 2017. Participants were randomized into two groups (kefir group and unfermented milk group as control) by the research physicians using a stratified block randomization method. The random allocation sequence was provided by the Department of Biostatistics, Hacettepe University. Participants visited the research center 5 times in total. The first visit included the screening of individuals in terms of inclusion and exclusion criteria. The second visit (Week 0) included recording general characteristics, medical history, and lifestyle behaviors of participants, assessing the nutritional status of participants using 24-h dietary recall and anthropometrical measurements, collecting the initial blood and fecal samples, measuring the blood pressure, and also proving information about the consumption and storage of test drinks. The third (Week 4) and forth (Week 8) visits included the assessment of the compliance in terms of consumption of test drinks and dietary intake. The fifth visit (Week 12) included the assessment of the nutritional status of participants using 24-h dietary recall and anthropometrical measurements, collection of the final blood and fecal samples, and the measurement of blood pressure (Figure 1).

The primary outcome of the study was the change in the relative abundance of microorganisms in gut microbiota by regular kefir consumption. The potential correlations between the changes in dietary intake, anthropometrical measurements, biochemical parameters, or blood pressure and the change in microbiota composition were all secondary outcomes.

**Figure 1.** Timeline of the study.

#### *2.3. Intervention*

During a 12-week intervention period, kefir group (*n* = 12) received kefir (180 mL/day) while control group (*n* = 10) received unfermented milk (180 mL/day) regularly. Participants were asked to maintain their habitual diet and physical activity. Additional products that contain probiotics were not allowed during the intervention period. No dietary supplement use was recorded before or during the study.

#### *2.4. Test Drinks*

Two dairy products (kefir and unfermented milk) were tested in parallel groups. Kefir was prepared using the culture of DC1500I (Danisco, Olsztzyn, Poland) containing *Lactococcus lactis* subsp. *lactis, Lactococcus lactis* subsp. *cremoris, Lactococcus lactis* subsp. *diacetylactis, Leuconostoc mesenteroides* subsp. *cremoris, Lactobacillus kefir, Kluyveromyces marxianus*, and *Saccharomyces unisporus* at Ege University Faculty of Agriculture, Department of Dairy Technology. Kefir was derived from the full-fat (3.5%) homogenized and pasteurized (at 85 ◦C) milk that was used as a control drink at the same time. The beverages were distributed and stored at 4 ◦C. The test drinks were received to participants twice a week, and they consumed the test drinks between 1 and 4 days of post-production.

#### *2.5. Dietary Assessment*

Dietary intake was assessed using 24-h dietary recall method by research dietician in each visit. A photographic atlas of food portion sizes was used to clarify the amounts of food items consumed. Dietary energy, macro- and micronutrient intakes were analyzed using BeBIS software (Ebispro for Windows, Stuttgart, Germany; Turkish Version BeBIS, Nutrition Information System, Version 8).

#### *2.6. Anthropometrical Measurements*

Body weight and composition (fat mass and fat-free mass) were measured by Tanita BC418 (USA), and height was measured by a calibrated stadiometer (Nan Tartı, TR). Body Mass Index (BMI) was calculated by dividing body weight (in kilograms) by the square of height (in meter). The waist circumference was measured at the midpoint between the lower ribs and the iliac crest, and hip circumference was measured horizontal at the largest circumference of hip. Waist-to-hip ratio (WHR) was calculated.

#### *2.7. Biochemical Analysis and Blood Pressure*

Venous blood samples were drawn after a 10-h overnight fasting excluding only the water at the Visit 1, Visit 2 (Week 0), and Visit 5 (Week 12). Serum glucose, insulin, HbA1c, total cholesterol, high-density lipoprotein (HDL-C), low-density lipoprotein (LDL-C), triglycerides, homocysteine, high-sensitivity C-reactive protein (hs-CRP), alanine aminotransferase (ALT), aspartate aminotransferase (AST), and gamma-glutamyl transferase (GGT) were analyzed at Ege University, Hospital of Medical School, Laboratory of Clinical Biochemistry. All biomarkers were analyzed using routine methods by Roche/Cobas analyzer series. Serum concentrations of tumor necrosis factor-α (TNF-α), interleukin-6 (IL-6), interleukin-10 (IL-10) and interferon-gamma (IFN-γ) were determined by enzyme-linked immunosorbent assay (ELISA) using standard kits, and the analyses were conducted as described by the manufacturer (DIAsource ImmunoAssays, Louvain-la-Neuve, Belgium). Insulin resistance was assessed using Homeostatic Model Assessment (HOMA-IR) model calculated with the equation of "the fasting insulin level (μU/L) × fasting plasma glucose (mg/dL)/405". Systolic (SBP) and diastolic blood pressure (DBP) were measured at the brachial artery of right upper arm after 15 min rest. Both blood pressures were measured twice at 5-min intervals and recorded on average.

#### *2.8. Specimen Processing, 16S rRNA Amplification and Sequencing*

Fecal samples from individuals enrolled in the study were collected in sterile containers and kept frozen at −80 ◦C. A sterile spatula was used to obtain 4–5 pieces of frozen chunks from the surface and internal portions of the specimen. They were combined for a 150–200 mg total weight for each and mixed by vortexing. Following a bead-beated step described by Tomas et al. and Wu et al. [25,26], DNA was extracted using Qiagen Stool Mini Kit (Qiagen, Hilden, Germany) as directed by the manufacturer. DNA amount of 50 ng/μL was prepared for each specimen, using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Hennigsdorf, Germany).

The 16S rRNA sequences were amplified using previously-described primers targeting the V3-V4 region, frequently used to study bacterial diversity [27], with Illumina adapter overhang sequences added, as directed by the manufacturer. Attachment of sequencing adapters to PCR products, amplification and library preparation were performed using the Nextera XT Index and Nextera DNA Library Prep kits (Illumina, San Diego, CA, United States), as suggested by the manufacturer. Product clean-up, library quantification, and optimization were carried out using the Agencourt AMPure XP reagent (Beckman Coulter Biosciences, Krefeld, Germany) standard protocol and a Qubit 2.0 Fluorometer (Thermo Fisher Scientific, Rochester, NY, USA). The sequencing runs were performed in an Illumina MiSeq sequencer (Illumina Inc., New York, USA).

#### *2.9. Data Handling, Phylogenetic and Statistical Analyses*

The raw sequencing data were de-multiplexed and extracted in fastq format. Sequence data handling and taxonomic assignment were carried out using Geneious v11.1 (Biomatters Ltd., Auckland, New Zealand), MALT V0.3.8 and MEGAN v6.11 [28]. Trimming for read quality and length and adaptor sequence removal were performed using Trimmomatic v0.35 [29,30]. Trimmed reads were

mapped to the NCBI-NT RefSeq 16S database via MALT V0.3.8, with hits down to 95% identity. For the operational taxonomic unit (OTU) identification and taxonomic binning, LCA-assignment algorithm (with 95% minimum identity) and 16s percent identity filter (species assignment at 99% identity) were employed. Relative bacterial abundance on the genus and species levels were calculated using the reads numbers of the corresponding OTUs.

Various alpha and beta diversity metrics were calculated for bacterial diversity and composition analyses. For this purpose, raw data were imported into QIIME2 [31], filtered and controlled for quality and chimeric sequences using DADA2, q2-demux, and dblur scripts [32,33]. The trimmed reads were subsequently mapped to the GreenGenes [34] and SILVA [35] databases for OTU identification and taxonomic binning. Faith phylogenetic diversity (PD), Pielou, Shannon, and Jaccard indices and Bray–Curtis and UniFrac distances were computed and evaluated using Kruskal–Wallis, Spearman, or permutational multivariate analysis of variance (PERMANOVA) tests as appropriate.

Data were analyzed using the Statistical Package for the Social Sciences (SPSS), version 22.0. (SPSS Inc., Chicago, IL, USA). Data normality was tested by Shapiro–Wilk test prior to further analyses. Kruskal–Wallis and Mann–Whitney U tests were employed for comparisons among groups where appropriate. Spearman's rank-order correlation was used to analyze the correlation analysis between microbial taxa and biochemical and blood pressure measurements. A value of *p* < 0.05 was considered as significant.

#### **3. Results**

Baseline characteristics of the participants were summarized in Table 1. There were no differences in terms of age, dietary intake, anthropometrical measurements, and biochemical parameters except serum insulin levels between groups.


**Table 1.** Baseline characteristics of kefir and unfermented milk groups.


**Table 1.** *Cont*.

Data are given as median (25th percentile–75th percentile). Mann–Whitney U test was used to compare differences between groups. Fisher's exact test was used to compare gender between groups. HDL: High Density Lipoprotein, LDL: Low Density Lipoprotein, HbA1c: Glycosylated Hemoglobin, HOMA-IR: Homeostasis Model of Assessment Insulin Resistance, hs-CRP: High-sensitivity C-reactive Protein, TNF: Tumor Necrosis Factor, IL: Interleukin, IFN: Interferon, ALT: Alanine Aminotransferase, AST: Aspartate Aminotransferase, GGT: Gamma-glutamyl Transferase. \* *p* < 0.05.

The changes in dietary intake, anthropometrical measurements, biochemical parameters, and blood pressure during the intervention period were given in Table 2. Intakes of energy and macronutrients did not change significantly during the intervention period in either kefir or unfermented milk groups (*p* > 0.05, for each). In terms of anthropometrical measurements, body weight and fat mass showed slight reductions after the 12-weeks intervention of kefir compared to unfermented milk, however, the changes in any of anthropometrical measurement from baseline to after the intervention did not differ significantly between the groups (*p* > 0.05, for each). Among the biochemical biomarkers, almost all parameters of lipid profile and glycaemic status showed amelioration by the intervention kefir group, however, only the difference in fasting insulin and thereby HOMA-IR from baseline to after intervention was significant (*p* = 0.050). Furthermore, TNF-α and IFN-γ showed a significant decrease after the intervention of kefir (*p* = 0.015 and *p* = 0.013, respectively), whereas IL-6 showed a slightly larger decrease in the unfermented milk group (*p* = 0.047). Both systolic blood pressure and diastolic blood pressure decreased significantly after the intervention in the kefir group (respectively *p* = 0.041 and *p* = 0.019), while only systolic blood pressure showed a modest decrease in the unfermented milk group (*p* = 0.047).


*Nutrients* **2019**

, *11*, 2089



#### *Nutrients* **2019** , *11*, 2089

**Table 2.** *Cont*.

In regard to the analysis of gut microbiota composition, the mean number of total reads per sample was 90627 (Standart Deviation (SD): 44912; range: 31198–183068) at baseline and 118025 (SD:38831; range 39248–171915) after the intervention in kefir group, while it was recorded as 138775 (SD: 29961, range: 100922–195456) at the baseline and 95058 (SD: 23740, range: 55853–124889) after the intervention in the unfermented milk group. The gut microbiome of the participants was composed of phyla *Bacteroidetes* (51%), *Firmicutes* (30%), *Proteobacteria* (11%), *Verrucomicrobia* (0.02%) and *Actinobacteria* (0.003%) at the baseline in kefir group; the relative abundance of these phyla were detected as 39%, 39%, 10%, 0.03%, and 0.04% respectively after the intervention of kefir. Only the increase in the relative abundance of *Actinobacteria* was found to be statistically significant (*p* = 0.023). In the unfermented milk group, the phyla *Bacteroidetes* (66%), *Firmicutes* (27%), *Verrucomicrobia* (0.01%), *Actinobacteria* (0.01%), and *Proteobacteria* (0.03%) were detected at the baseline, however, the relative abundance of these phyla were changed to 33%, 56%, 0.03%, 0.04%, and 0.02% respectively after the intervention (Figure 2). When the changes in the relative abundance of each phyla distribution from after the intervention to baseline were compared between the groups, no significant difference was obtained (*p* > 0.05, for each). In the kefir group, the median *of Firmicutes*/*Bacteroidetes* ratio was 0.62 (range:0.06–10.01) at the baseline, and 1.77 (range:0.14–46.16) after the intervention (*p* = 0.388). This ratio was 0.30 (range:0.03–25.27) and 2.22 (range:0.34–12.04) respectively at baseline and end of the intervention in the unfermented milk group (*p* = 0.333). No significant difference was obtained between the changes in *Firmicutes* to *Bacteroidetes* ratio of groups (*p* > 0.05).

**Figure 2.** Gut microbiota composition before and after the intervention in each group.

The changes in the relative abundance of *Bacteriodetes* and *Firmucutes* by the consumption of test drinks were given in Figure 3. The phyla *Bacteroidetes* was composed of five dominant genera; *Bacteroides*, *Odoribacteraceae*, *Porphyromonadaceae*, *Prevotellaceae*, and *Alistipes* in the gut microbiome of participants. *Bacteroides* (54% in kefir group and 44% in unfermented milk group), *Prevotellacea* (26% in kefir group and 40% in unfermented milk group) and *Alistipes* (11% in kefir group and 7% in unfermented milk group) were respectively most abundant families among *Bacteroidetes* at baseline. The relative abundance of *Bacteroides* changed to 59% in kefir group and 50% in unfermented milk group; *Prevotellacea* changed to 25% in kefir group and 29% in unfermented milk group; *Alistipes* changed to 9% in kefir group and 10% in unfermented milk group after the intervention. Despite the modest changes in the relative abundance of some genera, no significant difference was obtained when the changes in the relative abundances were compared between groups (*p* > 0.05, for each). Among *Firmicutes*, *Clostridia*, *Erysipelotrichaceae*, *Veillonellaceae,* and *Lactobacillales* were obtained in the fecal samples of participants. Although an increase in the relative abundance of *Clostridia* (from 73% to 85%) and *Lactobacillales* (2% to 5%), and also a decrease in the relative abundance of *Veillonellaceae* (from 9% to 6%) were obtained from baseline to after the intervention in kefir group, none of these changes were statistically significant (*p* > 0.05, for each). In the unfermented milk group, the relative abundance of *Clostridia* was increased from 75% to 89%, whereas *Lactobacillales* (5% to 2%) and *Veillonellaceae* (from 5% to 4%) were decreased from baseline to after the intervention. Similar to the phyla *Bacteriodetes*, no significant difference was obtained in the changes of the relative abundance of *Firmicutes* at genus level between the kefir and unfermented milk groups (*p* > 0.05, for each) (Figure 3). Among the phyla *Actinobacteria*, the relative abundance of *Bifidobacterium* was increased from 31% to 39% by the intervention of kefir, and from 23% to 32% by the intervention of unfermented milk. However, these changes were not found significant when compared between the groups (*p* > 0.05). Furthermore, *Bifidobacterium* species were detectable in only 50% of participants' the fecal samples at the baseline, whereas they could be detected in 91.7% after the intervention of kefir (data not shown). On the contrary, *Verrucomicrobia* was obtained less frequently from baseline (75% of participants) to after the intervention (58.3% of participants) in the kefir group.

The correlation between the change in gut microbiota composition and the change in physiological characteristics, including anthropometrical measurements, biochemical markers, or blood pressure were conducted to examine the potential associations. Correlations between the changes in anthropometrical measurements and fecal microbiota composition at phylum and subphylum level were summarized in Table 3. The body weight and BMI were positively correlated with the relative abundance of *Firmicutes* and *Proteobacteria.* However, they were negatively correlated with the relative abundance of *Clostridia* (*p* < 0.05, for each). The body fat mass was negatively correlated with the relative abundance of *Bacteroidetes* (*p* < 0.01), and positively correlated with the relative abundance of *Porphyromonadaceae, Firmicutes*, and *Actinobacteria* (*p* < 0.05, for each). The waist circumference was negatively correlated with the relative abundance of *Clostridia* and positively correlated with the relative abundance *Veillonellaceae* (*p* < 0.05, for each).

Table 4 summarizes the correlation between the change in gut microbiota composition and biochemical markers. The changes in the relative abundance of *Bacteroides* was negatively correlated with both total and LDL cholesterol while the changes in the relative abundance of *Veillonellaceae* was negatively correlated with only LDL cholesterol (*p* < 0.05, for each). The change in the relative abundance of *Bacteroidetes* was negatively, and the change in the relative abundance of *Odoribacteraceae* and *Alistipes* groups was positively correlated with the change in serum glucose. The change in the relative abundance of *Verrucomicrobia* was positively correlated with changes in serum homocysteine and insulin (*p* < 0.05, for each). When we analyzed the correlation between the change in gut microbiota composition and blood pressure in the phylum level, the change in the relative abundance of *Actinobacteria* and *Proteobacteria* was positive, whereas *Bacteroidetes* was negatively correlated with the change in blood pressure. In the sub–phylum level, the change in the relative abundance *Lactobacillales* was positively correlated with the change in systolic and diastolic blood pressure (*p* < 0.05).

We further assessed several alpha and beta diversity metrics to assess bacterial biodiversity in specimens collected in week 0 and 12 from kefir and unfermented milk groups. No statistically significant differences were observed in OTU counts (Figure 4). Alpha diversity indices, indicating species richness and evenness with/without phylogenetic relations; namely, Shannon, Jaccard, and Faith PD indices, were similar between the study groups (Figure 5) (Jaccard plots not provided). Biodiversity between study cohorts, assessed by Bray–Curtis and weighted/unweighted UniFrac distances revealed no significant variation among study groups or in different time points. The PCoA plots of the unweighted UniFrac distances were given in Figure 5. No differences in OTU counts, alpha or beta diversity measures were observed when specimens from week 0 and 12 were assessed, regardless of the study group.

No side effect was reported by the participants during or after the intervention period that included the consumption of kefir.

**Figure 3.** Bacterial changes in the relative abundance of Bacteriodetes (**a**) and Firmucutes (**b**) by the consumption of test drinks.


#### *Nutrients* **2019** , *11*, 2089


**4.**Correlationsbetweenchangesinbiochemicalmarkersandfecalmicrobiotacompositionatphylumandsub–phylumlevel

C-reactive Protein, TNF: Tumor Necrosis Factor, IL: Interleukin, IFN: Interferon, ALT: Alanine

Aminotransferase,

 AST: Aspartate

Aminotransferase,

 GGT:

Gamma-glutamyl

 Transferase.

**Figure**whiskers indicate the 90th and 10th percentiles (**A**: Observed OTUs, **B**: Shannon index, **C**: Faith phylogenetic diversity index).

**Figure 5.** Principal coordinate analysis (PCoA) plot of the unweighted (**A**) and weighted (**B**) UniFrac distance matrices in the study groups. The plots were generated using EMPeror [36]. Axis titles indicate the percentage variations. The colors indicate sampling time (red: week 0, blue: week 12).

#### **4. Discussion**

In this parallel–group randomized controlled study, regular kefir consumption during 12 weeks provided some improvements in anthropometrical measurements, lipid profile, glycaemic status, and inflammation in participants with MetS. In particular, insulin and HOMA–IR levels were significantly decreased, and also pro–inflammatory cytokines (TNF–α and IFN–γ) and blood pressure were ameliorated by kefir consumption. However, the magnitude of the improvements stayed insignificant when compared to unfermented milk. The effects of kefir on metabolic status were previously investigated in both animal models and human studies [37–39]. Some animal models suggested that kefir might have a potential to benefit the management of MetS by reducing body weight, fasting blood glucose, insulin, total, and LDL cholesterol, triacylglycerol, and pro–inflammatory cytokines, including IL–1β and IL–6 [38]. However, the evidence from human studies has been controversial. For instance, Ostadrahimi et al. reported that consumption of 600 mL/d kefir containing *Lactobacillus casei*, *Lactobacillus acidophilus*, and *Bifidobacteria* species had beneficial effects on fasting blood glucose and HbA1c compared to the control drink in patients with type 2 diabetes [37]. On the other hand, St–Onge et al. showed that 500 mL/day of kefir consumption for four weeks had no effect on lipid profile [40]. Furthermore, Fathi et al. showed that two servings of kefir in a day during eight weeks led a similar improvement both in lipid profile and weight management compared with milk [41,42]. The variation in response to the kefir consumption could be mainly explained by the variation of kefir composition, and the characteristics of study samples in different studies. Many different bacteria and yeast might be used for kefir production, and this might lead to distinct effects on metabolism and gut microbiota. Kefir used in this study contained *Lactococcus lactis* subsp. *lactis, Lactococcus lactis* subsp. *cremoris, Lactococcus lactis* subsp. *diacetylactis, Leuconostoc mesenteroides* subsp. *cremoris, Lactobacillus kefir, Kluyveromyces marxianus*, and *Saccharomyces unisporus,* and differed from the kefir samples used in other studies [37,38]. Furthermore, the initial metabolic profile of the participants was suggested an essential factor for the efficacy of probiotic interventions. Fuentes et al. showed that probiotics are more effective in patients with high baseline total cholesterol levels (251–300 mg/dL) compared to the patients with low baseline total cholesterol levels (200–250 mg/dL) [43]. Similarly, Nikbakht found

out that probiotic supplementation was only effective in patients with baseline fasting blood glucose level above 126 mg/dL [44]. In our study, the median of total cholesterol levels was 243.50 mg/dL and 220.00 mg/dL, and the baseline glucose levels were 105.00 mg/dL and 101.50 mg/dL for kefir group and unfermented milk group, respectively. This may partly explain the lack of efficacy of kefir on metabolic status in our study.

Alterations in gut microbiota diversity, composition, and function were suggested to play a significant role in the development of MetS [12]. Ameliorating the intestinal dysbiosis with prebiotics and probiotics have gained considerable attention in recent years for the management of MetS [45]. However, the studies have yielded inconsistent results regarding the influence of probiotics on fecal microbial diversity and composition [46–52]. Furthermore, the effects of kefir as a probiotic on gut microbiota have been examined very limited and mainly with animal studies. Kim et al. revealed that three-week oral administration of kefir provided a decrease in the number of *Firmicutes*, *Proteobacteria,* and *Enterobacteriaceae*, and an increase in the number of *Bacteroidetes*, *Lactobacillus*, *Lactococcus,* and total yeast compared to milk group in mice [53]. However, in their follow–up study, no significant difference apart from the increase in *Lactobacillus*/*Lactococcus* populations was observed in the kefir group compared to control [39]. Similarly, an increase in *Lactobacillus* and *Bifidobacterium* populations and a reduction in *Clostridium* populations by consumption of kefir have been reported in mice previously [54–56]. The present study is one of the first reports showing the impact of kefir on human microbiota composition in patients with MetS. In this study, regular kefir and milk consumption for 12 weeks resulted in some alterations in the gut microbiota composition. For instance, *Lactobacillus* and *Bifidobacterium spp.* were increased by kefir consumption. However, apart from the increase in the relative abundance of *Actinobacteria*, no significant change by kefir consumption was recorded. Furthermore, the changes in the relative abundance of bacterial populations did not differ significantly between the groups. In some studies, following the probiotic supplementation, increases in the supplemented genera without an additional impact on the main microbial groups were observed [57,58]. Previously, Yılmaz et al. showed that 400 mL/day kefir consumption for four weeks in patients with inflammatory bowel diseases resulted in the significant increase of Lactobacillus bacterial load in feces [59]. Our study also showed an increase from 2% to 5% in the relative abundance of *Lactobacillales* by kefir consumption, albeit lacking statistical significance. It has suggested that the change of the microbiota composition may be related to several factors such as age, gender, initial microbiota composition, dietary intake, lifestyle factors, menopausal status, and medical therapy of the individuals [60,61]. Moreover, the microbiota composition of the product (kefir) that was tested [17,18] and the consumption pattern, including the period and frequency of consumption and amount of the product should be considered as the factors that have potential to influence the magnitude of the changes in gut microbiota [62]. Therefore, evaluating the effect of probiotics or fermented foods such as kefir on an individual basis may be set as a goal for future studies.

In this study, the correlations between changes in microbiota and anthropometric measurements or biochemical status were demonstrated. Our results pointed out a negative correlation between body fat mass and abundance of *Bacteroidetes*, whereas a positive correlation with the abundance of *Firmicutes* and *Actinobacteria* were observed. Although the data regarding the abundance of *Bacteriodetes* and *Firmicutes* phyla in obese and lean individuals is inconsistent, an overall analysis of results indicates an increase in *Firmicutes* with obesity [63]. Our results supported the previous reports, which revealed increased *Firmicutes* and decreased the abundance of *Bacteroidetes* are associated with obesity [64]. Turnbaugh et al. revealed a higher proportion of *Actinobacteria* in obese individuals compared to lean individuals [65]. In this study, we observed a positive correlation between body fat mass and abundance of *Actinobacteria,* which is in line with Turnbaugh et al.'s work. Members of the phylum of *Proteobacteria* are gram–negative bacteria and include several common human pathogens. An association between the increased relative abundance of *Proteobacteria* and increased risk of cardio–metabolic disorders was suggested previously [66]. In parallel with these findings, our study showed a positive correlation between an increased relative abundance of Proteobacteria and both body weight and diastolic blood pressure.

In terms of glycaemic status, Larsen et al. showed a lower abundance of *Firmicutes* and *Clostridia,* and a higher abundance of *Bacteroidetes* and *Betaproteobacteria* in diabetic patients compared to the non–diabetics [67]. Accordingly, we observed a lower abundance of *Firmicutes* and a higher abundance of *Bacteroidetes* in both groups at the beginning of the study. However, only the change in the relative abundance of *Bacteroidetes* by the dietary intervention was negatively correlated with the change in fasting blood glucose. This was parallel to results of the study conducted by Egshatyan et al., which found that microbiota of glucose–intolerant subjects were represented by *Firmicutes* phylum and to a lesser degree by *Bacteroidetes* phylum [68]. In the subphylum level, studies mostly indicate higher levels of *Bacteroides* and *Prevotella* and lower levels of butyrate producing–bacteria in type 2 diabetic patients [68,69]. In this study, we have observed a positive correlation between the relative abundance of *Odoribacteraceae* and *Alistipes,* and fasting plasma glucose. The correlation between *Alistipes* and blood glucose was also observed in a previous study [70].

Many researchers demonstrated a link between dysbiosis of gut microbiota and blood pressure. Yang et al. reported an increase in *Firmicutes*/*Bacteroidetes* ratio in hypertensive rats and humans. They also recorded a lower abundance of *Actinobacteria* as well as acetate– and butyrate–producing bacteria [71]. Yan et al. indicated higher levels of *Proteobacteria* but lower levels of *Actinobacteria* in hypertensive subjects [72]. In this study, the change in the relative abundance of *Bacteroidetes* was negatively correlated with the change in systolic and diastolic blood pressure as reported previously. Surprisingly, we observed a strong positive correlation with systolic and a weak positive correlation with diastolic blood pressure, respectively, with *Actinobacteria* abundance. The phylum *Actinobacteria* includes *Bifidobacterium* genera, possessing probiotic features [73]. Studies that report reductions of *Actinobacteria* in hypertensive patients explained this relationship mostly with *Bifidobacterium* levels [71,74]. In our study, when we analyzed the association between *Bifidobacterium* and blood pressure, no significant correlation was noted. This may be due to the lack of significant changes in *Bifidobacterium* abundance after the intervention. The species other than *Bifidobacterium* within the *Actinobacteria* phylum might be further investigated in terms of their contributions to hypertension.

Apart from fermentation, unfermented dairy products may also affect the gut microbiota [75,76]. In our study, regular milk consumption that was used as control also led to some changes in microbiota composition compared to the baseline. *Firmucutes* and *Verrucomicrobiota* were increased with milk consumption. However, *Bacteroidetes* group was decreased compared to the baseline. In accordance with our results, Ntemiri et al. also found out that whole milk consumption was associated with an increase in taxons belonging to *Firmicutes* and a higher *Firmicutes*/*Bacteroidetes* ratio [76]. In another randomized cross–over study, consumption of probiotic yogurt and milk acidified with D-(+)-glucono-δ-lactone showed some distinct effects on microbiota composition. In both groups, the abundance of *Bilophila wadsworthia* was reduced. However, only the abundance of *Bifidobacterium* species was increased with acidified milk intake, and it was suggested that gluconic acid in milk might possess prebiotic activity [75]. In a like manner, exopolysaccharides such as kefiran derived from kefir also suggested as bioactive compounds due to their potential prebiotic effects and relation to alteration of intestinal microbiota [77]. These results support a strong interaction between diet and microbiota even without probiotic intervention.

Studies showed that not only the composition of microbiota but also its functionality plays a role in the metabolic status [78,79]. This study focused only on the composition of gut microbiota; any consideration of its functionality was not taken into account. Using the metabolites of microbiota, such as postbiotics, as markers of the efficiency might have provided a better understanding. This should be noted as the main limitation of the study. The small number of participants in each arm could also be considered as the other limitation of the study.

#### **5. Conclusions**

In conclusion, to our knowledge, this was the first report exploring the effect of kefir on microbiota composition in patients with metabolic syndrome. This study indicated that kefir consumption could provide some potential improvements, especially in glycaemic status, inflammation–related indicators, and blood pressure, however, none of these improvements might stay significant when compared the changes led by unfermented milk consumption. Regarding to microbiota composition, the relative abundance of *Actinobacteria* phylum were increased in the kefir group compared to the baseline, even though a similar change by unfermented milk was also reported. Furthermore, this study underlined the potential alterations in gut microbiota composition that can be correlated with some indicators of the metabolic status led by both kefir and milk consumption, even if the magnitude of the efficiency remained limited. Further studies, especially randomized controlled trials, are needed to clarify the efficiency of kefir on gut microbiota and its link to metabolic status.

**Author Contributions:** Conceptualization, E.B.-K. and Z.B.; Data curation, E.B.-K. and Z.B.; Formal analysis, E.B.-K., K.E. and Z.B.; Funding acquisition, Y.A., E.Y. and Z.B.; Investigation, E.B.-K., B.P.S.-Y., C.K. and A.G.O.; Methodology, E.B.-K., B.P.S.-Y., Y.A., F.A.-K., C.K., A.B., A.N., K.E. and E.Y.; Writing—original draft, E.B.-K., K.E. and Z.B.; Writing—review & editing, C.K., K.E. and Z.B.

**Funding:** This research was funded by the Turkish Council of Higher Education.

**Acknowledgments:** We thank all study participants for their cooperation.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **The Interplay between Maternal and Post-Weaning High-Fat Diet and Gut Microbiota in the Developmental Programming of Hypertension**

### **Chien-Ning Hsu 1,2, Chih-Yao Hou 3, Chien-Te Lee 4, Julie Y.H. Chan <sup>5</sup> and You-Lin Tain 5,6,\***


Received: 26 July 2019; Accepted: 21 August 2019; Published: 22 August 2019

**Abstract:** Excessive intake of saturated fat has been linked to hypertension. Gut microbiota and their metabolites, short-chain fatty acids (SCFAs), are known to be involved in the development of hypertension. We examined whether maternal and post-weaning high-fat (HF) diet-induced hypertension in adult male offspring is related to alterations of gut microbiota, mediation of SCFAs and their receptors, and downregulation of nutrient-sensing signals. Female Sprague–Dawley rats received either a normal diet (ND) or HF diet (D12331, Research Diets) during pregnancy and lactation. Male offspring were put on either the ND or HF diet from weaning to 16 weeks of age, and designated to four groups (maternal diet/post-weaning diet; *n* = 8/group): ND/ND, HF/ND, ND/HF, and HF/HF. Rats were sacrificed at 16 weeks of age. Combined HF/HF diets induced elevated blood pressure (BP) and increased body weight and kidney damage in male adult offspring. The rise in BP is related to a downregulated AMP-activated protein kinase (AMPK)–peroxisome proliferator-activated receptor co-activator 1α (PGC-1α) pathway. Additionally, HF/HF diets decreased fecal concentrations of propionate and butyrate and decreased G protein-coupled receptor 41 (GPR41), but increased olfactory receptor 78 (Oflr78) expression. Maternal HF diet has differential programming effects on the offspring's microbiota at 3 and 16 weeks of age. Combined HF/HF diet induced BP elevation was associated with an increased *Firmicutes* to *Bacteroidetes* ratio, increased abundance of genus *Akkermansia* and phylum *Verrucomicrobia*, and reduced abundance in genus *Lactobacillus*. Maternal gut microbiota-targeted dietary interventions might be reprogramming strategies to protect against programmed hypertension in children and their mothers on consumption of a fat-rich diet.

**Keywords:** AMP-activated protein kinase; butyrate; developmental origins of health and disease (DOHaD); gut microbiota; high fat diet; hypertension; nutrient-sensing signals; propionate; short chain fatty acids

#### **1. Introduction**

Non-communicable diseases (NCDs) are increasingly becoming the leading causes of global morbidity and mortality [1]. Among NCDs, hypertension-related diseases are the most common causes of deaths. Despite substantial advances in therapy, the global epidemic rise of NCDs remains a significant challenge. Early-life exposure can program the onset of chronic NCDs [2], now framed as the "developmental origins of health and disease" (DOHaD) [3].

Perinatal nutrition affects fetal development and long-term health of the offspring. Imbalanced maternal diet may induce fetal programming that permanently alters the morphology and function of fetal organs and systems, leading to various NCDs, including hypertension [4]. The high-fat (HF) diet model has been used to study obesity-related disorders like hypertension [5,6]. Along these lines, using a rat model of maternal plus post-weaning HF diets, we have demonstrated that adult male offspring exposed to HF intake develop hypertension [7].

Among the proposed mechanisms linking maternal nutritional insults to offspring adverse outcomes, changes of gut microbiota and their metabolites have recently received more attention [4,8]. Diet is an instrumental factor in shaping the gut microbiota. Increasing evidence links gut microbiota dysbiosis to the development of a variety of diseases [9]. During pregnancy, the diet–gut microbiota interactions can mediate epigenetic regulation of gene expression not only in mother but also in the fetus via the contact with their metabolites [10]. The offspring gut microbiota is highly sensitive to the early-life environmental stimuli. Accordingly, maternal diet can influence the gut microbiota of mothers and their offspring, consequently driving developmental programming of chronic diseases in adult offspring [11,12]. Although several microbial markers have been reported related to HF consumption, like increased abundance of phylum *Firmicutes* and decreased *Bacteroidetes* [9], whether a similar pattern of results can be obtained from offspring born to mothers fed with HF diet is largely unknown.

The gut microbiota produces a variety of metabolites like short-chain fatty acids detectable in host circulation [13]. Short-chain fatty acids (SCFAs, e.g., acetate, butyrate, and propionate) and their receptors are reported to be involved in the regulation of blood pressure (BP) [14]. In line with this, a recent study from our laboratory reported that prebiotic or probiotic therapy can alter gut microbiota, regulate SCFAs and their receptors, and mediate nutrient-sensing signals to protect adult male offspring against hypertension programmed by high-fructose diet [15].

Nutrient-sensing signals are regarded as key players in the developmental programming of hypertension, such as 5'-adenosine monophosphate-activated protein kinase (AMPK), peroxisome proliferator-activated receptor (PPAR), and PPARγ co-activator 1α (PGC-1α) [4,16]. Activation of AMPK by resveratrol can affect PGC-1α activity to regulate the downstream expression of PPAR target genes [17]. Our recently published study demonstrated that HF diet-induced hypertension is correlated to inhibitory AMPK/PGC-1α pathway and altered gut microbiota [18].

Our objective in this study was to examine whether maternal and post-weaning HF diet cause differential effects on BP, gut microbiota, SCFAs and their receptors, and nutrient-sensing signals in adult offspring.

#### **2. Materials and Methods**

#### *2.1. Animal Model*

This study was followed the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Institutional Animal Care and Use Committee of the Kaohsiung Chang Gung Memorial Hospital (IACUC permit number: 201721408). Virgin female Sprague–Dawley (SD) rats (*n* = 12) were obtained from BioLASCO Taiwan Co., Ltd. (Taipei, Taiwan) and maintained in an Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC)-approved animal facility in our hospital. The rats were housed in a in a controlled environment with 12:12 light-dark cycle and humidity of 55%, throughout the study. Male SD rats were caged with female rats until mating. The presence of the plug confirmed mating. Female rats were weight-matched and assigned to receive either a normal diet with regular rat chow (ND; Fwusow Taiwan Co., Ltd., Taichung, Taiwan; 52% carbohydrates, 23.5% protein, 4.5% fat, 10% ash, and 8% fiber) or a 58% high-fat diet (D12331, Research Diets, Inc., New Brunswick, NJ, USA; 58% fat (hydrogenated coconut oil), 25.5% carbohydrate, 16.4% protein, and 0% fiber) during pregnancy and

lactation. After birth, litters were culled to eight from each mother to standardize the received quantity of milk and maternal pup care. Since men are much more likely to be hypertensive than women at a younger age [19], only male offspring were used. Male offspring were weaned at 3 weeks of age, and onto either the normal diet (ND) or HF diet ad libitum from weaning to 16 weeks of age. Rats were assigned to four experimental groups (maternal diet/postweaning diet; *n* = 8/group): ND/ND, HF/ND, ND/HF, and HF/HF.

We used BP-2000 tail-cuff system (BP-2000, Visitech Systems, Inc., Apex, NC, USA) to measure BP in conscious rats at 3, 4, 8, 12, and 16 weeks of age [7]. Rats were allowed to adapt to restraint and tail-cuff inflation for 1 week prior to the experiment. Rats were placed on the specimen platform. Their tails were passed through a cuff and immobilized by adhesive tape. Following a 10-min warm-up period, 10 preliminary cycles were performed to allow the rats to adjust to the inflating cuff. For each rat, three stable measures were taken and averaged. Fresh feces samples were collected at 3 and 16 weeks of age, frozen, and stored at −80 ◦C until use. At 16 weeks of age, rats were anesthetized by intraperitoneally injecting ketamine (50 mg/kg body weight) and xylazine (10 mg/kg body weight) and were euthanized by intraperitoneally injecting an overdose of pentobarbital for sacrifice. Blood samples were collected. Kidneys were harvested and stored at −80 ◦C in a freezer for further analysis.

#### *2.2. Gas Chromatography-Flame Ionization Detector (GC-FID)*

We used gas chromatography-mass spectrometry (GCMS-QP2010; Shimadzu, Kyoto, Japan) with a flame ionization detector (FID) to measure levels of acetate, butyrate, and propionate in the plasma and feces [15]. We used internal standards in analytical standard grades for acetate and propionate (from Sigma-Aldrich, St. Louis, MO, USA), and for butyrate (from Chem Service, West Chester, PA, USA). The working solutions of used as internal and external standards were prepared at the concentration of 10 mM. These solutions were kept at −20 ◦C in a freezer. Dry air, nitrogen, and hydrogen were supplied to the FID at 300, 20 and 30 mL/min, respectively. A 2-μL aliquot of sample was injected into the column. The inlet and FID temperature were set at 200 and 240 ◦C, respectively. The total running time was 17.5 min.

#### *2.3. Analysis of Gut-Microbiota Composition*

Metagenomic DNA was extracted from frozen fecal samples after centrifugation. All polymerase chain-reaction amplicons were mixed together and sent to the Genomic and Proteomic Core Laboratory, Kaohsiung Chang Gung Memorial Hospital (Kaohsiung, Taiwan) for sequencing using an Illumina Miseq platform (Illumina, San Diego, CA, USA) [15]. Amplicons were prepared according to the 16S Metagenomics Sequencing Library Preparation protocol (Illumina, San Diego, CA, USA), and sequenced with the Illumina MiSeq platform (Illumina, San Diego, CA, USA). Sequences (Illumina, San Diego, CA, USA) with a distance-based similarity of 97% or greater were grouped into operational taxonomic units (OTUs) using the USEARCH algorithm. To determine the significantly differential taxa, we applied linear discriminant analysis effect size (LEfSe) to compare samples between groups. The LEfSe uses linear discriminant analysis (LDA) to estimate the effect size of each differentially abundant feature. The threshold of the linear discriminant was set to two.

#### *2.4. Western Blot*

Western blot analysis was performed using the methods published previously [7]. Protein samples (200-μg kidney cortex) were boiled with gel-loading buffer for 5 min, subjected to 10–15% SDS-PAGE, and then transferred to a nitrocellulose membrane (GE Healthcare Bio-Sciences Corp., Piscataway, NJ, USA). To verify equal loading, the membranes were incubated with Ponceau S red (PonS) stain solution (Sigma-Aldrich, St. Louis, MO, USA) for 10 min on the rocker. Two nutrient-sensing signals, AMPKα2 and PGC-1α, were analyzed. Additionally, we determined the protein abundance of three SCFA receptors, including G protein-coupled receptor 41 (GPR41), GPR43, and olfactory receptor 78 (Olfr78). We used the following primary antibodies: a rabbit polyclonal anti-rat phosphorylated

AMPKα1/2 antibody (1:1000, overnight incubation; Santa Cruz Biotechnology), a rabbit polyclonal anti-PGC-1α antibody (1:1000, overnight incubation; Abcam, Cambridge, MA, USA), a rabbit polyclonal anti-GPR41 antibody (1:500, overnight incubation; USBiological, Salem, MA, USA), a rabbit polyclonal anti-GPR43 antibody (1:500, overnight incubation; Millipore, Burlington, MA, USA), and a rabbit polyclonal anti-Olfr78 antibody (1:500, overnight incubation; Assay Biotech, Fremont, CA, USA). Next, the membrane was washed five times with 0.1% T-TBS, incubated for 1h with a peroxidase-labeled secondary antibody diluted 1:1000 in T-TBS, and then developed using Chemi Doc (Bio-rad Image Lab 5.0). Bands were quantified by densitometry as integrated optical density (IOD). IOD was then normalized to total protein PonS staining. The protein abundance was represented as IOD/PonS.

#### *2.5. Immunohistochemistry Staining*

Paraffin-embedded tissues sectioned at 3-μm thickness were deparaffinized in xylene and rehydrated in a graded ethanol series to phosphate-buffered saline. Following blocking with immunoblock (BIOTnA Biotech., Kaohsiung, Taiwan), the sections were incubated for 2 h at room temperature with an anti-phosphorylated AMPKα2 antibody (1:400, Cell Signaling, Danvers, MA, USA) or an anti-PGC-1α antibody (1:200, Abcam, Cambridge, MA, USA). Immunoreactivity was revealed using the polymer-horseradish peroxidase (HRP) labeling kit (BIOTnA Biotech). For the substrate–chromogen reaction, 3,30-diaminobenzidine (DAB) was used. An identical staining protocol omitting incubation with primary antibody was employed to prepare samples that were used as negative controls. Renal cells positive for immunostaining were examined in 10 randomly selected ×400 microscopic fields per section. The number of immunostained cells was expressed as we described previously [18].

#### *2.6. Statistical Analysis*

Data are reported as the mean ± standard error of mean (SEM). A value of *p* < 0.05 was considered statistically significant. Statistical analysis was conducted with one-way analysis of variance (ANOVA) with a Tukey post hoc test for multiple comparisons. Analyses were performed using the SPSS software 14.0 (SPSS Inc., Chicago, IL, USA).

#### **3. Results**

#### *3.1. The E*ff*ects of Maternal and Post-Weaning HF Diet on Morphological Values and BPs*

Post-weaning consumption of HF diet caused a greater body weight (BW) compared with controls and the HF/ND group, with the greatest BW in the HF/HF group (Table 1). The kidney weights and the ratios of kidney weight-to-body weight were lower in the ND/HF and HF/HF groups compared to controls and the HF/ND groups. At 16 weeks of age, maternal and post-weaning HF diet increased systolic BP by 5 and 11 mmHg compared to controls, respectively. There is a synergistic effect of maternal and post-weaning HF diet on systolic BP, resulting in an increase of ~26 mmHg in the HF/HF group versus control. Similarly, diastolic BP and mean arterial pressure were higher in the HF/ND and ND/HF group compared with those in the control group, with the highest in the HF/HF group. Figure 1 shows the systolic BPs of ND/HF and HF/HF group were significantly higher than those in the control group from 8 to 16 weeks. By 12 weeks of age, the systolic BP had significantly increased in the HF/HF group relative to the other three groups. The plasma creatinine level was higher in HF/HF group compared to the controls. These findings indicate that maternal or post-weaning HF diet more or less caused a rise in BW and BPs, which was enhanced to a greater extent in the combined HF/HF diets. However, only combined HF/HF diet resulted in kidney damage, represented by elevated creatinine levels.


**Table 1.** Measures of morphological values, blood pressure, and renal function in 16-week-old male offspring exposed to high-fat diet (HF).

ND/ND, maternal plus post-weaning normal diet; HF/ND, maternal high-fat diet plus post-weaning normal diet; ND/HF, maternal normal diet plus post-weaning high-fat diet; HF/HF, maternal plus post-weaning high-fat diet. BW, body weight; *n* = 8/group; <sup>a</sup> *p* < 0.05 vs. ND/ND; <sup>b</sup> *p* < 0.05 vs. HF/ND; <sup>c</sup> *p* < 0.05 vs. ND/HF.

**Figure 1.** Effects of maternal and postnatal high-fat (HF) diet on systolic blood pressure in male offspring from 3 to 16 weeks. ND/ND, maternal plus post-weaning normal diet; HF/ND, maternal high-fat diet plus post-weaning normal diet; ND/HF, maternal normal diet plus post-weaning high-fat diet; HF/HF, maternal plus post-weaning high-fat diet. \* *p* < 0.05 vs. ND/ND; # *p* < 0.05 vs. HF/ND; † *p* < 0.05 vs. ND/HF.

#### *3.2. The E*ff*ects of Maternal and Post-Weaning HF Diet on Nutrient-Sensing Signals*

We evaluated key elements in the nutrient-sensing pathway, including phosphorylated AMPKα2 and PGC-1α. As shown in Figure 2, the renal protein level of phosphorylated AMPKα2 (Figure 2B) was lower in the HF/ND, ND/HF, and HF/HF group compared with that in the ND/ND group. Additionally, the HF/HF diet caused a significant reduction of PGC-1α versus the controls in offspring kidneys (Figure 2C). We next evaluated phosphorylated AMPKα2 (Figure 3) and PGC-1α (Figure 4) in the offspring kidneys by immunohistochemistry.

**Figure 2.** (**A**) Representative western blots showing phosphorylated AMP-activated protein kinase (AMPKα2, ~63kDa), peroxisome proliferator-activated receptor co-activator 1α (PGC-1α, ~90kDa), G protein-coupled receptor 41 (GPR41, ~38kDa), GPR43 (~47kDa), and olfactory receptor 78 (Oflr78) (~35kDa) bands in offspring kidneys at 16 weeks of age. Relative abundance of renal cortical (**B**) phosphorylated AMPKα2, (**C**) PGC-1α, (**D**) GPR41, (**E**) GPR43, and (**F**) Oflr78 were quantified. ND/ND, maternal plus post-weaning normal diet; HF/ND, maternal high-fat diet plus post-weaning normal diet; ND/HF, maternal normal diet plus post-weaning high-fat diet; HF/HF, maternal plus post-weaning high-fat diet. *n* = 8/group. \* *p* < 0.05 vs. ND/ND; # *p* < 0.05 vs. HF/ND; † *p* < 0.05 vs. ND/HF.

**Figure 3.** (**A**) Light microscopic findings of phosphorylated AMPKα2 immunostaining in the kidney cortex in 16-week-old male offspring. Bar = 50 μm; (**B**) Quantitative analysis of phosphorylated AMPKα2-positive cells per microscopic field (400×); \* *p* < 0.05 vs. ND/ND; # *p* < 0.05 vs. HF/ND; † *p* < 0.05 vs. ND/HF.

Immunostaining of phosphorylated AMPKα2 in the glomeruli and renal tubules indicated intense staining in the ND/ND group (150 ± 15 positive cells), an intermediate level of staining in the HF/ND group (72 ± 11 positive cells) and ND/HF group (85 ± 21 positive cells), and little staining in the HF/HF group (24 ±17 positive cells) (Figure 3B). Similar to phosphorylated AMPKα2, maternal or post-weaning HF diet significantly decreased PGC-1α expression in the HF/ND group (112 ± 14 positive cells) and the ND/HF group (121 ± 21 positive cells) vs. the ND/ND group (220 ± 29 positive cells) (Figure 4A). Combined maternal and post-weaning HF diets caused the reduction of PGC-1α expression to a greater extent (36 ± 14 positive cells) (Figure 4B). Taken together, these findings indicated that HF/HF diet synergistically downregulated AMPK–PGC-1α pathway.

**Figure 4.** (**A**) Light microscopic findings of PGC-1α immunostaining in the kidney cortex in 16-week-old male offspring. Bar = 50 μm; (**B**) Quantitative analysis of PGC-1α-positive cells per microscopic field (400×); \* *p* < 0.05 vs. ND/ND; # *p* < 0.05 vs. HF/ND; † *p* < 0.05 vs. ND/HF.

#### *3.3. The E*ff*ects of Maternal and Post-Weaning HF Diet on SCFAs and Their Receptors*

It was reported previously that SCFAs are involved in the development of hypertension [14]. We investigated whether HF diet causes a rise in BP is related to alterations of SCFAs production and the expression of SCFA receptors. Our results demonstrated that post-weaning HF diet decreased fecal concentrations of acetate compared to the ND/ND and HF/ND group (Table 2). Fecal propionate and butyrate levels were lower in the ND/HF group than those in the ND/ND and HF/ND group. Similarly, combined maternal and post-weaning HF reduced fecal concentrations of propionate and butyrate compared to controls. We next evaluated the protein levels of SCFA receptors. Renal GPR41 expression was lower in the ND/HF and HF/HF group compared to that in the ND/ND group (Figure 2D). GPR43 protein level in offspring kidney was not different among the four groups (Figure 2E). However, combined HF/HF diets resulted in a significant increase of renal Olfr78 expression compared to the other three groups (Figure 2F).


**Table 2.** Fecal levels of acetate, propionate, and butyrate in in male offspring exposed to high-fat diet (HF) at 16 weeks of age.

ND/ND, maternal plus post-weaning normal diet; HF/ND, maternal high-fat diet plus post-weaning normal diet; ND/HF, maternal normal diet plus post-weaning high-fat diet; HF/HF, maternal plus post-weaning high-fat diet. BW, body weight; *n* = 8/group; <sup>a</sup> *p* < 0.05 vs. ND/ND; <sup>b</sup> *p* < 0.05 vs. HF/ND.

#### *3.4. The E*ff*ects of Maternal and Post-Weaning HF Diet on Gut Microbiota*

We further analyzed bacterial populations in the gut at the phylum and genus levels at 3 weeks (Figure 5) and 16 weeks of age (Figure 6). At 3 weeks, the age of weaning, the main phyla in the offspring born of dams fed with regular chow (ND) or HF diet were *Firmicutes*, *Bacteroidetes*, *Verrucomicrobia*, *Proteobacteria*, and *Actinobacteria*. Maternal HF intake caused a remarkable increase in the phylum *Firmicutes* (72.3 ± 4% vs. 52.9 ± 3.7%; *p* = 0.002), but a decrease in the *Verrucomicrobia* (10.3 ± 2.6% vs. 25.9 ± 4.5%; *p* = 0.007) and *Proteobacteria* (2.4 ± 0.2% vs. 4.8 ± 0.4%; *p* < 0.001) (Figure 5A). The *Firmicutes* to *Bacteroidetes* ratio has been considered a signature for hypertension [20,21]. In the current study, the *Firmicutes* to *Bacteroidetes* ratio was higher in the HF group (8.3 ± 1.6) compared to that in the control group (4.2 ± 0.6, *p* = 0.03) (Figure 5B). Additionally, the main bacterial genera were *Akkermansia*, *Blautia*, *Clostridium*, *Parabacteroides*, *Lactobacillus*, *Alkaliphilus*, *Ruminococcus*, *Sarcina*, *Natronincola*, and *Flavobacterium* (Figure 5C). Among them, maternal HF diet decreased abundance of genus *Akkermansia* (9.7 ± 2.4% vs. 25 ± 4.3%; *p* = 0.006) (Figure 5D). Conversely, abundance of genus *Clostridium* was induced in the HF group (19.9 ± 2.6%) compared with that in control (10.4 ± 1.2%; *p* = 0.03).

**Figure 5.** Effect of maternal high-fructose (HF) diet on offspring gut microbiota at 3 weeks of age. (**A**) Relative abundances of the top five phyla. (**B**) The *Firmicutes* to *Bacteroidetes* ratio. (**C**) Relative abundances of the top 10 genera. (**D**) Relative abundances of the genus *Akkermansia*. *n* = 16/group. \* *p* < 0.05 vs. ND.

As shown in Figure 6A, the main phyla in the offspring gut microbiota at 16 weeks were identical to those at 3 weeks of age. Combined HF/HF diet significantly reduced the abundance of the phylum *Bacteroidetes* (15.1 ± 1.7% vs. 30.2 ± 0.7%; *p* < 0.001), while augmenting the abundance of the *Verrucomicrobia* (15.6 ± 2.2% vs. 0.5 ± 0.2%; *p* < 0.001). Additionally, the *Firmicutes* to *Bacteroidetes* ratio was the highest in the HF/HF group compared to that in the other three groups (All *p* < 0.05) (Figure 6B).

Maternal HFintake decreased the abundance of genera *Lactobacillus*(HF/ND vs. ND/ND = 4.3 ± 0.8% vs. 13.7 ± 2.5%, *p* = 0.005) and *Turicibacter* (HF/ND vs. ND/ND = 0.9 ± 0.3% vs. 2.1 ± 0.4%, *p* = 0.035). The post-weaning HF diet caused an increase of genus *Akkermansia* (ND/HF vs. ND/ND = 9.4 ± 3.8% vs. 0.4 ± 0.2%, *p* = 0.002), and decreased the abundance of genera *Lactobacillus* (3 ± 0.5%, *p* = 0.002) and *Turicibacter* (0.6 ± 0.1%, *p* = 0.01). Combined HF/HF diet caused increases of several bacterial genera, including *Akkermansia*, *Clostridium*, and *Alkaliphilus* (Figure 6C; all *p* < 0.05). Conversely, the abundance of genera *Parabacteroides*, *Lactobacillus*, and *Ruminococcus* was reduced by HF/HF exposure (Figure 6C; all *p* < 0.05). Of note is that maternal (HF/ND: 4.3 ± 0.8%) and post-weaning HF diet (ND/HF: 3 ± 0.5%) both resulted in the reduced abundance in genus *Lactobacillus* compared to the ND/ND group (13.7 ± 2.5%; both *p* < 0.05). The combined HF/HF diet caused the reduction of genus *Lactobacillus* abundance to a greater extent (0.8 ± 0.3%, all *p* < 0.05) (Figure 6D).

**Figure 6.** Effect of maternal and post-weaning high-fructose (HF) diet on offspring gut microbiota at 16 weeks of age. (**A**) Relative abundances of the top five phyla. (**B**) The *Firmicutes* to *Bacteroidetes* ratio. (**C**) Relative abundances of the top 10 genera. (**D**) Relative abundances of the genus *Lactobaccilus*. *n* = 8/group. \* *p* < 0.05 vs. ND/ND; # *p* < 0.05 vs. HF/ND; † *p* < 0.05 vs. ND/HF.

The main bacterial species modified by the maternal HF diet were *Leptolyngbya laminosa* (LDA score = −3.1), *Enterococcus avium* (LDA score = −2.3), and *Enterococcus casseliflavus* (LDA score = −2.2) (Figure 7A). The post-weaning HF diet showed an increase in species *Lactococcus lactis* (LDA score = 2.6) and *Streptococcus dentirousetti*(LDA score = 2), and caused a decrease in the species *Leptolyngbya laminosa*

(LDA score = −2.5) and *Enterococcus casseliflavus* (LDA score = −2.1) as compared to the ND/ND group (Figure 7B). Of note, there was a remarkable decrease in several species of *Lactobaccilus* in the HF/HF group vs. the ND/ND group (Figure 7C). Conversely, HF/HF diet caused an increase of in species *Akkermansia muciniphila* (LDA score = 2.1).

**Figure 7.** Effect of maternal and post-weaning high-fructose (HF) diet on 16-week-old offspring gut microbiota at the species level. Linear discriminant analysis (LDA), along with effect size measurements, was applied to identify enriched bacterial species. Most enriched and depleted species (LDA score (log10) > 2.0) in the (**A**) HF/ND (red) vs. ND/ND (green), (**B**) ND/HF (red) vs. ND/ND (green), and (**C**) HF/HF (red) vs. ND/ND (green). *n* = 8/group.

#### **4. Discussion**

This study provides a novel insight into the mechanisms responsible for the development of hypertension programmed by maternal and post-weaning HF diet with particular emphasis on gut microbiota-derived metabolites SCFAs and nutrient-sensing signals. The main findings of this study are as follows: (1) combined maternal plus postweaning HF diets induced elevated BP and increased BW and kidney damage in male adult offspring; (2) The combined HF/HF diets caused a rise in BP, which is related to a downregulated AMPK–PGC-1α pathway; (3) The offspring exposed to HF/HF diets had decreased fecal concentrations of propionate and butyrate, decreased renal GPR41 protein levels, and increased renal Oflr78 expression; (4) At 3 weeks of age, the maternal HF diet increased the *Firmicutes* to *Bacteroidetes* ratio and abundance of genus *Clostridium*, and decreased the abundance of genus *Akkermansia* in the gut microbiota in offspring; and (5) The HF/HF diet caused the rise of BP at 16 weeks of age, which was associated with the increased *Firmicutes* to *Bacteroidetes* ratio and reduced abundance in genus *Lactobacillus*.

In line with previous studies showing that maternal HF intake induces elevated BP in offspring [22,23], our results demonstrated that systolic BP was approximately 5 mmHg higher in the HF/ND group than that in the ND/ND group. Maternal HF diet-induced programmed hypertension may be related to a downregulated AMPK–PGC-1α pathway, an increased *Firmicutes* to *Bacteroidetes* ratio, and a decreased abundance of the genera *Akkermansia* and *Lactobacillus*. Additionally, we found that there is a synergistic effect between maternal and post-weaning HF diet causing a rise in BP and body weight, in support of our previous study showing that effect of maternal nutritional insults on the fetus are not set in stone and can be amplified by changes in the postnatal environment [24,25].

The observed effect of maternal HF diet on BP increase may be related to the inhibition of AMPK–PGC-1α pathway. The interplay between AMPK and other nutrient-sensing signals, driven by maternal nutritional insults, is known to regulate PPARs and their target genes, thus leading to programming of hypertension [17]. AMPKα2 knockout mice expressed activation of the renin-angiotensin system (RAS) to favor the development of hypertension [26]. Also, uni-nephrectomized rats developed hypertension, which was associated with decreased AMPK expression and activation of the RAS [27]. On the contrary, the AMPK activation has been shown to regulate the RAS, resulting in protection from hypertension in different models of programming [28,29]. Recently, AMPK activation has emerged as a reprogramming strategy, via regulating other nutrient-sensing signals like PGC-1α, to protect against hypertension and kidney disease with developmental origins [30]. In the current study, maternal or post-weaning HF diet reduced phosphorylated AMPKα2 and PGC-1α expression. Remarkably, combined maternal and post-weaning HF diets caused the reduction of phosphorylated AMPKα2 and PGC-1α expression to a greater extent in the HF/HF group. These results reconfirmed our previous study showing that combined HF/HF diet-induced hypertension is associated with reduced phosphorylated AMPKα2 and PGC-1α expression. These changes were restored by AMPK activation through resveratrol treatment [7]. These observations suggest that pharmacological therapies aimed at AMPKα2 as a reprogramming intervention to prevent hypertension programmed by maternal HF intake deserve further evaluation.

The results of this study showed that changes of SCFAs and their receptors are another mechanism contributing to HF/HF-induced hypertension. Although maternal HF diet had a neglectable effect on fecal SCFA levels and their receptors, post-weaning HF diet significantly reduced fecal propionate and butyrate concentrations. Propionate and butyrate have been reported to induce vasodilatation via mediating GPR41 and GPR43 receptor [14]. Conversely, acetate is a ligand for Olfr78 to raise BP [14]. Our report showed that combined HF/HF diet decreased fecal propionate and butyrate levels, decreased GPR41 expression, and increased Oflr78 expression in adult offspring kidneys, all of which may favor the development of hypertension. AMPK can be activated by SCFAs, like propionate and butyrate [31,32]. SCFAs have been report to protect against ethanol-induced gut leakiness via AMPK activation [33]. On the other hand, AMPK activation altered microbial populations, which promotes SCFA production [34]. In line with increasing evidence of a link between gut microbiota, SCFAs, and AMPK [35], our study demonstrated that HF/HF-induced hypertension is associated with inactivation of AMPK signaling and the reduction of SCFA production. Additional studies warranted to clarify whether microbiota-derived SCFAs regulate AMPK signaling contributing to hypertension programmed by HF diet.

Additionally, we observed the major acetate-producing bacteria could be either decreased (e.g., *Lactobacillus*) or increased (e.g., *Clostridiums* and *Akkermansia*) in the HF/HF group with hypertension. Unlike a previous study demonstrating that hypertension-associated dysbiosis is characterized by increases in lactate-producing bacteria [21], results of this study showed that the abundance of genera *Lactobacillus* and *Turicibacter*, which are lactate-producing bacteria, were decreased in the ND/HF and HF/ND-induced hypertension groups. Thus, additional studies are required to clarify whether the imbalance of gut acetate-, butyrate-, and propionate-producing bacterial populations directly contribute to BP control in a variety of programming hypertension models.

The detrimental effects of HF diet may also relate to alterations in gut microbiota composition. Emerging evidence shows that the development of hypertension is related to gut microbiota dysbiosis in animal models of hypertension [20,21]. Microbiota dysbiosis in early life has deleterious effects and may have long-term consequences leading to many diseases in later life [31]. Our results go beyond previous studies, demonstrating that altered gut microbiota links early-life HF intake to the developmental programming of hypertension. Although the interactions between dietary fat with the gut microbiota have been well explored in human and experimental studies [36], little is known about the impact of maternal fat intake on the offspring gut microbiota. Previous studies showed that maternal HF consumption can alter the offspring microbiome in various animal species [7,37,38]. In

line with this, our study demonstrated that maternal HF diet resulted in a considerable impact on the infant microbiota (i.e., 3 weeks of age), as reflected in a higher *Firmicutes* to *Bacteroidetes* ratio, higher abundance of genus *Clostridium*, and lower abundance of genus *Akkermansia*. However, these changes in microbiota compositions seem not persistent until adulthood (i.e., 16 weeks of age). An increased *Firmicutes* to *Bacteroidetes* ratio has been related to obesity in animals fed with saturated fat [36]. Our results go beyond previous reports, showing that mother rats exposed to HF intake caused an increase of the *Firmicutes* to *Bacteroidetes* ratio in their offspring's microbiota. We found it notable that a certain change in this ratio was persistent until adulthood in the HF/HF group, which had significant BP increases. Given previously published studies using this ratio as a microbial marker for hypertension [20,21], we speculate that this ratio might be a marker to predict hypertension of developmental origins.

*Akkermansia muciniphila* is the main genus classified in the *Verrucomicrobia* phylum, and recent studies revealed its beneficial effects against obesity and cardiometabolic disease [39]. According to our data, maternal HF diet reduced abundance of genus *Akkermansia* in 3-week-old offspring microbiota. Conflicting with previous reports showing that *Akkermansia muciniphila* abundance inversely correlated with obesity and hypertension [40,41], our results demonstrated that the combined HF/HF diet caused a more than a 100-fold increase of *Akkermansia muciniphila* abundance. Also, HF/HF diet increased abundance of genus *Akkermansia* and phylum *Verrucomicrobia* in offspring microbiota at 16 weeks of age. One possible reason was because we were experimenting in the model of developmental programming, which is more complex than the established disease models. Thus, further studies are needed to elucidate whether *Akkermansia muciniphila* may serve as a microbial marker for hypertension in other developmental programming models. Additionally, we observed that combined HF/HF diets caused a remarkable decrease in abundance of *Lactobacillus*, which is generally considered as a beneficial microbe [42]. Certain probiotic strains like *Lactobacillus* have shown hypotensive effects [43]. As we observed, several *Lactobacillus* spp. were depleted in the HF/HF group, and our previous study demonstrated that maternal *Lactobacillus casei* treatment protected adult offspring against programmed hypertension [15], there is a need to further explore whether early probiotic supplementation may serve as a reprogramming strategy to prevent hypertension programmed by HF/HF diets as well as in other programming models.

Our study has a few limitations. First, we did not examine serial changes in the composition of offspring microbiota. The alterations in gut microbiota we observed in adult offspring may reflect postnatal plasticity rather than programmed processes. Second, we did not analyze other organs controlling BP. The hypertensive effect of HF diet might be attributed to other organs, such as the heart, brain, and vasculature. Third, we employed 16S rRNA gene amplicon analysis to determine proportional changes among bacterial taxonomies. Further studies addressing gene functions contributed by the gut microbiome rather than abundance of taxa to hypertension of developmental origin are required. With the exception of hypertension, maternal HF diet has been used to model other DOHaD-related NCDs [5,44]. It remains to be determined whether changes in microbial composition and their metabolite SCFAs observed in the current study are involved in the pathogenesis of other NCDs. Last, only male offspring were studied in the present study. Given that sex differences appear in gut microbiota and hypertension [19,45], additional studies are required to clarify whether sex-specific interactions between gut microbiota and hypertension exist in mechanisms underlying hypertension programmed by HF diet.

#### **5. Conclusions**

In conclusion, several important mechanisms are involved in the development of hypertension programmed by maternal and post-weaning HF diet, including alterations of gut microbiota, SCFAs and their receptors, and nutrient-sensing signals. Targeting AMPK signaling, gut microbiota, and SCFAs might be a reprogramming strategy to reverse the development of hypertension programmed by high fat consumption. Although reprogramming strategies from animal models still await further clinical translation, our findings highlight that pregnant women and children's caretakers must pay attention to avoid excessive foods that have high fat content.

**Author Contributions:** C.-N.H.: contributed to concept generation, data interpretation, drafting of the manuscript, critical revision of the manuscript and approval of the article; C.-Y.H.: contributed to data interpretation, critical revision of the manuscript and approval of the article; C.-T.L.: contributed to data interpretation, critical revision of the manuscript and approval of the article; J.Y.H.C.: contributed to methodology and approval of the article; Y.-L.T.: contributed to concept generation, data interpretation, drafting of the manuscript, critical revision of the manuscript, and approval of the article.

**Funding:** This work was supported by grant MOST 107-2314-B-182-045-MY3 from the Ministry of Science and Technology, Taiwan.

**Acknowledgments:** We would like to thank the Genomic & Proteomic Core Laboratory, Department of Medical Research and Development, Kaohsiung Chang Gung Memorial Hospital, Kaohsiung, Taiwan, for gut microbiota profiling.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### *Lactobacillus reuteri* **V3401 Reduces Inflammatory Biomarkers and Modifies the Gastrointestinal Microbiome in Adults with Metabolic Syndrome: The PROSIR Study**

**Carmen Tenorio-Jiménez 1, María José Martínez-Ramírez 2,3, Isabel Del Castillo-Codes 4, Carmen Arraiza-Irigoyen 2, Mercedes Tercero-Lozano 4, José Camacho 5, Natalia Chueca 6,7, Federico García 6,7, Josune Olza 8, Julio Plaza-Díaz 7,8,9,10, Luis Fontana 7,8,9, Mónica Olivares 11, Ángel Gil 7,8,9,10 and Carolina Gómez-Llorente 7,8,9,10,\***


Received: 27 June 2019; Accepted: 29 July 2019; Published: 31 July 2019

**Abstract:** Previous studies have reported that probiotics may improve clinical and inflammatory parameters in patients with obesity and metabolic syndrome (MetS). *Lactobacillus (L.) reuteri* V3401 has shown promising results on the components of MetS in animal studies. We aimed to evaluate the effects of *L. reuteri* V3401 together with healthy lifestyle recommendations on adult patients with MetS. Methods: We carried out a randomized, crossover, placebo-controlled, single-center trial in which we included 53 adult patients newly diagnosed with MetS. Patients were block randomly allocated by body mass index (BMI) and sex to receive a capsule containing either the probiotic *L. reuteri* V3401 (5 <sup>×</sup> 109 colony-forming units) or a placebo once daily for 12 weeks. Anthropometric variables, biochemical and inflammatory biomarkers, as well as the gastrointestinal microbiome composition were determined. Results: There were no differences between groups in the clinical characteristics of MetS. However, we found that interleukin-6 (IL-6) and soluble vascular cell adhesion molecule 1 (sVCAM-1) diminished by effect of the treatment with *L. reuteri* V3401. Analysis of the gastrointestinal microbiome revealed a rise in the proportion of *Verrucomicrobia*. Conclusions: Consumption of *L. reuteri* V3401 improved selected inflammatory parameters and modified the gastrointestinal microbiome. Further studies are needed to ascertain additional beneficial effects of other probiotic strains in MetS as well as the mechanisms by which such effects are exerted.

**Keywords:** metabolic syndrome; gastrointestinal microbiome; *Lactobacillus reuteri* V3401; probiotics; obesity

#### **1. Introduction**

Obesity is a chronic disease, affecting developed and developing countries, that has multiple comorbidities and deteriorates quality of life. It is characterized by an increase of fat mass, which can consequently produce hypertrophy of the adipocytes, leading to an altered adipose tissue functionality. Individuals who are obese can develop an insulin resistance syndrome, also called metabolic syndrome (MetS). MetS is defined by insulin resistance, dyslipidemia, hypertension, and increased abdominal circumference, and it is associated with the development of type 2 diabetes (DM2), cardiovascular disease (CVD), and nonalcoholic fatty liver disease (NAFLD). This condition is associated with a two-fold increase in the risk of coronary heart disease, cerebrovascular disease, and a 1.5-fold increase in the risk of all-cause mortality [1], constituting a major public health challenge worldwide.

Nowadays, there is sound evidence linking the metabolic dysfunction seen in MetS to a proinflammatory state. Adipose tissue is, in part, responsible of this low-grade inflammatory state through the increasing release of proinflammatory molecules, such as leptin and tumor necrosis factor α (TNF-α), and the inhibition of adiponectin secretion, an anti-inflammatory adipokine [2]. In recent years, it has become evident that alteration of the gastrointestinal microbiome, also called gastrointestinal dysbiosis, may also contribute to the development of insulin resistance associated with obesity [3–5]. Furthermore, different studies have linked gastrointestinal dysbiosis with the development of obesity and other hallmarks of MetS [6,7]. In this sense, a decreased ratio of *Bacteroidetes*/*Firmicutes* has been described in individuals who are obese compared to normal-weight individuals [7]. Likewise, individuals with a low bacterial richness have more dyslipidemia, insulin resistance, inflammatory phenotype, and overall adiposity than individuals with high bacterial richness [8]. In addition, an aberrant gastrointestinal microbiome can promote subacute systemic inflammation, insulin resistance, and increased risk of CVD by mechanisms that include exposure to bacterial products, such as lipopolysaccharide (LPS), which is responsible for the metabolic endotoxemia related to MetS [9].

In the last years, treatment of the hallmarks of MetS with probiotics has emerged as a promising therapy. Probiotics are living microorganisms that confer health benefits to the host when administered in adequate amounts [10]. *Bifidobacterium* and *Lactobacillus* are the most frequently used genera of probiotics used in humans. Some of the beneficial effects of probiotics are mediated by their capacity to normalize the gastrointestinal microbiome, reinforce the gut barrier function composition [11,12], and their immunomodulatory actions [12,13]. Therefore, the addition of probiotics to a healthy diet could represent an interesting tool to fight obesity, MetS, and associated inflammation when used alongside dietary management and lifestyle modifications (e.g., increased physical activity). In this regard, some studies have found an improvement of anthropometric parameters and a decrease in inflammatory biomarkers in this disease after probiotic administration [14]. However, the beneficial effects of probiotics on MetS components are contradictory [15], probably because of the different probiotic strains, doses, and clinical study designs.

*Lactobacillus (L.) reuteri* V3401 strain, deposited in the Spanish Type Culture Collection (CECT) with accession number CECT 8695, was isolated from cow's raw milk on Mark, Rogosa and Sharper (MRS) agar medium, and 16S gene sequence analysis was carried out for its identification. In addition, its carbohydrate fermentation ability was characterized by the Analytical Profile Index (API) CH50 test [16]. This strain has been shown to reduce the absorption of fluoresterol, a fluorescent cholesterol analogue, by HT-29 human enterocytes [16]. Furthermore, Wistar rats fed a hypercholesterolemic diet supplemented with the probiotic strain for 57 days showed HDL levels similar to those of a healthy control group fed a standard diet [16]. Regarding glycemic levels, hypercholesterolemic animals supplemented with the probiotic strain showed similar values to those of normocholesterolemic mice, whereas animals under a hypercholesterolemic diet without the probiotic strain exhibited higher levels

than normocholesterolemic mice [16]. Higher glucose levels are related to insulin resistance, which is normally associated with hypercholesterolemia and low HDL levels, both of them components of MetS. In this setting, supplementation with *L. reuteri* V3401 might offer an additional metabolic advantage together with healthy diet and exercise recommendations in patients with MetS.

All things considered, the present study aimed to evaluate whether the consumption of the probiotic strain *L. reuteri* V3401, together with healthy lifestyle (hypocaloric diet and physical activity) recommendations, was capable of improving MetS components. For this purpose, we designed a double-blind, crossover, placebo-controlled, single-center, randomized clinical trial (RCT).

#### **2. Materials and Methods**

#### *2.1. Ethical Statement*

All research and procedures performed during the study complied with the Declaration of Helsinki and the Guidelines of Good Clinical Practice. After receiving a complete verbal description of the study, patients signed a written informed consent. The study protocol was approved by the local Ethics Committee of both Granada and Jaén (references CEI-Jaén 25022016 and CEI-Granada 28022016, respectively).

#### *2.2. Subjects and Experimental Design*

We performed a randomized, double-blind, crossover, placebo-controlled, single-center trial in patients with a new diagnosis of MetS, according to the criteria of the International Diabetes Federation (IDF). The complete study design including sample size, randomization, and the trial protocol have been previously published [17] and registered at www.clinicaltrials.gov as NCT02972567. The study was conducted in agreement with the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) guidelines.

Sample size was calculated based on the range and median value of lipopolysaccharide (LPS) and assuming a power of 80% and a significant level of 5% [17]. In brief, a total of 53 out of 60 adult patients were recruited at the Endocrinology and Nutrition Clinical Management Unit, University Hospital of Jaén (Jaén, Spain) by qualified personnel. Patients were block randomly allocated, by BMI and sex, in a 1:1 ratio to receive a capsule containing either the probiotic *L. reuteri* V3401 (5 <sup>×</sup> 10<sup>9</sup> colony-forming units) or the placebo (maltodextrin) once daily for 12 weeks.

Both capsules, probiotic and placebo, were provided by Biosearch Life (Granada, Spain). In addition, participants received an intensive lifestyle intervention program that included nutritional and physical counseling to achieve and maintain a 7% loss of initial body weight and increase moderate-intensity physical activity for at least 150 min/week. In Figure 1 we summarized the experimental design of the study.

**Figure 1.** Illustration of the experimental design: A randomized, double-blind, crossover, placebocontrolled, single-center trial comparing the effect of consumption of *Lactobacillus reuteri* V3401 for 12 weeks on various clinical, biochemical, and inflammatory biomarkers and gastrointestinal microbiota. w: weeks; t: time; x: sample collection.

#### *2.3. Anthropometric, Biochemical, Inflammatory, and Cardiovascular Data*

We performed a systematic symptom evaluation on each visit, with special emphasis on gastrointestinal symptoms, and a physical examination. Body weight (kg), height (cm), and waist circumference (cm) were measured by the same person using standardized procedures. Blood pressure was taken 3 times by the same person, and the mean of the three values was included. The biochemical analyses, including lipid and glucose metabolism, were performed at the University Hospital of Jaén following internationally accepted quality control protocols. Homeostasis assessment model for insulin resistance (HOMA-IR) was calculated using fasting plasma glucose and insulin values.

Blood samples were collected from each patient and after 12 h of fasting, at the beginning and the end of each intervention period. Serum and plasma samples were collected by centrifugation of blood samples and kept at −80 ◦C until analysis.

Plasma adipokines as well as cardiovascular and inflammatory biomarkers—adiponectin, leptin, resistin, IL-6, IL-8, TNF-α, total plasminogen activator inhibitor-1 (PAI-1), hepatocyte growth factor (HGF), monocyte chemoattractant protein 1 (MCP-1), soluble intracellular adhesion molecule 1 (sICAM-1), soluble vascular cell adhesion molecule 1 (sVCAM-1), and myeloperoxidase (MPO)—were analyzed on a Luminex 200 system (Luminex Corporation, Austin, TX, USA) with human monoclonal antibodies (EMD Millipore Corp, Billerica, MA, USA) using MILLIplex™ kits (HADK1MAG-16K, HSTCMAG-28SK, HAD2MAG-61K, HCVD2MAG-67K) according to the manufacturer's recommendations.

LPS and LPS-binding protein (LBP) were determined in serum samples using CEB526GE and SEB406 HU ELISA kits (Cloud-Clone Corp, TX, USA), respectively, following the manufacturer's instructions.

#### *2.4. Fecal Samples, DNA Extraction, and Next-Generation Sequencing*

Fecal samples were collected from each patient at each time (t1, t2, t2, t4, t5, and t6). Fecal samples were placed inside of a sterile plastic bottle and kept at −80 ◦C until analysis. DNA was extracted using a QIAamp DNA stool Mini Kit (QIAGEN, Barcelona, Spain) according to the manufacturer's instructions, with the exception that samples were incubated with the lysis buffer at 95 ◦C instead of 70 ◦C to guarantee the lysis of both Gram-positive and Gram-negative bacteria. Extracted DNA samples were sequenced at facilities of the Department of Microbiology, University Hospital Campus de la Salud (Granada, Spain). A 16S metagenomics sequencing was performed following the Illumina protocol.

In summary, the V3-V4 region of the bacterial 16S rRNA gene was amplified using the primers described by Klindworth et al., 2013 [18]. The PCR mixture was composed of 5 μL for each forward and reverse primers (1 μM, Macrogen, Seoul, Korea), 2.5 μL of DNA template samples, and 12.5 μL of 1x Hot Master Mix (KAPA HiFi HS RM, Roche, Basilea, Switzerland) to a final volume of 25 μL. Five microliters of elution solution was used for the negative control. The PCR conditions were: initial denaturation at 95 ◦C for 3 min, followed by 25 cycles of denaturation at 95 ◦C for 30 s, primer annealing at 55 ◦C for 30 s, extension at 72 ◦C for 30 s, and a final elongation at 72 ◦C for 5 min. The PCR products were demonstrated by electrophoresis on a 2% agarose gel. No amplification product was observed in the negative control. The amplifications were subjected to purification using Ampure beads (Agencourt Bioscience, La Jolla, CA, USA), the eluted DNA product was quantified using the assays of the Qubit kit (Invitrogen, Life Technologies, Waltham, Massachisetts, USA), and then all samples were pooled in equal concentrations for sequencing. Bioanalyzer 2100 was used with the DNA 1000 Chip kit (Agilent, Palo Alto, CA, USA) to evaluate the quality of the final products for each sample individually. Sequencing was carried out using Illumina MiSeq paired-end sequencing in an Illumina MiSeq device (Illumina Inc., San Diego, CA, USA) with 600 cycles (300 cycles for each paired reading and 12 cycles for the sequence of bar codes) according to the manufacturer's instructions. Sequence analysis was performed using the metagenomic workflow based on 16S of MiSeq Reporter v2.3 (Illumina Inc., San Diego, CA, USA).

#### *2.5. Taxonomic Analysis*

The "Quantitative Insights Into Microbial Ecology 2" (QUIIME 2) package was used to analyze sequence data [19]. Denoising quality, chimera check, and clustering were performed using the DADA2 plugins implemented in QUIIME 2. Amplicon sequence variants (ASVs) with a relative proportion lower than 0.1% were eliminated; as a result, the total numbers of ASV were reduced to 2015 but with a very low impact on the total data. The GreenGenes database (version 13.8), together with the naïve Bayes algorithm, was used as the reference 16S database.

#### *2.6. Statistical Analysis*

For the anthropometric, biochemical, and inflammatory biomarkers, results are presented as the mean values ± standard deviation (SD), unless otherwise indicated. For those variables not following a normal distribution, we applied the logarithmic transformation (insulin, HOMA index, glycated hemoglobin, total cholesterol, triacylglycerols, alanine aminotransferase (GPT), gamma glutamiltransferase (γGT), C reactive protein (CRP), IL-6, IL-8, adiponectin, resistin, HGF, sICAM, sVCAM and LBP) or the inverse transformation (high-density lipoprotein (HDL), aspartate aminotransferase (GOT).

Only patients with less than 5 missing data were considered, resulting in a final number of 34 patients. Missing data in these patients were imputed using principal component analysis (PCA) and trimmed score regression (TSR) [20]. The treatment effect in anthropometric, biochemical, and inflammatory biomarkers was evaluated according to the approach described by Wellek et al. [21]. Two tests were carried out: (i) a pretest for significance of carryover effects, and (ii) a test for significance of treatment effects. The treatment effects were considered significant for those biomarkers for which the null hypothesis of the pretest was not rejected and the null hypothesis of the test was rejected (*p* < 0.05), confirming that the biomarker presented statistically significant differences only due to treatment effects. *p*-value computations were confirmed with different state-of-the-art multivariate approaches, including multivariate analysis of variance (MANOVA) [22], partial least-squares discriminant analysis (PLS-DA) [23], and ANOVA simultaneous component analysis (ASCA) [24]. TheMEDA toolbox (https://github/josecamachop/MEDA-Toolbox) and the MANCOVAN toolbox (http://www.mathworks. com/matlabcentral/fileexchange/27014-mancovan) in Matlab (Mathworks) were used to perform the statistical analysis.

For the gastrointestinal microbiota analysis, the generated sequences, ASV, were normalized by means of the rarefaction method (Figure S1). The alpha diversity was measured by means of the Shannon index, whereas the unique fraction metric (Unifrac), both weighted and unweighted, was used to determine the beta diversity. When comparing the incremental of relative bacteria proportions before and after treatment (delta), a pairwise Wilcoxon signed-rank test was used. *p*-values were adjusted by False discovery rate-FDR (*q*-values).

#### **3. Results**

#### *3.1. Anthropometric, Biochemical, and Inflammatory Data*

Anthropometric and biochemical characteristics of the subjects are described in Table 1, whereas in Table 2 the inflammatory biomarkers determined in blood samples are described.

In the case of BMI, diastolic blood pressure, GOT, and LBP, we found that the washout period was not long enough to avoid the carryover effects. We found significant differences for Il-6, sVCAM (Figure 2), and insulin levels (Table 1); however, we did not find any significant results for HOMA index (Table 1).

**Figure 2.** Inflammatory biomarkers throughout the study. The levels of interleukin 6 (IL-6) and soluble vascular cell adhesion molecule 1 (sVCAM) were modified by the probiotic consumption (*p* < 0.05). Continuous line: probiotic group. Discontinuous line: placebo group.

#### *3.2. Gastrointestinal Microbiome Composition*

We characterized the gastrointestinal microbiome composition of the participants at the beginning, middle, and end of each intervention period (Table S1). However, we were unable to determine the specific presence of the *L. reuteri* V3401 strain in fecal samples due to the lack of specific primers for this strain. As shown in Figure 3, at the beginning of the intervention, the most abundant phyla were *Firmicutes* and *Bacteroidetes* followed by *Proteobacteria*, *Actinobacteria*, *Verrucomicrobia*, and *Cyanobacteria*.



Values are expressed as means ± SD. BMI: body mass index; SBP: systolic blood pressure; DBP: diastolic blood Pressure; HOMA-IR: Homeostasis assessment model for insulin resistance; LDL: low-density lipoprotein; HDL: high-density lipoprotein; GOT: aspartate aminotransferase; GPT: alanine aminotransferase; γ-GT: gamma glutamyltransferase.



LPS:

lipopolysaccharide;

 LBP:

lipopolysaccharide-binding

 protein.

**Figure 3.** Baseline gastrointestinal microbiome composition. Taxonomic composition of the gastrointestinal communities at the beginning of the intervention. The figure shows bar charts of the relative abundance of bacteria at the phylum level. Each column represents a participant.

Regarding the bacterial diversity, we did not find significant differences in the alpha diversity throughout the study, measured as the Shannon index (H) (Figure S2), or in the beta diversity (Figure S3). Therefore, our next analysis was to determine the evolution of the relative proportion of specific taxa, namely *Firmicutes*, *Bacteroidetes*, *Verrucomicrobia*, *Actinobacteria*, *Proteobacteria*, *Fusobacteria*, *Cyanobacteria*, *Elusimicrobia*, *Tenericutes*, and *Lentisphaerae*. It is worth mentioning there was an increase in the relative proportion of the *Verrucomicromia* phylum in the participants that consumed the probiotic strain (Figure 4A). The same results were found in the *Akkermansia* genus (Figure 4B).

Based on the results described above, we decided to determine whether there were significant differences in the relative abundance of these taxa due to the treatment (probiotic versus placebo). We, therefore, performed a pairwise comparison [25]. During the first intervention (t1, t2, and t3) we observed a significant increase in the delta values (t3–t1) in the *Verrucomicrobia* phylum due to the treatment (probiotic versus placebo). However, during the crossover intervention, the differences (t6–t4) were not statistically significant, although we found a significant trend (FDR *p* = 0.07) (Figure 5).

**Figure 4.** Trend of the relative proportion of *Verrucomicrobia* phylum and *Akkermansia* genus. The thick lines with error bars represent the means of both groups (Group 1: blue lines; group 2: orange lines). Dashed lines represent the means ± 2 and 3 standard deviations. Group 1 started the intervention receiving the placebo (t1, t2, and t3) and then was switched to receive the probiotic strain (t4, t5, and t6). Group 2 started the intervention receiving the probiotic strain (t1, t2, and t3) and then was switched to receive the placebo (t4, t5, and t6). (**A**) Temporal trend of *Verrucomicrobia* phylum; (**B**) Temporal trend of *Akkermansia* genus.

**Figure 5.** Relative abundance of the delta values of *Verrucomicrobia* phylum. The box plots indicate the relative abundance of *Verrucomicrobia* phylum between the final and the beginning points of each intervention—t3-t1 (**A**) and t6-t4 (**B**)—due to the treatment and the number of patients (n) of each group and treatment. (**A**) shows data for the first part of this intervention study (First intervention, 1: placebo; 2: *L. reuteri* V3401), whereas (**B**) shows data from the intervention after the crossover (Second intervention, 1: *L. reuteri* V3401; 2: placebo). (**C**) describes the Wilcoxon signed-rank test values and the significant levels by means of *p* and FDR *p*-values. FDR: False discovery rate; : outliers data values.

#### **4. Discussion**

To the best of our knowledge, the PROSIR study is the first randomized, crossover clinical trial in humans that evaluates whether the strain *L. reuteri* V3401 is capable of improving the components of MetS in humans when added to a healthy lifestyle. We did not find any differences in the clinical features of the syndrome between groups. This may be due to the fact that all subjects included in the study lost weight and improved their metabolic status as a result of the counseling to follow a healthy lifestyle that included diet and physical activity. However, we did find a decrease in IL-6 and sVCAM levels in patients who consumed the probiotic strain, together with a modification of the gastrointestinal microbiome, in particular, an increase in the *Verrucomicrobia* phylum.

Other studies have shown that consumption of *Lactobacillus casei* Shirota reduces sVCAM-1 levels in individuals who suffer from MetS, although in this study no effects on insulin sensitivity, endothelial function, or the inflammatory biomarkers were observed [26]. Bernini et al. [14] showed in another work that consumption of fermented milk enriched with *Bifidobacterium lactis* HN019 resulted in a reduction in BMI, an improvement in the lipid profile, and a significant decrease in proinflammatory cytokines (TNF-α and IL-6).

Systemic low-grade inflammation has an important role in the development of MetS. In this sense, IL-6 is a cytokine that has been associated with insulin resistance. Specifically, IL-6 is able to induce insulin resistance in both liver and adipocytes through reduction of phosphorylation of the insulin receptor substrate (IRS), or by transcription inhibition of the IRS [27,28]. In addition, adhesion molecules, such as sVCAM-1, are necessary for normal development and function of the heart and blood vessels, and they have been related to the development of CVD [29]. In the adult Spaniard population, impaired glucose metabolism has been related to increased levels of sVCAM-1 [30].

Results regarding the utility of probiotics in the treatment of MetS have been contradictory. This may be due to various facts: (i) the particular probiotic strain used in each trial; (ii) the experimental design—most of the studies have been parallel-group randomized trials, whereas a crossover study is a more appropriated approach to determine health benefits of clinical interventions; in crossover studies each participant serves as their own control, but in addition, this clinical design demands a lower sample size than parallel-group studies [21]; and (iii) the duration of the treatment. In our study, 12 weeks might not have been a long enough treatment to reverse, or at least improve, a chronic proinflammatory state as the one observed in MetS.

In recent years, it has become clear that the gut microbiota plays a role in the development of MetS. Specific bacterial groups have been described to be involved in obesity and related metabolic diseases. Among these bacteria, *Akkermansia muciniphila* has been proposed as a contributor to the maintenance of gut health and glucose homeostasis [31]. Administration of *A. muciniphila* to diet-induced obese animals improve their metabolic endotoxemia adipose tissue inflammation and insulin resistance [32]. It is worth mentioning that *Akkermansia* was the only genus of the *Verrucomicrobia* phylum present in the gastrointestinal samples [33]. In humans, *A. muciniphila* has been found to be decreased in prediabetic patients compared to normal glucose tolerance subjects [34]. Conversely, other studies have shown an increase of *A. muciniphila* in type 2 diabetes [35]. More recently, *A. muciniphila* has been reported to be associated with a healthy metabolic status in overweight and obese individuals, in agreement with previous results in murine studies [31]. Additionally, higher *A. muciniphila* abundance has been described in subjects with high bacterial gene richness, which is associated with a healthier metabolic status, in French and Danish population [8,31]. Although we found an increase in the proportion of the *Verrucomicrobia* phylum in the group that received the probiotic, we did not find any significant correlation between the delta proportions of *Verrucomicrobia* and any inflammatory biomarker.

It is important to highlight the age of the subjects that participated in our study. Our patients were younger compared to other previously published trials [14,15,36], and still we observed a decrease of inflammatory biomarkers and an increase in the abundance of *Verrucomicrobia* phylum in this pretreatment phase of the disease. This is consistent in both intervention periods, although in the case of gastrointestinal microbiome, the results only showed a trend in the second intervention, probably because of the dropout number we had, which is usual in large clinical intervention studies.

#### **5. Conclusions**

In conclusion, our data point to a beneficial effect of supplementation with *L. reuteri* V3401 in subjects with MetS when added to a hypocaloric diet and regular physical activity. In particular, these effects may be mediated by an improvement of dysbiosis and a decreased proinflammatory state, both features of this condition. However, further studies with longer periods of intervention are needed, in animals and clinical studies, to confirm these results and to elucidate the underlying mechanisms of action.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6643/11/8/1761/s1: Figure S1, Rarefaction curves; Figure S2, Alpha diversity measured by means of the Shannon index (H); Figure S3, Bacterial beta diversity; Table S1, Gastrointestinal microbiome normalized data.

**Author Contributions:** M.J.M.-R., M.O., Á.G. and C.G.-L. conceived and designed the study. C.T.-J., M.J.M.-R., I.D.C.-C., M.T.-L. and C.A.-I. were responsible for the clinical intervention and the follow-up of the patients, as part of the clinical team. J.P.-D., J.O., N.C. and F.G. performed the analysis. J.O., J.C., Á.G., C.T.-J., L.F. and C.G.-L. reviewed and did the statistics. C.T.-J. and C.G.-L. wrote the manuscript. L.F. revised the English and helped in writing the manuscript. All the authors read and approved the final manuscript.

**Funding:** This study is part of the grant entitled "Guía para la sustanciación de declaraciones de salud en alimentos: funciones inmune, cognitiva y síndrome metabólico", funded by the company Biosearch life (Granada, Spain), reference 3006, managed by Fundación General Universidad de Granada, Spain.

**Acknowledgments:** This paper will be part of Carmen Tenorio Jimenez's doctorate, which is being completed as part of Programa de Doctorado en Seguridad de los Alimentos at the University of Jaén, Spain. We also wish to thank all the participants who generously contributed to the study.

**Conflicts of Interest:** M.O. is Head of the Department of Research and Development at Biosearch Life. The funders had no role in the recruitment, biological sample analysis, statistical analysis, data interpretation, writing of the manuscript, nor in the decision to publish the results.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Probiotic Ingestion, Obesity, and Metabolic-Related Disorders: Results from NHANES, 1999–2014**

**Eva Lau 1,2,\*, João Sérgio Neves 1,3, Manuel Ferreira-Magalhães 2,4, Davide Carvalho 1,5 and Paula Freitas 1,5**


Received: 24 May 2019; Accepted: 26 June 2019; Published: 28 June 2019

**Abstract:** Gut microbiota dysbiosis has been recognized as having key importance in obesityand metabolic-related diseases. Although there is increasing evidence of the potential benefits induced by probiotics in metabolic disturbances, there is a lack of large cross-sectional studies to assess population-based prevalence of probiotic intake and metabolic diseases. Our aim was to evaluate the association of probiotic ingestion with obesity, type 2 diabetes, hypertension, and dyslipidemia. A cross-sectional study was designed using data from the National Health and Nutrition Examination Survey (NHANES), 1999–2014. Probiotic ingestion was considered when a subject reported consumption of yogurt or a probiotic supplement during the 24-h dietary recall or during the Dietary Supplement Use 30-Day questionnaire. We included 38,802 adults and 13.1% reported probiotic ingestion. The prevalence of obesity and hypertension was lower in the probiotic group (obesity-adjusted Odds Ratio (OR): 0.84, 95% CI 0.76–0.92, *p* < 0.001; hypertension-adjusted OR: 0.79, 95% CI 0.71–0.88, *p* < 0.001). Accordingly, even after analytic adjustments, body mass index (BMI) was significantly lower in the probiotic group, as were systolic and diastolic blood pressure and triglycerides; high-density lipoprotein (HDL) was significantly higher in the probiotic group for the adjusted model. In this large-scale study, ingestion of probiotic supplements or yogurt was associated with a lower prevalence of obesity and hypertension.

**Keywords:** intestinal microbiota; probiotics; nutrients

#### **1. Introduction**

Obesity is a pro-inflammatory state that plays a central role in the progression of several diseases, such as type 2 diabetes, hypertension, and dyslipidemia [1]. The pathophysiology of obesity and metabolic-related diseases is complex, resulting from the imbalance between environmental and genetic factors. The human gastrointestinal tract is populated by a complex ecosystem—the gut microbiota—which is responsible for the regulation of essential functions for the maintenance of health, including protective, structural, and histological functions [2]. New insights emphasize the role of gut microbiota in energy homeostasis, giving rise to the "The Metagenome Hypothesis" as a key player in the comprehension of metabolic diseases [3–5]. Accordingly, recent studies have shown

the relationship between intestinal dysbiosis, which is defined as a change in the composition of gut microbiota and glucose and lipid metabolism deregulation in obesity and type 2 diabetes [6,7]. Obesity has been linked to an increase in Firmicutes and a decrease in Bacteroidetes [5,8]. Likewise, two large metagenome-wide association studies reported that type 2 diabetes had a lower proportion of butyrate-producing *Clostridiales* (*Roseburia* and *Faecalibacterium prausnitzii*) and greater proportions of *Clostridiales* that do not produce butyrate [7,9]. On one hand, distinct differences in gut microbiota result in a greater increase in harvesting energy from the diet by fermentation and the absorption of dietary polyssacharides, promoting hepatic lipogenesis. On the other hand, gut microbiota regulate intestinal permeability and an increase in the translocation of lipopolysaccharide-containing gut microbiota increases the inflammatory state, which is named metabolic endotoxemia, accompanied by weight gain and insulin resistance [10]. Understanding this interplay between gut microbiota and the host has created interest in shaping microbiota to prevent, treat, or delay obesity, type 2 diabetes, and metabolic-associated complications.

Probiotics are food components or supplements with living microorganisms that confer health advantages to the host [11]; specific strains have been increasingly studied as a potential therapeutic approach to shape gut microbiota composition, with possible benefits for weight control and diabetes management [12–15]. The administration of probiotics may restore the crosstalk between human host and gut microbiota, controlling homeostatic functions during obesity and metabolic-related disorders. Although there is increasing evidence of the potential benefits of probiotics in metabolic diseases, there is a lack of large cross-sectional studies to evaluate the population-based prevalence of probiotic intake and metabolic differences in those exposed to probiotics compared to those who are not. Large population surveys can create powerful information about population health status and trends. To date, there are no published studies about the large-scale use of probiotic supplements and yogurts and possible associations with metabolic diseases. This type of analysis can produce high-quality data for the real-life use of these types of food and supplements. Our aim was to assess the association of probiotic ingestion, through yogurt or supplements, with the prevalence of obesity and associated metabolic disturbances, namely type 2 diabetes, hypertension, and dyslipidemia.

#### **2. Materials and Methods**

#### *2.1. Study Design and Settings*

We designed a cross-sectional analysis, using data from the National Health and Nutrition Examination Survey (NHANES). NHANES is a national research survey designed to collect demographic, socio-economic, health, and nutritional statuses from a representative sample of the non-institutionalized civilian resident population of the United States of America. NHANES is a major program of the National Centre for Health Statistics (NCHS), which is part of the Centers for Disease Control and Prevention (CDC), and the detailed methodology is described in the literature [16]. NHANES was approved by the NHANES Institutional Review Board (IRB) and the NCHS Research Ethics Review Board (ERB) (after 2003).

#### *2.2. Participants and Data Collection*

We included adults aged 18 years or older, who had been included in NHANES between 1999 and 2014. Pregnant women were excluded. NHANES participants without physical examination or laboratory data and with no dietary data or implausible dietary data (24-h dietary recall) were also excluded. Figure S1 (supplementary material) shows the flowchart of the study population. NHANES data collection was performed through an in-home interview for demographic and basic health information data collection, together with a health examination in a Mobile Examination Centre (MEC), where participants were examined and surveyed. NHANES MEC examinations included anthropometric measurements, blood pressure assessment, and blood workup. Data were collected

by a trained interviewer who had completed an intensive training course administered by the US Department of Agriculture and the US Department of Health and Human Services.

#### *2.3. Assessment and Definition of Probiotic Exposure*

In all NHANES cycles, from 1999 to 2014, a 24-h dietary recall was collected. Using an automated multiple-pass method, a detailed dietary intake (quality and quantity) for the 24-h period before the interview was recorded. For participants in the 1999–2002 NHANES, only one in-person 24-h dietary recall was performed. From 2003 onward, an additional telephone dietary recall interview was also performed 3 to 10 days following the in-person dietary interview. For the participants of the 2003–2014 NHANES, we used the mean of the nutritional information from both recalls (in-person recall and telephone recall). To assess probiotic supplementation exposure, we also used the Dietary Supplement Use 30-Day (DSQ), which assesses food supplement use during the preceding 30 days. Table S1 (supplementary materialsupplementary material) lists the probiotic supplements included.

Probiotic ingestion was considered when a subject reported consumption of a probiotic supplement or yogurt (as a dietary source of probiotics) during the 24-h dietary recall or of a probiotic supplement during the DSQ. Non-yogurt foods containing probiotics were classified as probiotic supplements.

#### *2.4. Definition of Metabolic Comorbidities, Smoking, and Physical Activity*

Obesity was defined as a body mass index (BMI) <sup>≥</sup>30 Kg/m2. Type 2 diabetes was defined as glycated hemoglobin (HbA1c) ≥6.5%, fasting plasma glucose level ≥126 mg/dL, or current glucose-lowering drug use. Dyslipidemia was assumed if participants had low-density cholesterol (LDL) ≥160 mg/dL, high-density cholesterol (HDL) <40 mg/dL, triglycerides ≥200mg/dL, total cholesterol ≥240 mg/dL, or if they were being treated with lipid-lowering drugs. Systolic and diastolic blood pressures (BP) were determined by the mean of 3 or 4 consecutive blood pressure readings. Hypertension was defined as systolic blood pressure ≥140 mmHg, or diastolic blood pressure ≥90 mmHg, or current medication for hypertension.

Smoking status was classified as former and current smokers. Current smokers were those who reported smoking at least 100 cigarettes during their lifetime and were currently smoking every day, or some days. Former smokers were those who reported smoking at least 100 cigarettes during their lifetime, but do not currently smoke.

Physical activity was measured differently along NHANES cycles. We classified participants using variables that allowed categorization of physical activity level into three categories (low, intermediate, and high). From 1999 to 2006 the physical activity level was assessed with the question "compare activity with others of the same age" (participants answering "less active" were classified into category "low", "about the same" into category "intermediate", and "more active" into category "high"). From 2007 to 2014, the weekly metabolic equivalents (MET) minutes of physical activity (accounting for vigorous work-related activity, moderate work-related activity, walking or bicycling for transportation, vigorous leisure-time physical activity, and moderate leisure-time physical activity) was divided into tertiles (participants were classified as "low" if included in the lower MET-minute tertile, as "intermediate" if in the middle MET-minute tertile, and as "high" if in the higher MET-minute tertile).

#### *2.5. Statistical Analysis*

Statistical analysis took into account the complex survey design of the NHANES dataset and was performed according to the CDC analytic recommendations [17].

Continuous variables were described as mean ± standard deviation (SD) and categorical variables were described as absolute and relative frequencies. To assess the association between probiotic exposure and metabolic comorbidities (obesity, diabetes, hypertension, and dyslipidemia), we performed unadjusted and adjusted logistic regression models. To evaluate the association between probiotic exposure and cardiomatebolic parameters (BMI, HbA1c, fasting plasma glucose, systolic BP, diastolic BP, LDL, HDL, and triglycerides), we performed unadjusted and adjusted multivariate linear regression models. We excluded those participants who were being treated with anti-hypertensive drugs from systolic and diastolic BP analysis, participants treated with anti-dyslipidemic drugs from analysis concerning lipid profiles, and patients treated with antidiabetic drugs from HbA1c and fasting plasma glucose analysis.

In the adjusted analyses, we used the following models (Table S2): Model 1, including age, sex, ethnicity (Mexican American, other Hispanic, non-Hispanic white), annual family income (<\$25,000, \$25,000 to \$75,000, >\$75,000), and education (<9th grade, ≥9th grade); model 2, including all model 1 covariates plus alcohol intake, smoking status (never a smoker, current smoker, or former smoker), physical activity (low, intermediate, high), ingested kcal per day, ingested carbohydrates/kcal per day, ingested protein/kcal per day, ingested fiber/kcal per day, and ingested polyunsaturated/saturated fatty acids ratio. In model 2, we also included BMI in all analyses except in the obesity analysis and sodium intake per day only in the hypertension and blood pressure analyses. As a supplementary analysis, we performed an additional model (model 3) that classified the diet pattern using the Dietary Approaches to Stop Hypertension (DASH) score. In model 3 we included all model 1 covariates plus alcohol intake, smoking status, physical activity, and the DASH dietary pattern score. Model 3 also included BMI in all analyses except the obesity analysis and sodium intake per day only in the hypertension and blood pressure analyses. The DASH score is based on 9 target nutrients (sodium, total fat, saturated fat, protein, fiber, cholesterol, calcium, magnesium, and potassium), as previously described [18]. Individuals meeting the DASH goal were given a score of 1.0 for that nutrient; if they attained an intermediate goal, they were given a score of 0.5 for that nutrient. The DASH score is the sum of the score for each individual nutrient. Logistic regression results were expressed as an odds ratio (OR) and a 95% confidence interval (95% CI). A two-sided *p*-value of <0.05 was considered statistically significant. Analyses were performed with Stata (version 14.2).

#### **3. Results**

#### *3.1. Baseline Characteristics According to Probiotic Consumption*

We included 38,802 adults, of whom 13.1% had exposure to probiotic supplements or yogurt. Baseline population characteristics according to probiotic consumption are described in Table 1. Participants in the group exposed to probiotic supplements or yogurt were more likely to be female, older, non-Hispanic white and to have a higher income and education level. Ingestion of kcal/day was similar between groups.


**Table 1.** Baseline population characteristics according to probiotic consumption (*n* = 38,802).


**Table 1.** *Cont.*

BMI: body mass index; BP: blood pressure; HDL: high-density lipoprotein; LDL: low-density lipoprotein; n.a.: Not applicable; \* statistically significant. # See methods regarding physical activity level definition.

#### *3.2. Prevalence of Metabolic Comorbidities According to Probiotic Consumption*

The prevalence of metabolic comorbidities according to probiotic supplement or yogurt exposure is represented in Figure 1. All four studied comorbidities were lower in the exposed group. The comorbidities that showed higher differences in prevalence between groups were obesity and hypertension; 5.4% and 2.9% lower, respectively. Diabetes prevalence difference was only 1.6%, but this was still significant in our analysis.

#### *3.3. Modulation of Metabolic Comorbidities According to Probiotic Consumption*

Table 2 summarizes the results of metabolic disturbances by using unadjusted and adjusted models of the prevalence of comorbidities according to probiotic supplement or yogurt exposure. For unadjusted analysis, participants exposed to probiotics manifested a 22% reduction in the odds of having obesity (OR: 0.78, 95% CI 0.71–0.86; *p* < 001), a 16% reduction in the odds of having diabetes (OR: 0.84, 95% CI 0.73–0.96; *p* = 0.020), and a 12% reduction in the odds of having hypertension (OR: 0.88, 95% CI 0.81–0.96; *p* = 0.004). No significant differences were found in dyslipidemia prevalence.

After adjusting for potential confounders, obesity and hypertension prevalence remained significantly lower in the probiotic exposed group (Table 2). In turn, diabetes prevalence became similar between groups.

**Figure 1.** Prevalence of obesity, diabetes, hypertension, and dyslipidemia, according to probiotic exposure.


**Table 2.** Odds ratios of disease in subjects exposed to probiotics compared to non-exposed subjects.

Model 1: Age, sex, race, income and education; model 2: Model 1 + alcohol intake, smoking status, physical activity, kcal per day, carbohydrates/kcal per day, protein/kcal per day, fiber/kcal per day, and polyunsaturated/saturated fatty acids ratio. Model 2 also includes BMI in all analyses except in the obesity analysis and includes sodium intake per day only in the hypertension analysis. \* Statistically significant.

Table 3 summarizes the results of cardiometabolic parameters according to probiotic supplement or yogurt exposure. In the unadjusted analysis, all the studied markers were lower in the probiotic-exposed group, with the exception of HDL (which was higher in the exposed group) and LDL (no differences were seen). After adjusting for potential confounders, BMI, and systolic BP, diastolic BP and triglycerides remained significantly lower in the probiotic-exposed group and HDL remained significantly higher (Table 3). Tables S3 sand S4 summarize the odds ratios of disease and the variation of cardiometabolic parameters, respectively, according to probiotic supplement or yogurt exposure, after accounting for the DASH dietary pattern score. The associations of probiotic supplement or yogurt ingestion with cardiometabolic parameters were not different after adjusting for DASH diet adherence.

**Table 3.** Variation of cardiometabolic parameters in participants exposed to probiotics compared to non-exposed participants.


Model 1: Age, sex, race, income and education; model 2: Model 1 + alcohol intake, smoking status, physical activity, kcal per day, carbohydrates/kcal per day, protein/kcal per day, fiber/kcal per day, and polyunsaturated/saturated fatty acids ratio. Model 2 also includes BMI in all analyses except in the BMI analysis and includes sodium intake per day only in the BP analyses. <sup>a</sup> Excluding participants treated with anti-hypertensive drugs. <sup>b</sup> Excluding participants treated with anti-dyslipidemic drugs. <sup>c</sup> Excluding participants treated with antidiabetic drugs. \* Statistically significant.

The odds of metabolic comorbidities according to the origin of probiotics is presented in Figure 2. Among those participants exposed to probiotics, 95.7% were exposed to yogurt and 5.4% were exposed to probiotic supplements. The odds of obesity, diabetes, hypertension, and dyslipidemia for participants exposed to yogurt or to probiotic supplements alone were similar to the odds for participants exposed to any type of probiotic. Although the confidence intervals were wider in the analysis of probiotic supplements (due to the smaller number of exposed individuals), the point estimates for the association with comorbidities was similar to the association with exposure to any probiotic or exposure to yogurt, with the exception of diabetes. Although the association in both analyses was not significant, the odds ratio for diabetes was 0.92 (95% CI 0.77–1.10) among participants exposed to yogurt and 1.28 (95% CI 0.71–2.23) among those exposed to probiotic supplements. 


**Figure 2.** Odds ratio of disease in participants exposed to probiotics compared to those not exposed, according to the origin of probiotics (all probiotics, yogurt, or probiotic supplements). Logarithmic regression models adjusted for age, sex, race, income, education, alcohol intake, smoking status, physical activity, carbohydrates/kcal per day, protein/kcal per day, fiber/kcal per day, and polyunsaturated/saturated fatty acids ratio (model 2). Model 2 also includes BMI in all analyses except in the obesity analysis and includes sodium intake per day only in the hypertension analysis.

#### **4. Discussion**

We conducted a cross-sectional analysis on a large and representative US population, for a total of 38,802 adults, and found that 13.1% reported the use of probiotic supplements or yogurt ingestion. Although there are several studies addressing the possible beneficial associations of probiotic ingestion and several metabolic outcomes, there is a lack of large cross-sectional studies to objectively assess population-based prevalence of probiotic intake and metabolic differences in those exposed and not exposed to probiotics. To our knowledge, this was the first large cross-sectional analysis aiming to assess the association of probiotic ingestion, either by probiotic supplements or yogurt, with metabolic disturbances. Probiotic ingestion was associated with a 17% lower prevalence of obesity and a

21% lower prevalence of hypertension. Furthermore, HDL cholesterol was significantly higher and triglyceride levels were significantly lower in the probiotic group.

Probiotics modulate gut microbial communities, exerting beneficial metabolic effects through the regulation of multitudinous physiological metabolic pathways. Among the molecular mechanisms, the regulation of adipogenesis, stimulation of insulin signaling, improvement of gut barrier function, reduction of metabolic endotoxemia, and down-regulation of cholesterol levels are some of the suggested key players in the crosstalk between probiotics and metabolic disorders [19].

#### *4.1. Obesity*

We found that probiotic ingestion, via supplements or yogurt, was associated with a lower prevalence of obesity (17% reduction), before and after adjusting for demographics and potential confounders. Although subjects that consumed probiotics had a higher consumption of carbohydrates, fiber, and protein, the effects of probiotics on BMI were significant, even after adjustment for confounders (−0.41 Kg/m2 between groups in model 2). Furthermore, there were no differences in total energy intake per day between groups and, also, there was no linear association of physical activity level (low, intermediate, and high) and probiotic ingestion. The only thing observed was a higher proportion of intermediate physical activity level, but lower proportions of high and low physical activity levels in the probiotic-exposed group. Putting this all together, these results support our hypothesis of the beneficial impact of probiotic ingestion per se on metabolic health, namely its effects on the regulation of body weight. In agreement with our study, a recent meta-analysis of randomized clinical trials with 957 subjects, with a mean BMI of 27.6 kg/m2, showed that probiotic administration significantly reduced body weight by <sup>−</sup>0.60 kg and BMI by <sup>−</sup>0.27 kg/m<sup>2</sup> [20]. NHANES's data between 1999–2004 also showed that yogurt consumption was associated with a lower likelihood of having obesity (OR: 0.57, 95% CI 0.40–0.82; *p* < 0.05) [21]. A prospective study including 120,877 US individuals evaluated lifestyle factors and weight change at four-year intervals, with multivariable adjustments. The four-year weight change was negatively associated with yogurt ingestion [22], which further supports our results.

#### *4.2. Diabetes*

In the unadjusted analysis, probiotic supplement and yogurt consumers had lower odds of having diabetes and, accordingly, lower glycemia and HbA1c levels. However, when adjusted for individual characteristics and confounders, the difference was no longer significant. One possible explanation for this is that diabetes is largely determined by individual demographic characteristics, such as age, ethnicity and, mainly, BMI [23]. Therefore, there are no longer differences after adjusting for individual factors. Furthermore, in another meta-analysis of randomized controlled trials, Ruan I et al. [14]. concluded that probiotics had a greater effect on fasting blood glucose for people with diabetes. On the contrary, those without diabetes only show a trend of a glucose-lowering effect, which shows that probiotic supplementation may have a greater benefit for individuals with higher fasting glucose levels [14]. In three prospective cohorts in the US, yogurt intake was consistently and inversely associated with type 2 diabetes risk [24]. Probiotic supplementation seems to be more effective in reducing HbA1c in diabetic patients with higher baseline BMIs and, furthermore, probiotic supplements with greater bacterial species may be more effective [25], which we could not evaluate in our study. Yao K et al. [26] conducted a meta-analysis in patients with type 2 diabetes to investigate the effects of probiotics on glucose metabolism, and, similar to our results, they did not find a significant effect on fasting blood glucose levels, although HbA1c was improved with probiotic supplementation. In the analysis according to the origin of probiotics, we found a non-significant trend to higher odds of diabetes in participants exposed to probiotic supplements, which was not seen in participants exposed to yogurt. This may be explained by a reverse-causality relationship. Participants with diabetes may be more prone to consuming probiotic supplements due to their known potential to improve glucose control in diabetes.

#### *4.3. Dyslipidemia*

We did not find differences in the odds of dyslipidemia according to probiotic ingestion, however, we did observe some beneficial aspects in the lipid profile of the probiotic-exposed group. HDL was significantly increased and triglycerides were significantly decreased in the probiotic group, even after adjustment. Our results are in line with a study by Fu et al. [27], with 893 subjects, which showed that gut microbiota was associated with a 4.5% variance in BMI, a 6% variance in blood levels of triglycerides, and a 4% variance in HDL, but had little effect on LDL or total cholesterol. In contrast, a meta-analysis including 1624 participants (828 in the probiotic and 796 in the placebo group) demonstrated that probiotics reduced total cholesterol and LDL cholesterol by 7.8 mg/dL and 7.3 mg/dL, respectively but had no significant effects on HDL cholesterol or triglycerides [28]. Human clinical studies have yielded different results on the association of lipid profiles with probiotic supplementation. Differences in the type of probiotics and in the experimental designs, including the clinical heterogeneity of participants, namely their baseline levels of blood lipids, may affect the role of probiotics in lipid metabolism and explain the different results. Given the clinical correlation between obesity and related metabolic disorders, it is possible that the observed associations between gut bacterial composition and lipid levels can be mediated, in part, through the effects on BMI.

#### *4.4. Hypertension*

Ingestion of probiotic supplements or yogurt resulted in a 21% reduction in the odds of hypertension after adjusting for potential confounders. Both systolic BP and diastolic BP were significantly lower. A meta-analysis of randomized clinical trials supports our results, showing that probiotic consumption significantly reduced systolic BP by −3.56 mmHg and diastolic blood pressure by −2.38 mmHg, compared with control groups [29]. In our study, for the adjusted model, probiotic ingestion was associated with a lower systolic BP by 1.48 mmHg and a lower diastolic BP by 0.86 mmHg. The modulation of BP by probiotics may be linked to several mechanisms, including their capacity to improve lipid profiles, to reduce BMI, and to produce bioactive peptides with angiotensin-converting inhibitory activity [30–32].

#### *4.5. Strengths and Limitations*

Our study has a number of strengths and limitations, which need to be highlighted. We conducted an analysis of a large cross-sectional survey, which was representative of the US population. To the best of our knowledge, this is the first cross-sectional study aiming to assess the association of probiotic ingestion, through supplements or yogurt, and metabolic disturbances. A previous study on the NHANES population was carried out, aiming to evaluate the association between dairy products (which included yogurt ingestion) with obesity and other disturbances of metabolic syndrome; however it did not evaluate probiotic supplementation [21]. We also developed an analytic strategy that included adjusted logistic regression models to obviate confounders such as physical activity, alcohol consumption, and smoking status. One of the limitations of this type of analysis is that we cannot deduce causation; however, we were able to show strong associations between probiotic ingestion and the prevalence of some metabolic disturbances. The participants in the probiotic-exposed group were more likely to be non-Hispanic white. It has been previously stated that gut microbiome varies by geographic ancestry [33], which may limit the extrapolation of these results to other ethnic groups. Furthermore, inter-individual differences in the composition of gut microbiota were previously associated with different responses to probiotics [34], including non-responders to gut microbiota modulation. The absence of individual gut microbiota analyses in our study may have hampered the analysis of these type of responders. The assessment of probiotic exposure was based on self-reported information, however, NHANES only provides dietary information that is considered to be reliable [35]. In our study, we defined probiotic ingestion as being either yogurt or probiotic supplement consumption. Assuming that there could be differences between the types of ingestion,

we performed a sensitivity analysis based on the origin of probiotics. There were no significant differences in associations according to the source of exposure to probiotics. Probiotic supplements and yogurts vary in the amounts of bacteria per serving they are composed of. In our study, the population was classified according to whether or not they were exposed to probiotic supplements or yogurt. The duration and quantity of exposure were not taken into account, which may have diluted the magnitude of the association between probiotic consumption and metabolic disturbance.

In summary, our study supports the beneficial association of probiotic supplement or yogurt ingestion with metabolic health, specifically obesity and hypertension. Furthermore, probiotic ingestion was significantly associated with higher HDL cholesterol and lower triglyceride levels. Our study supports the possibility of gut microbiota modulation by the use of probiotics as an attractive therapeutic target to prevent and treat obesity and related cardiometabolic disorders. Future research should focus on understanding the gut microbiota ecosystem and on identifying individuals who benefit the most from selective modulation of microbiota.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6643/11/7/1482/s1. Figure S1: Flowchart of the study population; Table S1: Probiotic supplements included; Table S2: Variables included in the adjusted models; Table S3: Odds ratio of disease in subjects exposed to probiotics compared to non-exposed (model 3); Table S4: Variation of cardiometabolic parameters in participants exposed to probiotics compared to non-exposed (model 3).

**Author Contributions:** Conceptualization, E.L. and J.S.N.; methodology, E.L., J.S.N., and M.F.-M.; validation, P.F. and D.C.; formal analysis, J.S.N. and M.F.-M.; writing—original draft preparation, E.L.; writing—review and editing, P.F. and D.C.

**Funding:** This research received no external funding.

**Acknowledgments:** All data sets used in the present study were provided by the U.S. Centers for Disease Control and Prevention as part of the ongoing National Health and Nutrition Examination Survey (NHANES).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*

## **Aerobic Exercise Training with Brisk Walking Increases Intestinal Bacteroides in Healthy Elderly Women**

### **Emiko Morita 1,2, Hisayo Yokoyama 1,3,\*, Daiki Imai 1,3, Ryosuke Takeda 3, Akemi Ota 4, Eriko Kawai 1, Takayoshi Hisada 5, Masanori Emoto 6, Yuta Suzuki 1,3 and Kazunobu Okazaki 1,3**


Received: 19 March 2019; Accepted: 16 April 2019; Published: 17 April 2019

**Abstract:** This study examined the effect of an exercise intervention on the composition of the intestinal microbiota in healthy elderly women. Thirty-two sedentary women that were aged 65 years and older participated in a 12-week, non-randomized comparative trial. The subjects were allocated to two groups receiving different exercise interventions, trunk muscle training (TM), or aerobic exercise training (AE). AE included brisk walking, i.e., at an intensity of ≥ 3 metabolic equivalents (METs). The composition of the intestinal microbiota in fecal samples was determined before and after the training period. We also assessed the daily physical activity using an accelerometer, trunk muscle strength by the modified Kraus–Weber (K-W) test, and cardiorespiratory fitness by a 6-min. walk test (6MWT). K-W test scores and distance achieved during the 6MWT (6MWD) improved in both groups. The relative abundance of intestinal *Bacteroides* only significantly increased in the AE group, particularly in subjects showing increases in the time spent in brisk walking. Overall, the increases in intestinal *Bacteroides* following the exercise intervention were associated with increases in 6MWD. In conclusion, aerobic exercise training that targets an increase of the time spent in brisk walking may increase intestinal *Bacteroides* in association with improved cardiorespiratory fitness in healthy elderly women.

**Keywords:** intestinal microbiota; intestinal *Bacteroides*; cardiorespiratory fitness; trunk muscle training; aerobic exercise training; brisk walking

#### **1. Introduction**

"All disease begins in the gut"., a quotation from the ancient Greek physician Hippocrates, highlights the potential roles of intestinal microbiota in various disease risks, which have recently attracted considerable attention from researchers. The presence of an imbalanced, low-diversity, intestinal microbiota is known as dysbiosis and it is associated with a variety of pathologies, including constipation [1], obesity [2], diabetes [3], colon cancer [4], coronary artery disease [5], inflammatory bowel disease [6], and depression [7]. Aging also strongly affects the composition of the intestinal microbiota. In general, the intestinal microbiota of the elderly show reduced species diversity [8]. In addition, intestinal *Bifidobacterium* and *Bacteroides*, which are known to be related to obesity, are also reduced [8], which potentially contributes to the high prevalence of obesity in the elderly population. Overall, the intestinal microbiota could be regarded as an indicator of host health.

Multiple factors, including host genetics [9], method of childbirth (i.e., by vaginal delivery or caesarian section) [10], age [8], nutrition [11], and antibiotic intake [8], have been suggested to affect the composition of the intestinal microbiota. Recent studies demonstrated the association between exercise training, i.e., a low-cost health strategy, and lower risks of colon cancer [12,13], a disease that is known to at least partly arise from imbalanced intestinal microbiota [4]. Therefore, exercise may also have potential for modifying the composition of the intestinal microbiota, although these studies did not directly examine the effect of exercise on intestinal microbiota.

In fact, animal studies have demonstrated the changes in the composition of the intestinal microbiota by exercise training [14–16]. A number of cross-sectional human studies have confirmed the associations between physical activity or cardiorespiratory fitness and the composition of the intestinal microbiota [17–19]. For example, rugby players were found to have a greater diversity of intestinal microbiota and an enlarged abundance of *Akkermansia*—which is known to prevent diabetes—when compared to sedentary adults [17]. Other studies showed that cardiorespiratory fitness or physical activity level is associated with greater microbial diversity in healthy humans [18–20]. Furthermore, trained elite race walkers show increased relative abundance of *Bacteroides,* in combination with high fat diet [21]. However, these studies did not examine the effect of exercise alone on intestinal microbiota independent of the dietary habits that may have the greater impact on intestinal microbiota than exercise. Therefore, the potential impact of exercise interventions on human intestinal microbiota has not been fully clarified.

In the present study, we examined the effects of exercise interventions on intestinal microbiotic composition in healthy elderly women. We hypothesized that an improvement of cardiorespiratory fitness would be crucial to exercise-induced changes in the intestinal microbiota. We compared the effects of two exercise modalities on the intestinal microbiota: aerobic exercise, which specifically enhances cardiorespiratory fitness, and trunk muscle training as a control condition to verify this hypothesis.

#### **2. Materials and Methods**

#### *2.1. Subjects*

Thirty-two healthy sedentary women that were aged 65 years and over were recruited from the residents of Osaka City, Japan, by an advertisement in a local magazine. The selected 32 subjects voluntarily opted for enrollment in either of the two exercise programs, aerobic exercise training (AE) or trunk muscle training (control condition; TM). Prior to the study, none of the subjects engaged in a regular exercise for more than 1 h per week. Health status and the use of medication were assessed by structured interview. Applicants presenting a history of ischemic heart disease, chronic heart failure, stroke, severe hypertension, diabetes, or neuropsychiatric disorder were excluded from the study. Applicants who were judged by a physician to be unable or ill-equipped to participate in the exercise program were also excluded. Consequently, none of the 32 subjects was excluded. The Institutional Review Board of Osaka City University Graduate School of Medicine approved the study protocol (approval no. 3501, approved on August 30, 2016). The authors also confirm that all of the ongoing and related trials for this intervention are registered in the University Hospital Medical Information Network Clinical Trials Registry (UMIN 000023930). Written informed consent was obtained from all of the participants after explanation of the study purpose. The study protocol also conformed to the ethical guidelines of the 1975 Declaration of Helsinki.

#### *2.2. Study Design*

The study design involved a 12-week non-randomized, comparative trial, in which the allocation of the participants to either of the two exercise groups, AE and TM, was based on their own preference. This study was conducted between the first recruitment of the participants on 12 September 2016 and the final follow-up of the participants on 24 January 2018. Before study enrollment, all of the applicants visited our research center at Osaka City University for baseline measurements, e.g., body composition, motor ability, and clinical laboratory analyses, as well as an assessment of daily physical activity levels, nutrient intake, and bowel habits. In addition, fecal samples were collected. All of the baseline assessments were conducted at least 1 week before the first training session. Finally, 18 and 14 applicants who met the inclusion criteria were enrolled in the AE group and the TM group, respectively, after which they were started on the selected 12-week exercise programs. The measurements during the baseline session were repeated at least one week after the final session of the exercise program.

#### *2.3. Exercise Intervention*

The subjects in the TM group received a 1-h group training weekly for 12 weeks, which aimed at strengthening the trunk muscles. All of the sessions were held at Sumiyoshi Sports Center, a gymnasium located in Osaka City, and supervised by a trained instructor. A training session comprised 5–10 min. of warm-up, followed by 45 min. of targeted resistance training of the trunk muscles and 5–10 min. of cool down exercises. Figure 1 shows examples of the trunk muscle training. The training was composed of several kinds of exercises, including arching–swaying, plank, pelvic rotation in the supine position, and diagonal lifting while standing on all fours. The contraction duration was set at 3 to 5 s, and each exercise was performed in two sets of 10 repetitions. The subjects were also instructed to work out at home daily. Adherence to group sessions, as well as to the home exercises, was recorded weekly by the instructor throughout the 12-week intervention period.

**Figure 1.** Exercises during trunk muscle training. (**A**) Arching–swaying while standing on all fours, (**B**) plank, (**C**) lying pelvic rotation, and (**D**) diagonal lifting while standing on all fours.

The subjects in the AE group were instructed to perform 60 min. of brisk walking at an intensity of <sup>≥</sup> 3 metabolic equivalents (METs) daily for 12 weeks. They wore a three-axis accelerometer (Mediwalk® MT-KT02DZ, TERUMO, Tokyo, Japan [22,23]) throughout the intervention period, except while sleeping and bathing, to record their daily number of steps and time that is spent in brisk walking. The instructor shared the accelerometer data with the participants once a week and was encouraged them to increase the intensity and duration of their brisk walking regimen gradually as much as possible. The subjects were also instructed to keep good posture while walking.

#### *2.4. Analysis of Intestinal Microbiota*

The fecal samples were collected in a container with guanidine thiocyanate as a preservative solution (TechnoSuruga Laboratory, Shizuoka, Japan) and refrigerated at 4 ◦C until transfer to the laboratory within seven days. We conformed to the protocol [24] for the representative extraction of DNA from bacterial populations in feces. Terminal restriction fragment length polymorphism (T-RFLP) analyses to determine the relative abundance of intestinal microbiota phylogenetic groups from each fecal sample were performed at the TechnoSuruga Laboratory (Shizuoka, Japan) [25,26]. T-RFLP analysis is one of the most well-established and reliable 16S ribosomal RNA-based methods, especially when considering its high throughput and reproducibility. Briefly, the fecal samples (approximately 4 mg each) were suspended in a 1200 μL solution containing 100 mM Tris-HCl (pH 9.0), 40 mM ethylenediaminetetraacetic acid, 4 M guanidine thiocyanate, and 0.001% bromothymol blue. A FastPrep 24 device homogenized the Fecal solids in the suspension (MP Biomedicals, Irvine, CA, USA) with zirconia beads being set at 5 m/s for 2 min. DNA was then extracted from a 200 μL suspension using magLEAD 12gC (Precision System Science; Chiba, Japan). MagDEA® Dx SV (Precision System Science) was used as the reagent in automatic nucleic acid extraction. PCR was performed with a Takara Thermal Cycler Dice TP650 (Takara Bio, Shiga, Japan) in 20 μL of a reaction mixture containing 1× PCR buffer, with each deoxynucleotide triphosphate at a concentration of 200 μM, 1.5 mM MgCl2, each primer at a concentration of 0.2 μM, 10 ng of fecal DNA, and 0.2 U of HotStarTaq DNA polymerase (Qiagen, Hilden, Germany). 5 FAM-labeled 516f (5'-TGC-CAGCAGCCGCGGTA-3'; *Escherichia coli* positions 516−532) and 1510r (5'-GGTTACCTTGTTACGA-CTT-3'; *E. coli* positions 1510−1492) were the primers used. The amplification program used was as follows: preheating at 95 ◦C for 15 min, 35 cycles of denaturation at 95 ◦C for 30 s, annealing at 50 ◦C for 30 s, extension at 72 ◦C for 90 s, and finally, terminal extension at 72 ◦C for 10 min. Electrophoresis and purified using a MultiScreen PCR μ96 Filter Plate verified amplified DNA (Millipore, Billerica, MA, USA). The purified 16S rDNA amplicons were treated with 10 U of FastDigest BseLI (Thermo Fisher Scientific, Waltham, MA, USA) for 10 min. An ABI PRISM 3130xl genetic analyzer (Thermo Fisher Scientific) was used to analyze the resultant DNA fragments, i.e., fluorescent-labeled terminal restriction fragments (T-RFs). GeneMapper software (Thermo Fisher Scientific) was used to determine the T-RF length and the peak area for each sample. T-RFs were divided into 29 operational taxonomic units (OTUs). The individual OTUs were quantified as the percentage of all OTUs combined based on the area under the curve (% AUC). The reference database, Human Fecal Microbiota T-RFLP profiling (http://www.tecsrg-lab.jp/t\_rflp\_hito\_OTU.html), was used to putatively match the bacteria in each classification unit to the corresponding OTU. T-RFLP analyses enabled the classification of the sampled intestinal microbiota into the following 10 groups: *Bifidobacterium*, *Lactobacillales*, *Bacteroides*, *Prevotella*, *Clostridium* cluster IV, *Clostridium* subcluster XIVa, *Clostridium* cluster IX, *Clostridium* cluster XI, *Clostridium* cluster XVIII, and others.

#### *2.5. Anthropometrical Measurements*

The body mass index (BMI) was calculated as body weight/(height)2, as expressed in kg/m2. Bioelectrical impedance analysis using a body composition analyzer estiated the percentages of fat and muscle mass of the trunk and lower extremities (Nippon Shooter Ltd., Physion MD, Tokyo, Japan).

#### *2.6. Physiological Performance*

Quadriceps muscle strength was assessed using a strain gage dynamometer (ST-200S, MUL-TECH, Tokyo, Japan). Each subject performed two attempts on each leg and the maximum value of these four trials was marked for later analysis. The modified Kraus–Weber (K-W) test was used to assess trunk muscle strength [27]. This simple exercise test was based on the K-W Minimum test that was

developed by Drs. Hans Kraus and Sonja Weber in the 1950s [28] to assess the strength and endurance of the trunk muscles. The trunk muscle strength of each subject was rated based on the total scores (full marks = 40) of the test (Supplementary Figure S1).

Four physical performance tests were conducted to evaluate motor ability and fitness: maximal step length (MSL), Timed Up and Go (TUG) test, single-leg standing, and the 6-min. walk test (6MWT). MSL was determined as the maximum possible stride per step of a subject. In the TUG test, we measured the time that is required for a subject to stand up from a chair, walk 3 m, turn, walk back to the chair, and sit down. In single-leg standing, we measured the maximum time that a subject could stand on one leg. In case a subject continued single-leg standing for over 120 s, the test was discontinued. All of the functional tests were conducted twice and the best scores were marked for analysis. Cardiorespiratory fitness was evaluated by the 6MWT according to the guidelines of the American Thoracic Society [29]. In short, he subjects were instructed to walk back and forth on a 25-m course as fast as possible for 6 min under the supervision of a medical doctor. They were permitted to stop and rest in case of fatigue. The investigator encouraged the subjects with routine phrases (e.g., "you are doing well" and "keep up the good work") once per minute during the test. The total distance (in meters) walked after 6 min. (6MWD) was recorded and used as an indicator of cardiorespiratory fitness, since performance on the 6MWD strongly correlates with peak oxygen uptake [30,31].

#### *2.7. Daily Physical Activity Level*

The parameters reflecting the daily physical activity level of the participants included the number of steps and the time spent in brisk walking—i.e., at an intensity of three METs or more—was estimated using the same three-axis accelerometer as that used during training in the AE group. This device also automatically calculates ethe nergy expenditure (EE) from METs based on a widely-accepted formula (EE (kcal) = 1.05 × METs × time (h) × body weight (kg) [32]). All of the subjects were instructed to wear the accelerometer throughout the one-week measurement period, except while sleeping and bathing, and to continue with daily activities as usual. The assessments were conducted before and after the 12-week intervention. The data, which were automatically stored on the device, were subsequently transferred to a computer while using specialized software (HR Joint® Smile Data Vision, TERUMO, Tokyo, Japan). The mean daily values of all parameters recorded during the one-week monitoring period were used for further analysis.

#### *2.8. Laboratory Measurements*

The blood samples were collected at 9 AM under standardized 12-h fasting conditions. Serum samples were stored at −80 ◦C until further analysis. The hexokinase UV method measured the plasma glucose levels, whereas serum insulin levels were determined by chemiluminescent enzyme immunoassay. Serum triglycerides, low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C) were determined by enzymatic methods. The homeostasis model assessment of insulin resistance (HOMA-IR), which is an established surrogate index of insulin resistance [33], was also determined. The HOMA-IR was obtained from fasting plasma glucose (FPG) and serum insulin (FIRI) levels according to the original method by Matthews et al. [34] while using the following formula:

$$\text{HOMA-IR} = \text{FPG (mmol/L)} \times \text{FIRI (\mu\text{U/mL})} / 22.5 \tag{1}$$

A higher HOMA-IR value represents higher insulin resistance.

#### *2.9. Nutrient Intake*

Nutrient intake was estimated using a food frequency questionnaire (FFQ), which the Japan Public Health Center-based Prospective Study developed and previously validated [35]. The FFQ consists of 138 food and beverage items and measures nine intake frequency categories: never or seldom, 1–3 times/month, 1–2 times/week, 3–4 times/week, 5–6 times/week, once/day, 2–3 times/day, 4–6 times/day, and more than seven times/day. All of the subjects were asked to complete the questionnaire before and after the 12-week intervention. FFQ data were analyzed with the help of Education Software Co., Ltd. (Tokyo, Japan), and then converted to quantitative estimates of the daily consumed amounts of energy, protein, lipid, carbohydrates, saturated fat, and dietary fiber.

#### *2.10. Defecation Assessment*

Defecation patterns were assessed using the Japanese version of the Constipation Assessment Scale (CAS-J), which was modified from the original scale that was developed by McMillan et al. [36] to assess constipation in Japanese populations [37]. The CAS-J comprises eight questions, i.e., "The abdomen appears distended or swollen", "The amount of flatus", "The frequency of defecation" "The rectum appears to be filled with feces", "Pain of the anus during defecation", "The amount of feces", "Ease of defecation", and "Diarrhea or watery stools". Each item includes a three-point rating scale: 0 ("no problem"), 1 ("some problem"), and 2 ("severe problem"). Thus, the maximum possible CAS-J score is 16, with higher scores indicating more severe cases of constipation.

#### *2.11. Statistical Analyses*

The data are presented as median and interquartile ranges. Changes in clinical parameters and relative abundances of specific classes of intestinal microbiota following intervention in each group were examined by the Wilcoxon Signed-rank test. The Spearman's rank correlation coefficient test examined the relationships between the parameters and changes in the relative abundance of specific types of intestinal microbiota. Stepwise regression analysis was also performed to identify the factors that determined the change in the relative abundance of specific microbiota. Finally, the Mann–Whitney U-test was used to compare the changes in the relative abundance of specific types of intestinal microbiota between the exercise groups according to the increase in time spent in brisk walking. All of the statistical procedures were performed using SPSS statistical software (version 24.0, IBM, New York, NY, USA). *P* values less than 0.05 were considered to be statistically significant.

#### **3. Results**

#### *3.1. Clinical Characteristics of the Subjects*

Figure 2 shows the procedural flowchart of the enrollment, measurement, intervention, and data analysis of this study. Two participants in the TM group and one in the AE group dropped out during the intervention period. A total of 12 participants in the TM group and 17 participants in the AE group completed the study. We could confirm that all of the subjects in the TM group participated in 90% or more of the sessions and that the mean adherence to the home exercise was 96.0%. The mean percentage of attendance at weekly meetings with the instructor was 97.1% in the AE group. Table 1 summarizes the clinical characteristics of both groups. The median age was 70 (65–77) years in the TM group and 70 (66–75) years in the AE group.

**Figure 2.** Flowchart of the screening, enrollment, intervention, and data analysis of the study. Abbreviations: TM, trunk muscle training; AE, aerobic exercise training.


**Table 1.** Clinical characteristics of the subjects.

Data are presented as median (interquartile range) for age, BW, body fat, SBP, and DBP, and as *n* (%) for present illness, past history, and medication. Abbreviations: TM, trunk muscle training; AE, aerobic exercise training; BW, body weight; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure.

#### *3.2. Changes in Body Composition, Muscle Strength, Physical Performance, and Daily Physical Activity Following the Intervention*

Table 2 shows the changes in body composition, muscle strength, physical performance, and daily physical activity following the intervention in both groups.


**Table 2.** Changes in the parameters following the intervention.

All values are presented as median (interquartile range). Changes in clinical parameters following intervention in each group were examined by the Wilcoxon Signed-rank test. \*: *p* < 0.05 compared with baseline. Abbreviations: TM, trunk muscle training; AE, aerobic exercise training; BMI, body mass index; K-W test score, Kraus–Weber test score; Quad. muscle strength, Quadriceps muscle strength; MSL, maximal step length; TUG, Timed Up & Go; 6MWD, distance in the 6-min. walk test; EE, energy expenditure; FPG, fasting plasma glucose; TG, triglyceride; LDL-C, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; HOMA-IR, homeostatic model assessment of insulin resistance.

The number of steps (*p* = 0.004) and the time spent in brisk walking (*p* = 0.003), as well as the exercise-induced EE (*p* = 0.003), were significantly increased following the intervention in the AE group only. Total EE was significantly increased in the AE group (*p* = 0.012), while it was decreased in the TM group (*p* = 0.049) following the intervention. The K-W test scores (TM group: *p* = 0.008; AE group: *p* < 0.001) and 6MWD (TM group: *p* = 0.028; AE group: *p* = 0.001) were equally improved following the intervention in both groups. No further significant changes were observed in other parameters of motor ability following the interventions in either group.

#### *3.3. Changes in Laboratory Measurements Following the Intervention*

Table 2 shows the changes in the laboratory measurements following the intervention in both groups. FPG and blood levels of triglycerides, LDL-C, HDL-C, and insulin as well as HOMA-IR remained unchanged after the intervention in both groups.

#### *3.4. Changes in Nutrient Intake and Defecation Pattern Following the Intervention*

Table 3 shows the changes in nutrient intake and the CAS-J scores following the intervention. Significant differences in nutrient intake patterns as well as total energy intake were found neither at baseline nor after the interventions. Regarding the patterns of defecation, the CAS-J scores were significantly decreased in the AE group only (*p* = 0.036) following the interventions. For individual components of the CAS-J, the score on "Ease of defecation" was significantly decreased following the intervention in the AE group. On the other hand, the score in "The rectum appears to be filled with feces" was significantly improved following the intervention in the TM group only.


**Table 3.** Changes in nutrient intake and defecation pattern following the intervention.

All values are presented as median (interquartile range). Changes in clinical parameters following intervention in each group were examined by the Wilcoxon Signed-rank test. \*: *p* < 0.05 compared with baseline. Abbreviations: TM, trunk muscle training; AE, aerobic exercise training; CAS-J, Japanese version of the Constipation Assessment Scale.

#### *3.5. Composition of Intestinal Microbiota*

Figure 3 shows the composition of the intestinal microbiota in both groups. Following the interventions, the relative abundance of *Bacteroides* was significantly increased, and that of *Clostridium* subcluster XIVa was only decreased in the AE group. The relative abundance of *Clostridium* cluster IX was only significantly increased in the TM group. After the interventions, the relative abundance of other microbiota groups remained unchanged in both of the groups.

**Figure 3.** Changes in the composition of the intestinal microbiota following the intervention. The relative abundance of intestinal *Bacteroides* was significantly increased, and that of the *Clostridium* subcluster XIVa was decreased only in the AE group (by the Wilcoxon Signed-rank test). The relative abundance of *Clostridium* cluster IX was significantly increased only in the TM group. \*: *p* < 0.05 compared with baseline. Abbreviations: TM, trunk muscle training; AE, aerobic exercise training.

#### *3.6. Relationship between Changes in the Parameters and Change in the Relative Abundance of Intestinal Bacteroides after the Intervention*

We examined the relationships between age, the relative abundance of intestinal *Bacteroides* before the intervention (pre-*Bacteroides*), or the changes in the parameters that were modulated by the exercise intervention and the change in the relative intestinal abundance of *Bacteroides* (Δ%*Bacteroides*). Pre-*Bacteroides* was negatively correlated with Δ%*Bacteroides* (*r* = −0.519, *p* = 0.004) when analyzing all of the subjects combined. A significant positive correlation was also found between the change in 6MWD (Δ6MWD; *r* = 0.431, *p* = 0.020) or that in time spent in brisk walking (ΔTime spent in brisk walking; *r* = 0.371, *p* = 0.047) and the Δ%*Bacteroides* following the intervention in all subjects (Figure 4). There were no significant correlations between changes in other parameters and Δ%*Bacteroides* among all the subjects combined (Table 4).

To identify the factors that contribute to Δ%*Bacteroides*, we performed stepwise multiple regression analysis, in which Δ%*Bacteroides* was included as the dependent variable and age, pre-*Bacteroides*, Δ6MWD, and ΔTime spent in brisk walking were included as the possible independent variables. In this analysis, Δ6MWD (β = 0.370, *p* = 0.034) and pre-*Bacteroides* (β = −0.356, *p* = 0.041) were found to be independent contributors (*R*<sup>2</sup> = 0.317).


**Table 4.** Correlation coefficients in simple regression analysis between clinical factors and the changes in the relative abundance of intestinal *Bacteroides* following the exercise intervention in all subjects.

The relationships between the parameters and changes in the relative abundance of specific types of intestinal microbiota were examined by Spearman's rank correlation coefficient test. \*: *p* < 0.05. Abbreviations: Pre-*Bacteroides*, the relative abundance of the intestinal *Bacteroides* before the intervention; K-W test score, Kraus–Weber test score; 6MWD, distance in the 6-min. walk test; EE, energy expenditure; CAS-J, Japanese version of the Constipation Assessment Scale.

**Figure 4.** Relationship between changes in the distance during the 6MWT (6MWD) (**A**), changes in the time spent in brisk walking (**B**), and changes in the relative abundance of intestinal *Bacteroides* by the intervention. Improvements in 6MWD and time spent in brisk walking were positively correlated with increases in the relative abundance of intestinal *Bacteroides* in all subjects. Abbreviations: TM, trunk muscle training; AE, aerobic exercise training; 6MWT, 6-min. walk test; 6MWD, distance in the 6MWT.

#### *3.7. E*ff*ect of Increased Daily Physical Activity on Changes in the Relative Abundance of Intestinal Bacteroides Following the Intervention in the AE Group*

Although the improvement of 6MWD was observed in each group, a significant increase in the relative abundance of intestinal *Bacteroides* was only found in the AE group. Therefore, we focused on the effect of the increased time spent in brisk walking on Δ%*Bacteroides* following the intervention in the AE group. The subjects in the AE group were divided into two groups according to whether they had increased their time spent in brisk walking by more or less than 20 min. following the intervention. As shown in Figure 5, Δ%*Bacteroides* in subjects who added > 20 min. of time spent in brisk walking was greater than that in the subjects who added ≤ 20 min. (9.7% (4.7%–14.2%), *n* = 10 vs. –3.5% (–4.2% – 2.4%), *n* = 7; *p* = 0.025).

**Figure 5.** Effect of increased daily physical activity on changes in the relative abundance of intestinal *Bacteroides* following the intervention in the AE group. Increases in intestinal *Bacteroides* in subjects who increased the daily time spent in brisk walking for 20 min. or more were greater than in those who did not (by the Mann–Whitney *U*-test). Horizontal bars indicate the minimum values, the 25th, 50th, 75th percentile levels, and the maximum values. Abbreviations: AE, aerobic exercise training.

#### **4. Discussion**

The aim of the present study was to investigate whether exercise intervention modifies the composition of intestinal microbiota in healthy elderly women. Our main findings were that a 12-week aerobic exercise program that consists of daily episodes of brisk walking increased the relative abundance of intestinal *Bacteroides,* while improving cardiorespiratory fitness without any changes to nutrient intake. Moreover, the increase relative abundance of intestinal *Bacteroides* was especially marked in subjects who increased the time spent in brisk walking by more than 20 min. We also found that aerobic exercise improved the pattern of defecation independently of Δ%*Bacteroides*. Meanwhile, the elderly subjects who engaged in the trunk muscle training showed neither a significant Δ%*Bacteroides* nor changes in the defecation pattern.

To date, the primary findings of animal studies suggested that the level of exercise may modulate the composition of the intestinal microbiota. In rodents, six days of wheel running exercise increased *Bifidobacterium* and *Lactobacillus*, which are widely recognized as health-promoting intestinal bacteria [14]. It was also reported that the exercise-induced changes of the intestinal microbiota in mice depend on the exercise modalities (voluntary wheel running or forced treadmill running) and that voluntary wheel running reduced *Turicibacter* spp., which are associated with immune dysfunction and bowel diseases [15]. Another study demonstrated that a six-week schedule of interval treadmill running in mice enhanced the diversity of intestinal microbiota, with marked increases in the relative abundance of *Bacteroidetes* [16]. On the other hand, few data in humans have been published regarding the effect of exercise interventions on the intestinal microbiota. In a recent report by Allen et al. a six-week aerobic exercise training altered the intestinal microbiota differently, depending on body weight status [38]. The present study could further elaborate these results by demonstrating that a 12-week aerobic exercise program that consists of brisk walking—in contrast to training of trunk muscles—increased the relative abundance of intestinal *Bacteroides*. This suggests that aerobic exercise may beneficially modify the intestinal microbiota in healthy elderly women. In previous studies, the maintenance of an optimal intestinal environment has been shown to contribute to the prevention

of various types of diseases [1–7]. The results of our study suggest a practical approach, i.e., aerobic exercise, as a strategy to attain the optimization of the intestinal microbiota in humans.

The 12-week aerobic exercise training increased the relative abundance of intestinal *Bacteroides*, in parallel with an improvement in cardiorespiratory fitness. Interestingly, a cross-sectional study has shown that cardiorespiratory fitness is associated with a larger proportion of *Bacteroides* in the intestinal microbiota of premenopausal women [18]. The results from our interventional study are consistent with—and augment—the significance of this observation. We also demonstrated that increases in the relative abundance of *Bacteroides* in the large intestine were greater in the subjects who improved the daily time spent in brisk walking at an intensity of ≥ 3 METs by more than 20 min. On the basis of the findings of the Nakanojo Study, brisk walking at an intensity of > 3 METs for 20 min. or more on most days are recommended for the elderly to reduce the risk of lifestyle-related diseases [39]. In particular, such exercise levels reduce the incidence of osteoporosis [40], metabolic syndrome, hypertension, and hyperglycemia [41]. Based on these considerations, we initially set the target volume of brisk walking for the AE group in our study at 20 min. per day. Our results demonstrate that this regimen also effectively modifies and optimizes the composition of the intestinal microbiota.

By contrast, trunk muscle training did not change the composition of the intestinal microbiota in subjects within our TM group, although it did improve the cardiorespiratory fitness. An improvement in cardiorespiratory fitness in the TM group may have resulted from the strengthening of the respiratory muscles by the trunk muscle training [42]. It may have also occurred because the subjects commuted to the sports center once weekly during the study period. However, the improvement in the cardiorespiratory fitness in the TM group did not coincide with a changed composition of the intestinal microbiota. Furthermore, in the present study, the increase in the time spent in brisk walking was positively correlated with the increase in the relative abundance of intestinal *Bacteroides*. To put these result into perspective, cardiorespiratory fitness may need to be improved by aerobic exercise, such as brisk walking, when the goal is to modify the intestinal microbiota.

There are some candidate mechanisms by which aerobic exercise might increase intestinal *Bacteroides*. Changes in the colonic transit time result in changes in pH within the colonic lumen that may be key in affecting the composition of the intestinal microbiota. Prolonged colonic transit time is known to limit the diversity of intestinal microbiota [43], and this coincides with a greater rise in pH during transit from the proximal to the distal colon [44]. Aerobic exercises, such as jogging and cycling at a moderate intensity, decrease intestinal transit time in healthy people [45] as well as middle-aged patients with chronic constipation [46], probably via increases in the visceral blood flow, increased release of gastrointestinal hormones, mechanical stimulation, and strengthening of the abdominal muscles [46]. Furthermore, aerobic exercise increases the fecal concentrations of the short-chain fatty acids (SCFA) [47], which slightly lowers the colonic-luminal pH [48]. *Bacteroides* species prefer mildly acidic conditions (pH 6.7) for their survival in the colonic lumen, whereas they grow poorly at pH 5.5 [49]. This may explain why aerobic exercise increases intestinal *Bacteroides*, although, of course, a more detailed analysis of the underlying factors remains necessary.

*Bacteroides* species are opportunistic bacteria: whether they positively or negatively affect host health depends on the characteristics of their intestinal environment. *Bacteroides* spp. play a role in protecting against inflammatory bowel disease [50,51], whereas they may increase infants' susceptibility to chronic allergic disease, such as early-onset atopic eczema [52]. Thus, future studies will need to detail the clinical consequences of aerobic-exercise-induced increases in *Bacteroides*. Nonetheless, it is widely accepted that lower levels of *Bacteroides* are associated with the higher prevalence of obesity and metabolic syndrome and that *Bacteroides* species may help in suppressing metabolic dysfunction [53,54], although we unfortunately could not evaluate waist circumference as a surrogate index of visceral fat accumulation in the present study. However, in the present study, the increase in intestinal *Bacteroides* in the AE group did not decrease insulin resistance, as assessed by HOMA-IR. This may be because the sedentary but healthy subjects that were included in the present study presented with good insulin sensitivity at baseline, making further improvements in insulin sensitivity following exercise difficult

to attain. Further studies should clarify whether aerobic exercise might improve insulin sensitivity through the increase of intestinal *Bacteroides* in obese and/or insulin-resistant subjects.

It is widely accepted that *Bifidobacteria* and *Lactobacillales* contribute to intestinal health, preventing diarrhea and various infectious, allergic, and inflammatory conditions [55]. The relative abundances of these bacteria are decreased in elderly people [8], which may at least partly result in intestinal barrier dysfunction in the population [56]. In addition to some factors, such as probiotics and dietary fiber [56,57], vigorous exercise also has the potential for increasing these bacteria based on rodent studies [14]. However, in the present study, the relative abundance of *Bifidobacterium*, the only *Bifidobacteria* that can be identified by our T-RFLP analysis, as well as *Lactobacillales* remained unchanged in both groups. This may be because the quantity (time and intensity) of our brisk walking was not enough to increase these bacteria. Exploring an exercise prescription that can increase these bacteria will benefit intestinal health in elderly people.

A few limitations of the present study should be noted. First, our non-randomized study design with a relatively small subject sample size may have been insufficiently powered to detect differences in efficacy between the two exercise programs to affect the clinical outcomes, such as trunk muscle strength, cardiorespiratory fitness, defecation pattern, and the composition of the intestinal microbiota. Second, we confirmed that the participants had no substantial exercise habits before exercise intervention. However, exercise-induced energy expenditure at the baseline was greater in the AE than in the TM group. The reason for this may have been that the subjects who opted for the AE training were more aware of the health benefits of walking, which may have resulted in a superior effect of brisk walking to trunk muscle training on increasing intestinal *Bacteroides*. Finally, we classified fecal intestinal microbiota into only 10 major groups, which were present in the fecal samples of all the subjects. Therefore, it was impossivly to evaluate the effects of the exercise intervention on the diversity of the intestinal microbiota. A greater diversity of the intestinal microbiota is generally considered to provide various health benefits. It might have been possible to detect an increase in the diversity of the intestinal microbiota following the exercise intervention if the adopted microbiotic classification scheme had included several hundred subdivisions.

#### **5. Conclusions**

Aerobic exercise training targeting an increase of the time spent in brisk walking may have a potential for increasing intestinal *Bacteroides,* while also improving cardiorespiratory fitness in healthy elderly women. Exercise intervention may provide a practical means of acquiring a more optimal composition of intestinal microbiota. Further studies are needed to clarify the mechanism by which exercise exerts it effect on the composition of the intestinal microbiota.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6643/11/4/868/s1, Figure S1: Modified Kraus–Weber test.

**Author Contributions:** Conceptualization, H.Y.; Data curation, E.M., H.Y. and T.H.; Formal analysis, E.M. and H.Y.; Funding acquisition, H.Y.; Investigation, E.M., H.Y., D.I., R.T., A.O., E.K. and T.H.; Methodology, H.Y. and T.H.; Project administration, H.Y.; Supervision, H.Y. and K.O.; Writing—original draft, E.M., H.Y. and T.H.; Writing—review & editing, H.Y., D.I., M.E., Y.S. and K.O.

**Funding:** The present study was funded by Nakatomi Foundation and by TIPNESS Inc. in Tokyo, Japan. The present research was also supported by the Osaka City University (OCU) Strategic Research Grant 2016 for basic researches, and in part by the OCU "Think globally, act locally" Research Grant for Female Researchers 2017 and 2018 through the hometown donation fund of Osaka City.

**Acknowledgments:** We are grateful to TIPNESS Inc. for skillfully instructing the study subjects during the exercise programs. We would also like to thank TERUMO Co., Ltd., for providing us with Mediwalk® devices to monitor the physical activity levels of the participants in the present study. We also acknowledge Nooshin Naghavi for her skillful technical assistance.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*

### **Diet Supplemented with Antioxidant and Anti-Inflammatory Probiotics Improves Sperm Quality after Only One Spermatogenic Cycle in Zebrafish Model**

### **David G. Valcarce 1, Marta F. Riesco 1, Juan M. Martínez-Vázquez <sup>1</sup> and Vanesa Robles 1,2,\***


Received: 20 March 2019; Accepted: 11 April 2019; Published: 13 April 2019

**Abstract:** Infertility is a medical concern worldwide and could also have economic consequences in farmed animals. Developing an efficient diet supplement with immediate effects on sperm quality is a promising tool for human reproduction and for domesticated animal species. This study aims at elucidating the effect of a short-time probiotic supplementation consisting of a mixture of two probiotic bacteria with proven antioxidant and anti-inflammatory activities on zebrafish sperm quality and male behavior. For this purpose, three homogeneous groups of males in terms of motility (<60%) were established. The control group was fed with a normal standard diet. The other received supplements: One group (vehicle control) was fed with maltodextrin and the other received a probiotic preparation based on a mixture (1:1) of *Lactobacillus rhamnosus* CECT8361 and *Bifidobacterium longum* CECT7347. The feeding regime was 21 days corresponding with a single spermatogenesis in zebrafish. The preparation did not modify animal weight, positively affected the number of fluent males, increased sperm concentration, total motility, progressive motility, and fast spermatozoa subpopulations. Moreover, the animals fed with the supplement showed different behavior patterns compared to control groups. Our results suggest a diet-related modulation on the exploration activity indicating a lower stress-like conduct. The studied formulation described here should be considered as advantageous in male reproductive biotechnology.

**Keywords:** sperm quality; probiotics; zebrafish; motility; behavior

#### **1. Introduction**

Infertility is a highly ubiquitous global health problem. It has been recognized as a public health issue worldwide by the World Health Organization (WHO) and is predicted to affect 9% of the world population on average [1]. Altered production of functional and motile spermatozoa is a causal factor in up to 70% of infertility cases [2]. Sperm concentration has reduced by half in Western countries in the last four decades, without evidence of improvement [3]. Moreover, a high proportion of young men have sperm counts below the fertile threshold [4]. A variety of environmental factors may be contributing to reduced semen quality [4]. One of them is diet [5]. For example, there exists vast evidence regarding the adverse consequences of high-fat diets on male reproductive success [6]. Describing the dietary factors that can influence male fertility potential is of high interest. Nowadays, there is strong and consistent evidence about antioxidants as essential factors for sperm defense [7]. Selenium, vitamin E, vitamin C, folate, carotenoids, zinc, or carnitine are antioxidants naturally found in semen samples. The main function of these molecules is helping to overcome reactive oxygen species

(ROS) production from free radicals [8]. Since subfertile men have been identified as having lower levels of these scavengers in their semen [9], they have become an object of study for reproductive biologists [10].

Probiotics are living microorganisms that improve animal health status when integrated in the diet [11]. These microorganisms act by balancing the gut microbiota, regulating the intestinal transit, modulating intestinal villi, and protecting nutrient digestion and absorption. The intestinal microbiome is a complex ecosystem, which provides numerous crucial functions to the host organism [12]. During the last decades, gut microbiota (16S rRNA surveys are used to taxonomically identify the microorganisms in the environment [13]) has emerged as a key factor which regulates host metabolism and different gut microbiome phenotypes (the genes and genomes of the microbiota, as well as the products of the microbiota and the host environment [13]) have been associated with diseases [14]. Therefore, being able to regulate intestinal microbiota is of huge interest for scientists due to the potential implications in several fields of knowledge. To date, a myriad of species from *Bacillus*, *Enterococcus*, *Lactococcus*, *Streptococcus*, *Bifidobacterium*, or *Lactobacillus* have been used as probiotics [15], with the last two genera being the most used for this purpose [16].

Nowadays, zebrafish (*Danio rerio*) is accepted by the scientific community as a vertebrate model for the study of genetics, development, and diseases among others [17,18]. This teleost is also a good model for probiotic-related experiments [19] since zebrafish microbiota is comparable to that of human as well as gut colonization [20,21]. The aim of this study was to provide, by in vivo experimentation, new insights into the potential positive effects of probiotics on male reproductive biology. The finding of new ways to increase sperm quality would be useful for clinical protocols and the possible applications derived from the potential beneficial effects of probiotics on reproductive biology will be of interest not only in human reproduction, but also in animal production where nutrition is a key element.

Our hypothesis is that probiotics during a single spermatogenesis cycle can improve sperm quality as well as animal welfare. In order to validate this hypothesis, in this work we use zebrafish since it is an optimal model for reproductive biology because of its easy reproduction, low-cost maintenance, and fast cystic spermatogenesis (21 days) [22]. To verify whether a short-term exposure to probiotics has effects on sperm quality and animal behaviour, we exposed zebrafish adult males with different initial sperm quality to a multistrain probiotic combination containing two previously described bacteria with antioxidant and anti-inflammatory activities: *Lactobacillus rhamnosus* CECT8361 and *Bifidobacterium longum* CECT7347 [23,24].

#### **2. Materials and Methods**

#### *2.1. Ethics Statement*

The institutional Animal Care and Use Committee at the Marine Culture Plant El Bocal of the Spanish Institute of Oceanography in Santander (Spain) approved the experimental design and all protocols and procedures including animals (PI-10-16). All animals were manipulated in accordance with the Guidelines of the European Union Council (86/609/EU, modified by 2010/62/EU), following Spanish regulations (RD/1201/2005, abrogated by RD/2013) for the use of laboratory animals.

#### *2.2. Animals*

Wild-type zebrafish (Ab strain) were housed in the Marine Culture Plant El Bocal zebrafish platform of the Spanish Institute of Oceanography in Santander (Spain). Fish were bred and maintained according to standard protocols. In all trials, males were anesthetized in 110 mg/L buffered tricaine methane sulfonate (MS222). All efforts were made to reduce suffering and a humane endpoint was applied with a lethal dose of anesthetic if fish reached a moribund state.

#### *2.3. Visible Implant Elastomer Tags (VIE) Tagging*

VIE tags were prepared following manufacturer's indications adapted for the minimal volume (Northwest Marine Technology, Shaw Island, WA, USA). Green and red fluorescent elastomers (viscoelastic polymers) were used in the experiment. A code combining colors, number of tags, and positions (taking as reference the anteroposterior and dorsoventral axis) was generated and individually associated to a specific male in the zebrafish colony (Figure 1A). An expert hand injected small amounts of elastomers (dot shaped) in each anesthetized male. After injection, tag retention and injury evaluation (not registered) was evaluated in the recovery tanks. Health status was checked daily.

**Figure 1.** Study design summary. (**A**) Animals participating in the experiment (*n* = 40) were tracked with fluorescent visible implant elastomers. Each male carried a unique code visible under white and UV light. (**B**) Only males that reached the inclusion criteria described in the flowchart (*n* = 36) were selected for creating homogeneous experimental groups in terms of total motility. (**C**) Each group (*n* = 12) received a different diet during 21 days corresponding to a *Danio rerio* spermatogenesis cycle. "C", "M", and "P" refer to the experimental groups: Control, maltodextrin, and probiotics, respectively.

#### *2.4. Inclusion Criteria, Experimental Group Definition, Study Design, and Feeding Regimes*

Adult zebrafish males were anesthetized and sperm samples were collected and evaluated with a CASA (computed assisted sperm analysis) system (see below for procedure). Only males showing a total motility under 60% were selected for the experiment (Figure 1B). The inclusion criterion was established this way to analyze the effect of the probiotic strains mixture on diverse quality sperm samples. The cutoff value of 60% was chosen with the aim that there would be an improvement margin in the samples after the treatment. Semen samples over 60% can be considered acceptable samples in terms of motility. Males reaching the inclusion criteria were used to generate three homogeneous groups (*n* = 12) in terms of motility (Figure 1B). Each group had a different feeding regime: (1) The control group "CTRL" ingested only a commercial diet; (2) the vehicle control group "MALTO" received the commercial diet and two doses of 0.11 g of maltodextrin; and (3) the experimental group "PROBIO" received the commercial diet and a probiotic treatment consisting of a daily 10<sup>9</sup> Colony Forming Units (CFU) mixture (1:1) of lyophilized *L. rhamnosus* CECT8361 and *B. longum* CECT7347 strains carried in 0.22 g of maltodextrin. Strains were kindly provided by the company Biopolis S.L. (Valencia, Spain) and the commercial diet provided to all animals within the experiment twice a day (55% min. crude protein, 15% min. crude fat, 1.5% max. crude fiber, and 12% max. moisture) was purchased from Aquatic Animals (Apopka, FL, USA). In order to guarantee the ingestion, supplements (vehicle or probiotic mixture) were provided to experimental males in rearing water 30 min before each routine feeding. All experimental groups were held under the same conditions during all experiments. The experiment was replicated three times, including four males per experimental group each time (final population for each experimental group, *n* = 12). Sperm analysis was performed at t = 0 days and t = 21 days of each experimental replicate. Feeding regimes were maintained during the 21 days according to a spermatogenesis cycle in the species (Figure 1C).

#### *2.5. BiometricAnalysis*

At day 0 and day 21, fish weight was determined using a microbalance (Mettler MT5, Mettler Toledo, Spain).

#### *2.6. Sperm Sampling*

At 0 days and at 21 days, semen was collected approximately 1h after the lights of the zebrafish facility were turned on. Each fish was identified by checking its VIE tag code and after that, they were anesthetized one by one. Once the absence of reflexes was corroborated, the animals were gently located on a sponge, the surrounding area of the urogenital pore was dried and sperm collection was performed by abdominal massage using glass flat forceps as tools to smoothly press both sides according to routine protocols. Ejaculates were collected with a micropipette and diluted in 10 μL of buffered Hank's solution (0.137 M NaCl; 5.4 mM KCl; 0.25 mM Na2HPO4; 0.44 mM KH2PO4; 1.3 mM CaCl2; 1.0 mM MgSO4; 4.2 mM NaHCO3). The diluted samples were stored at 22ºC until analysis (5 min).

#### *2.7. CASA Sperm Analysis*

The activation of motility was performed by diluting 1 μL of sperm with 9 μL of system water (~300 mOsm/L) at 28 ◦C. Sperm motility, kinetics, and concentration were analyzed using a CASA system with ISAS software (ISAS, PROiSERR+D, S.L. Spain). Activated sperm was loaded into a Makler counting chamber (10 μm depth; Sefi Medical Instruments, Haifa, Israel). The CASA system consisted of a tri-ocular optical phase-contrast Nikon Eclipse Ts2R microscope (Nikon, Tokyo, Japan) using a 10× objective equipped with Basler A312fc digital camera (Basler Vision Technologies, Ahrensburg, Germany). The ISAS software was used with specific settings for fish spermatozoa (1 μm2 < particle area < 20 μm2; cell description according to VCL (curvilinear velocity): 10 μm/s < slow < 45 μm/s < medium < 100 μm/s < fast); and it rendered the following parameters: (1) Concentration; (2) percentage

of motile spermatozoa (MOT,%); (3) percentage of progressive spermatozoa (P-MOT,%) defined as the percentage of spermatozoa which swim forward in 80% of a straight line; (4) curvilinear velocity (VCL, μm/s) defined as the time per average velocity of a sperm head along its actual curvilinear trajectory; (5) average path velocity (VAP, μm/s) defined as the time per average velocity of a sperm head along its spatial average trajectory; (6) straight-line velocity (VSL, μm/s) defined as the time per average velocity of a sperm head along the straight line between its first-detected position and its last position; (7) linearity of the curvilinear path (LIN, %), expressed as VSL/VCL; (8) straightness (STR,%) defined as VSL/VAP; (9) wobble (WOB,%) expressed as VAP/VCL; (10) amplitude of the lateral head displacement (ALH, μm); and (11) beat cross frequency (BCF, Hz) based on VCL crossing VAP per second. Motility parameters were evaluated at 15 s after activation to avoid drifting and to corroborate that all samples were measured at an exact equal time post activation. At least, 200 spermatozoa were analyzed for each sample. Three fields per sample were evaluated. If samples reported very low concentrations, more than three fields were captured.

#### *2.8. Behavior Analysis*

To test the exploratory behavior of the animals, we used a novel tank test (NTT), which evokes motivational conflict between the "protective" diving behavior and subsequent vertical examination following established procedures. Briefly, each animal was individually placed in the evaluation arena (20 cm (*x*) × 18 cm (*y*) × 8 cm (*z*); swimming volume: 3.5 L). Males were let to acclimate to the new environment for 3 min and right after animal behaviour was filmed (1920 × 1080 px) for 3 min. Individual male swimming activity was monitored using the free digital video tracking software Tracker (physlets.org/tracker/). The actual position of the animal was manually located every 20 frames to avoid possible inaccuracies of the automatic option of the software. Then, each resulting track was evaluated using a virtual grid pattern with upper and lower subareas in order to allow quantification and comparison between experimental groups. For each animal we quantified two estimates of exploratory behavior: Number of crossings between the upper-half subarea and the lower one and the percentage of time spent in the upper half of the arena.

#### *2.9. Data Analysis*

Results are expressed as the mean ± standard error. Statistical differences between mean values of each variable at 0 and 21 days were determined using a t-Student test for correlated variables for normally distributed variables or a Wilcoxon test for paired samples for non-parametric variables. A principal component analysis was performed for the set of observed variables for CASA parameters. All statistical analysis were performed using Prism 8 (GraphPad Software, San Diego, CA, USA) and SPSS V. 22 (SPSS Inc., Chicago, IL, USA). *P*-values < 0.0500 were considered statistically significant.

#### **3. Results**

#### *3.1. E*ff*ects of Probiotic Mixture Supplementation on Total Body Weight and Spermiation Capacity*

In order to investigate the effects of probiotic supplementation in male zebrafish on growth parameters, we weighed the animals included in the experiment at t = 0 days and t = 1 days (Figure 1). As expected, taking into account the short temporal frame of our experiment, our analysis revealed no statistical differences (*p* > 0.0500) at day 21 in any of the experimental groups: Control (C; CTRL); maltodextrin, the vehicle control (M; MALTO); or probiotic-fed (P; PROBIO) (Figure 2A; Supplementary File Table S1). These data provide confidence about the suitability of maltodextrin as a carrier in our probiotic-fed group.

**Figure 2.** Probiotic mixture supplementation effects on zebrafish males and sperm quality after a cycle of spermatogenesis exposure. (**A**) None of the studied diets modified male total body weight. (**B**) Spermiation ability of studied males before and after the experiment. (**C**) Concentration, (**D**) total motility, and (**E**) progressive motility at 0 and 21 days obtained for each experimental group represented with violin graphs. "C", "M", and "P" refer to the experimental groups: Control, maltodextrin, and probiotics respectively. Furthermore, "c", "d", and "e" are before–after graphs for the PROBIO group where "M#" indicates the number of the male. Asterisks show statistically significant differences: \* (*p* < 0.0500), \*\* (*p* < 0.0100).

As a first general parameter regarding spermatogenesis, we focused our attention on spermiation ability. As a result of a correct spermatogenesis, mature spermatozoa are released from cysts into the lumen of the tubules and therefore ejaculated. At day 0, we arranged population homogeneously with three (25%; CTRL), four (33%; MALTO), and three (25%; PROBIO) non-spermiating males in each group. After a single cycle of spermatogenesis (21 days) the non-fluent male percentage changed as follows: four (33%; CTRL), two (20%; MALTO), and one (8.33%; PROBIO). Interestingly, a spermiation modulation was suggested with this data in males supplemented with bacteria strains (PROBIO). In this group, only one male did not provide sperm at day 21 sampling (Figure 2B).

#### *3.2. E*ff*ects of the Probiotic Mixture on Concentration, Total Sperm Motility, and Progressive Motility*

To study the effects of the ingested probiotic strains on a single cycle of spermatogenesis, we studied individually, using VIE tagging for male tracking, the sperm samples in terms of concentration, total motility, and progressive motility. At day 0, all males included in this study presented an initial total motility below 60%. Groups were created including a wide range of sperm motility values from 0% motility to the 60% threshold (Figure 1B). Results regarding sperm concentration, total motility, and progressive motility are presented in Figure 2.

Bacteria ingestion strongly modified (*p* = 0.0050) sperm count (10<sup>8</sup> cells/mL; mean <sup>±</sup> s.e.) in the PROBIO group after 21 days of supplementation (Figure 2C; Table S2). The mean value for this variable increased from 44.58 ± 16.40 to 110.10 ± 23.13. Controls reported lower concentrations at day 21: 30.19 ± 10.15 (CTRL) and 41.57 ± 18.16 (MALTO), respectively.

Regarding total motility (%; mean ± s.e.), controls showed similar values (*p* > 0.0500) before and after the experiment. Mean values were: 26.44 ± 6.528 (day 0) vs. 28.97 ±6.194 (day 21) for the CTRL group and 24.56 ± 5.53 (day 0) vs. 24.97 ±6.77 (day 21) for the MALTO group (Figure 2D; Table S2). In contrast, the animals with a feeding regime supplemented with probiotics (PROBIO) revealed a substantial rise (*p* = 0.0018) in total motility from 28.39 ± 6.46 (day 0) to 48.36 ± 7.32 (day 21). When a detailed individual evaluation of data was performed, results showed strong increments (>40%) in 11 of 12 studied males. Four fish increased their sperm motility more than 100%: M3, M8, M10, and M12 (Figure 2D.d). These results evidence a very strong positive effect of probiotic ingestion on zebrafish sperm quality. Please find individual before–after graphs for CTRL and MALTO groups in Figure S1.

Moreover, and concomitantly with total motility, progressive motility (P-MOT) was also significantly raised (*p* = 0.0137) after 21 days in the PROBIO group from 15.22% ± 4.71% to 22.73% ± 5.09% (Figure 2E; Table S2). P-MOT is another key parameter in sperm quality since it influences fertilization success, and it is a focus of attention in zebrafish research [25,26]. In our experiment, we considered progressive those cell in which swimming track was forwards in 80% of a straight line.

#### *3.3. E*ff*ects of the Probiotic Mixture on Sperm Kinematic Parameters*

We studied sperm kinetics in depth since it has been reported that external-fertilizing fish have the highest known intensity of sperm competition of any external fertilizing vertebrates. Thus, the presence of fast subpopulations within the motile cells seems to be an advantage and, therefore, it can be considered a parameter of sperm quality. In this experiment subpopulations were clustered in terms of VCL. Four groups were established and set up in the CASA software: Static, slow, medium, and fast according to the following thresholds: 10 μm/s < slow < 45 μm/s< medium < 100 μm/s < fast. Interestingly, only the PROBIO group showed statistical differences at the end of the experiment (Figure 3A) in the four subpopulations: Static (*p* = 0.006), slow (*p* = 0.0208), medium (*p* = 0.0270), and fast (*p* = 0.0323) cells.

Concerning kinematic parameters, there was no overall difference (*p* > 0.0500) in sperm velocities (VCL, VSL, VAP), linearity (LIN), straightness (STR), wobble (WOB), amplitude of the lateral head displacement (ALH), or beat cross frequency (BCF). Figure 3B shows these results for the PROBIO experimental group. These results suggest that the effect of probiotic bacteria do not fine tune zebrafish sperm kinetics. PCA results can be found in Figure S2.

**Figure 3.** Probiotic mixture supplementation effects on zebrafish sperm kinematics after a cycle of spermatogenesis exposure. (**A**) Sperm subpopulations within the motile population according to speed parameters before and after the experiment for each group. (**B**) Detailed sperm kinetics rendered by the CASA system for the PROBIO group. Asterisks show statistically significant differences; \* (*p* < 0.0500). Abbreviations: VCL—curvilinear velocity, VSL—straight line velocity, VAP—average path velocity, LIN—linearity of the curvilinear path, STR—straightness, WOB—wobble, ALH—amplitude of the lateral head displacement, BCF—beat cross frequency.

#### *3.4. E*ff*ects of the Probiotic Mixture on Male Behavior*

After three minutes of adaptation time to a new environment, tracking analysis were performed to evaluate the anxiety status of the fish. Quantification of the novel tank test (NTT) was carried out attending to two variables: (1) The percentage of positions scored in each of the two virtual zones (upper and lower) of the novel tank and (2) the number of crossings from one to another (Figure 4A). The novel tank test (NTT) is the conceptual equivalent of the rodent open field (OF) paradigm; NTT induces motivational conflict between the "defensive" diving behavior and following vertical exploration [27,28]. Each male was analyzed at the beginning and at the end of the experiment a day before sperm squeezing (*t* = −1 day and *t* = 20 days). Summaries of each animal behavior were created for easier evaluation (Figure 4B). As can be checked in Figure 4C,E, there was a non-statistical (*p* > 0.0500) trend toward PROBIO fish spending more time in the top of the tank, close to the 50% in mean values, at day 21 (47.92% ± 8.37%) compared to day 0 (39.25% ± 4.71%). The number-of-crossings evaluation revealed again significant differences (*p* = 0.0373) only in the PROBIO population doubling the mean values of the variable (Figure 4D; Table S3). Moreover, a moderate correlation (*p* = 0.0018; *r* = 0.5595) between fish behavior and total motility of squeezed ejaculates was observed (Figure 4F).

**Figure 4.** Probiotic bacteria ingestion modulates male behavior in zebrafish. (**A**) Schematic representation of the novel tank test (NTT). (**B**) Individual summaries of the track for each animal in the experiment at 0 and 21 days. (**C**) Organized summaries from higher to lower scores in the lower subarea. (**D**) Comparison of the scores registered in the upper zone before and after treatment. (**E**) Comparison of the number of crossings between the two subareas of the arena at 0 and 21 days. (**F**) Correlation between "number of crossings" and "total motility" variables integrating data from the three experimental groups after probiotic treatment. Asterisk shows statistically significant differences \* (*p* < 0.0500).

#### **4. Discussion**

Nutrition could have a positive or negative impact on reproduction. Nowadays, decrease in sperm quality could be considered a global health problem. Indeed, asthenozoospermia is one of the male subfertility pathologies described by the WHO (2010) as a condition in which the percentage of progressively motile sperm is abnormally low [29]. Since the development and optimization during the last decades of the artificial reproductive technologies (ARTs): Intracytoplasmic sperm injection (ICSI), ovarian stimulation, intrauterine insemination (IUI), or in vitro fertilization (IVF) many infertile couples have found a solution to conceive. Although they have become a major worldwide focus of attention, these techniques are expensive and invasive. Possible clinical approaches may include antioxidant ingestion as a preliminary or concomitant treatment to reproductive techniques to improve fertility outcomes.

The definition of oxidative stress (OS) is the overabundance of reactive oxygen species (ROS) or a deficiency of antioxidants [30]. The imbalance produced by ROS causes cell damage. The deleterious effects of this damage on spermatozoa have been known since the 80 s [31]. There exists evidence regarding the need of certain amounts of ROS for normal sperm functions of both in mammals [32] or teleost [33] mainly produced by the mitochondria. However, excessive quantities become pathophysiological and lead to DNA damage and even apoptosis [7]. Endogenous or exogenous factors may be the cause of high levels of ROS. The most common exogenous causes of OS are obesity, smoking, environmental contaminants, alcohol intake, and malnutrition [9]. Natural antioxidants can scavenge ROS, inactivate them, and repair the cellular damage [34]. Spermatozoa, due to their high specialization, do not present cytoplasm after spermatogenesis and they depend on seminal plasma, which is rich in antioxidants [35]. Moreover, in spermatozoa, polyunsaturated fatty acids (a highly oxidizable substrate) enrich the cell membrane, provoking a high vulnerability to lipid peroxidation from ROS both in mammals [5] and teleost [36]. Oxidative damage affects the sperm flexibility and therefore motility, which is the excellence parameter to assess sperm quality. The spermatozoal heightened vulnerability to OS has caused enormous interest in the role of diet antioxidants in the management of infertile men [7].

Nowadays, the probiotic market is increasing globally as a cheap and well accepted (by the consumers) supplement source all around the world. There exists an increase in the demand of these kinds of products to improve health or prevent human illness. The developing observation that the gut microbiota plays a central role in regulating the host's physiology has supported the significance of the probiotic concept. The modulation of the intestinal microbiota composition has been proposed as one of the main mechanism of probiotic activity [37]. In a previous study, our group reported that the effects of a commercial probiotic diet supplement (Bactocell®, Lallemand Animal Nutrition S.A., Blagnac, France), containing a lactic acid bacteria strain (*Pediococcus acidilactici*) improved molecular sperm quality markers in zebrafish testicular cells after a short period (10 days) [38], providing initial data regarding the potential use of probiotic supplementation on zebrafish male reproductive performance. In the present study, the ingestion of a supplement containing probiotic strains on a single cycle spermatogenesis evidenced a positive effect of the host's sperm quality after a single cycle of spermatogenesis in zebrafish model. Specifically, the present study was undertaken to evaluate the combined effects of two strains: *L. rhamnosus* CECT8361 and *B. longum* CECT7347. These strains were selected because they belong to the most-used genera as probiotics nowadays [16] and they have been previously described as strains with antioxidant activity [23]. Additionally, *B. longum* CECT7347 has been assigned with anti-inflammatory activity. This strain has been described to reduce the inflammatory effects of the dysbiotic intestinal microbiota of individuals with coeliac disease on peripheral blood mononuclear cells partially via the induction of IL-10 production [24,39]. *B. longum* CECT7347 has also been demonstrated to decrease the cytotoxic and inflammatory effects of gliadin peptides on epithelial cell in vitro degradation [40,41]. Furthermore, in the gliadin-induced enteropathy animal model, this strain has been shown to reduce the peripheral CD4<sup>+</sup> T cells, rise IL-10, and shrink TNF-α production [42]. The other strain used in this experiment belongs to the *Lactobacillus* genera, which has been repeatedly shown as the predominant bacteria in the semen, accompanied by a flexible composition of other taxa [43–45]. *L. rhamnosus* species has been described as a highly adhesive bacteria in zebrafish [21]. The specific mechanism by which these bacterial strains are modifying fish behaviour and sperm quality in the present study is unclear. It is known that the ingestion of antioxidants can improve sperm motility [7], but the anti-inflammatory properties of *B. longum* CECT7347 could not be ignored and cannot be separately evaluated in our study. In fact, in humans, it has been reported that the ingestion of a combination of the two strains *L. rhamnosus* and *B. longum* modulated the gut microbiota composition, leading to a significant reduction of potentially harmful bacteria and an increase of beneficial ones [46]. Indeed, the combination of specific bacterial strains belonging to these two species can act in optimal synergy for restoring the intestinal balance [47] even better than individually [48].

In the present study, it was demonstrated that in the zebrafish model, males fed with the probiotic formulation increased sperm quality. In particular, in terms of sperm counts, 11 of the 12 males within the PROBIO group showed an improvement in concentration (Figure 2C), independent of initial values after 21 days of ingestion. The individual track of animals was available thanks to the use of VIE tagging. This technique is starting to spread among facilities since this inert, non-immunogenic polymer is useful for many purposes. The results achieved by this in vivo study clearly showed that the ingestion of the combination of *L. rhamnosus* CECT8361 and *B. longum* CECT7347 increased the percentage of motile cells after a single cycle of spermatogenesis (Figure 2D). After 21 days of exposure, a clear induction of total motility was found in all males within the PROBIO group, contrary to control cohorts (Figure 2D). These results are in line with our previous observations in asthenozoospermic human samples [23] on the ability of the same couple of probiotics reporting an increase of total motility after treatment. In the current study, total motility improved with a 1.7-fold change. Concomitantly to the increment in total motility values, progressive motile cells were also improved (Figure 2D), although the fold-change before and after the probiotics ingestion was lower. In our results, the increment of total motility was also accompanied by a modulation of sperm subpopulations within the motile population (Figure 3). All slow, medium, and fast motile cell populations were increased after 21 days in the probiotic-fed animals contrary to diet-controlled and vehicle-fed ones (Figure 3). Interestingly, the spermatozoa kinetics did not show differences after the experimental time indicating that probiotic bacteria are not able to alter these parameters in the zebrafish model. Altogether, the capability of *L. rhamnosus* CECT8361 and *B. longum* CECT7347 to modulate sperm quality was remarkably corroborated. Our results regarding sperm quality improvement are further supported by some studies [49,50] involving the use of probiotics in other animal models, which reported a potentially positive effect of probiotics in terms of sperm quality parameters. In particular, Dardmeh and colleagues [50] demonstrated that *L. rhamnosus* PB01 (DSM 14870) may have an effect on weight after eight weeks of treatment as well as a modulation of sperm kinetics and hormone levels in mice with diet-induced obesity. It has also been suggested that the use of *Bacillus amyloliquefaciens* TOA5001 as a probiotic has potential positive effect on broiler breeders, since the strain was able to increase sperm count and sperm viability after six weeks of treatment [49].

The scientific community is starting to elucidate the mechanisms provoking these beneficial effects of probiotic ingestion on sperm quality. Recently, Kelton Tremellen has published a novel theory [51], the GELDING theory (Gut Endotoxin Leading to a Decline IN Gonadal function) in which it is postulated that "the trans-mucosal passage of bacterial lipopolysaccharide from the gut lumen into the circulation is a key inflammatory trigger underlying male hypogonadism". The author has also linked the theory to a described positive effect of probiotics on human sperm samples from infertile patients [52]. This new and interesting theory is remarkable after analysing our results. The synergy between antioxidant and anti-inflammatory properties of the two bacteria used in the present study may explain the registered improvement in zebrafish sperm quality. According to Tremellen's theory, this assumption may be accepted.

Noticeably, in the present study, we offer evidence that in adult male zebrafish, short ingestion of probiotics modulates behavior pattern (Figure 4). Zebrafish is an interesting model organism to investigate behavior. Founded on geotaxis—an innate escape "diving" behavior of fish in novel environments—the novel tank test (NTT) has long been used to analyze adult zebrafish behaviors [53] and drug responses [54]. Adult zebrafish initially spend more time at the lower part of the tank when they are exposed to a novel environment. Concomitantly, they reduce "top" swimming and reveal more unpredictable movements and show freezing/immobility events [55]. Subsequent, because of habituation to the NTT novelty, animals gradually explore the top area (theoretically less safe for zebrafish in their wild habitats due to predator risk) [28]. Although our results did not report statistically significant differences in the number of scores in the upper subarea of the novel arena before and after probiotic administration, the number of crossing between the bottom and the upper area revealed differences (Figure 4). These results suggest that *L. rhamnosus* CECT8361 and *B. longum* CECT7347 modulate the exploration activity of the males after only 21 days of exposure showing a lower stress-like conduct. The microbiota signals to the central nervous system (CNS) via several potential pathways [56]. Probable mechanisms of communication embrace production of various metabolites that cross the intestinal barrier into the circulatory system, and/or microbe-derived metabolites that can signal through the immune system [57]. Moreover afferent pathways of the vagus nerve from the enteric nervous system (ENS) to the CNS have been associated as a key route of communication concerning the microbiota and CNS [58]. Our results are in accordance with a number of recent findings reporting that the use of various *Lactobacillus* and/or *Bifidobacterium* strains can lighten anxiety-and depressive-like behavior and alleviate stress responses in animal models [58–61].

#### **5. Conclusions**

In conclusion, our study showed that 21 days of treatment (a spermatogenesis cycle) with a probiotic mixture with described antioxidant and anti-inflammatory activities significantly improved zebrafish sperm quality and increased the number of fluent males. These data highlight the promising use of this probiotic mixture to improve reproductive performance in different quality sperm samples by increasing sperm total motility, progressive motility, concentration, and fast sperm populations. Furthermore, behavior analysis revealed a modulation in probiotics-fed males suggesting a lower anxiety-like pattern, which may be correlated with the improvement of sperm quality parameters in this model. Considering the simplicity and economical effectiveness of the studied multistrain product, the results presented here strengthen the potential use of this preparation in male reproductive biotechnology, which may be useful in the aquaculture industry and reproductive biology fields.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6643/11/4/843/s1, Figure S1: Before–after graphs for (**A**) concentration, (**B**) total motility, and (**C**) progressive motility at 0 and 21 days obtained for each experimental group. "M" and "P" refers to the experimental groups: Control and maltodextrin respectively. "M#" indicates the number of the male; Figure S2: (**A**) PCA analysis for Computer Assisted Sperm Analysis (CASA) variables for probiotic-fed group. (**B**) Representation of the experimental group in a principal component plane; Table S1: Weight values (mean ± s.e.) for the groups before and after treatment; Table S2: Sperm quality values (mean ± s.e.) for the groups before and after treatment; Table S3: novel tank values for behavior estimators (mean ± s.e.) for the groups before and after treatment.

**Author Contributions:** Conceptualization, V.R., D.G.V., and M.F.R.; data curation, D.G.V., M.F.R., and J.M.M.-V.; formal analysis, D.G.V. and M.F.R.; funding acquisition, V.R.; investigation, V.R., D.G.V., and M.F.R.; methodology, V.R., D.G.V., M.F.R., and J.M.M.-V.; project administration, V.R.; resources, V.R.; software, D.G.V. and J.M.M.-V.; supervision, V.R.; validation, D.G.V.; visualization, D.G.V.; writing—original draft, V.R., D.G.V., and M.F.R.; writing—review and editing, V.R. and D.G.V.

**Funding:** This work was supported by project AGL2015 68330-C2-1-R (MINECO/FEDER).

**Acknowledgments:** The authors acknowledge contract PTA2016-11987-I (MINECO/FEDER), Biopolis S.L. (Valencia, Spain), AQUA-CIBUS International Net 318RT0549 (CYTED), and the staff from Planta de Cultivos El Bocal (IEO).

**Conflicts of Interest:** The authors declare that they have no competing interests.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Plausible Biological Interactions of Low- and Non-Calorie Sweeteners with the Intestinal Microbiota: An Update of Recent Studies**

**Julio Plaza-Diaz 1,2,3, Belén Pastor-Villaescusa 1,4,5, Ascensión Rueda-Robles 1,2, Francisco Abadia-Molina 2,6 and Francisco Javier Ruiz-Ojeda 1,3,7,\***


Received: 8 April 2020; Accepted: 15 April 2020; Published: 21 April 2020

**Abstract:** Sweeteners that are a hundred thousand times sweeter than sucrose are being consumed as sugar substitutes. The effects of sweeteners on gut microbiota composition have not been completely elucidated yet, and numerous gaps related to the effects of nonnutritive sweeteners (NNS) on health still remain. The NNS aspartame and acesulfame-K do not interact with the colonic microbiota, and, as a result, potentially expected shifts in the gut microbiota are relatively limited, although acesulfame-K intake increases Firmicutes and depletes *Akkermansia muciniphila* populations. On the other hand, saccharin and sucralose provoke changes in the gut microbiota populations, while no health effects, either positive or negative, have been described; hence, further studies are needed to clarify these observations. Steviol glycosides might directly interact with the intestinal microbiota and need bacteria for their metabolization, thus they could potentially alter the bacterial population. Finally, the effects of polyols, which are sugar alcohols that can reach the colonic microbiota, are not completely understood; polyols have some prebiotics properties, with laxative effects, especially in patients with inflammatory bowel syndrome. In this review, we aimed to update the current evidence about sweeteners' effects on and their plausible biological interactions with the gut microbiota.

**Keywords:** nonnutritive sweeteners; sweetening agents; gut microbiota

#### **1. Introduction**

Excessive sugar consumption has become an important public health concern due to its adverse effects on health and metabolic consequences such as obesity, insulin resistance, metabolic syndrome, cardiovascular diseases, and type 2 diabetes. One century ago, sweetening agents or sweeteners—sugar substitutes that mimic the sweet taste—emerged as an alternative to sucrose and glucose–fructose syrups consumption to reduce energy intake [1,2]. However, the impact of sugar consumption on health continues to be a controversial topic in relation to its effects on metabolic disease [3]. Some contradictory results were published in 2015 about sweeteners and gut microbiota. Suez et al. concluded that some sweeteners may affect the human microbiome, and consequently more studies are needed [4]. In contrast, Frankelfeld et al. [5] reported no differences in median bacterial abundance across consumers and non-consumers of sweeteners.

Sweeteners are between several hundred to thousands of times sweeter than sucrose and they do not contain too many calories. They include nonnutritive sweeteners (NNS), which have a higher sweetening intensity than other sweeteners, such as acesulfame K (ace-K), advantame, aspartame, aspartame–acesulfame salt, cyclamate, neohesperidin dihydrochalcone, neotame, saccharin, steviol glycosides (including 10 different glycosides), sucralose, and thaumatin, low-calorie sweeteners (LCS), such as polyols or sugar alcohols and other new sugars which are low-digestible carbohydrates derived from the hydrogenation of their sugar or syrup sources. Polyols are around 25%–100% as sweet as sugar and include erythritol, hydrogenated starch hydrolysates (sometimes listed as maltitol syrup, hydrogenated glucose syrup, polyglycitol, polyglucitol, or simply HSH), isomalt, lactitol, maltitol, mannitol, sorbitol, and xylitol. All of them are considered safe for human consumption as long as they are consumed within the acceptable daily intake [6]. This safety was claimed by the European Food Safety Authority (EFSA) except for cyclamate, which is not approved by the US Food and Drug Administration [1,7,8].

In 2019, we reviewed the effects of sweeteners on the gut microbiota, considering both experimental studies and clinical trials, and we reported that, among NNS, only saccharin and sucralose shift the populations of the gut microbiota, although more human studies are needed to clarify those observations. Within nutritive sweeteners (NS), only stevia extracts seem to affect gut microbiota composition, while some polyols, such as isomalt and maltitol which can reach the colon, increase *Bifidobacterium* in healthy subjects and might act as prebiotics. Besides, lactitol can decrease *Bacteroides*, *Clostridium*, coliforms, and *Eubacterium*, increasing butyrate and IgA secretion in humans [1]. Thus, we concluded that still more studies are needed; however, as the plausible biological interaction between sweeteners and intestinal microbiota has not been reported elsewhere, we aimed to review and update the current knowledge about sweeteners and gut microbiota interactions in humans.

A comprehensive literature search was conducted in PubMed, Embase®, and Scopus using different combinations of the following keywords: aspartame, acesulfame-K, cyclamate, sucralose, saccharin, steviol glycosides, erythritol, isomalt, lactitol, maltitol, sorbitol, mannitol, xylitol, and microbiota, with special attention and interest to what was published from February 2018 to March 2020.

#### **2. Biological Plausibility: Which Low- and Non-Calorie Sweetener (LNCS) Could Potentially A**ff**ect the Colonic Microbiota?**

Biological plausibility is one component of a method of reasoning that can establish a cause-and-effect relationship between a biological factor and a particular disease or adverse effect based on assessing the strength of evidence, since the work of Bradford Hill [9]. Here, we will assess biological plausibility between different sweeteners and gut microbiota composition. Although we usually refer to the different low- and non-calorie sweeteners (LNCS) as if they were a single molecule, it is well known that they do not share their absorption, distribution, metabolism, and excretion (ADME) profiles. Therefore, the extrapolation of the effect of a particular LNCS on the intestinal microbiota to all LNCS is unappropriated. These differences are crucial to understanding if each LNCS has the potential to alter the intestinal microbiota, directly or indirectly. For more detailed information on the metabolic fate of each LNCS beyond its relationship with the colonic microbiota, the excellent review by Magnuson et al. (2016) [10] can be consulted.

#### *2.1. E*ff*ects of Non-Nutritive Low-Calorie Sweeteners on the Gut Microbiota*

#### 2.1.1. Aspartame

Aspartame is a methyl ester of a dipeptide composed of L-phenylalanine and aspartic acid. When ingested, this dipeptide undergoes enzymatic hydrolysis in the gastrointestinal lumen and in the cells of the internal intestinal mucosa (by peptidases and intestinal esterases), so that virtually no aspartame enters the general circulation [11,12]. Hence, aspartame as an intact molecule cannot interact directly with the colonic microbiota. The three digestion products (aspartic acid, L-phenylalanine, and methanol) are rapidly absorbed in the duodenum and jejunum [12], reaching the systemic circulation without passing through the colon [10]. These degradation products are presented in the same way as when they are absorbed from vegetables, fruits, dairy, or meat, and at much lower concentrations than when they are derived from such foods [11]. These products follow their usual metabolic pathways.

Methanol enters the portal circulation into the liver and, by the enzymatic action of alcohol dehydrogenase, is metabolized to formaldehyde, which in turn, by the action of formaldehyde dehydrogenase, is oxidized to formic acid. Formic acid can be eliminated by the respiratory tract as carbon dioxide or excreted into the urine [10,11]. Aspartate undergoes a transamination reaction in the enterocytes, becoming oxalacetate. Oxalacetate and aspartate are interconverted in the body and can participate in the urea cycle and gluconeogenesis in the liver. Excess aspartate is eliminated in the urine [10]. Phenylalanine is absorbed in the gastrointestinal tract mucosa. It enters the liver through portal circulation, where, by the action of phenylalanine hydroxylase, can be converted into tyrosine. Phenylalanine that reaches the systemic circulation can be distributed throughout the body [11]. Its excess is excreted in the urine [13].

Based on the abovementioned information, the finding of a positive association between intake of aspartame and alteration of the colonic microbiota of rodents could be in fact due to the effect of what the animals stop eating rather than to the effect of aspartame intake itself. This last concept applies to all LNCS. However, a recent study carried out in female Sprague Dawley rats subjected to a high-fat/sucrose (HFSD), a HFSD + aspartame (5–7 mg kg−<sup>1</sup> day<sup>−</sup>1), or a HFSD + stevia (2–3 mg kg−<sup>1</sup> day<sup>−</sup>1) diet showed an increase of body fat in the offspring at weaning following maternal consumption of aspartame and stevia in the HSFD. In addition, glucose tolerance was altered, particularly with aspartame. *Akkermansia muciniphila* and *Enterobacteriaceae* concentrations were higher in mothers compared with their offspring. Regarding the cecal microbiota, a reduced abundance of *Enterococcaceae*, *Enterococcus*, and *Parasutterella* and an increased abundance of *Clostridium* cluster IV were found in the aspartame group. Moreover, fecal transplantation from offspring to germ-free mice produced an altered gut microbiota, causing impaired adiposity and glucose tolerance. In addition, increased concentrations of *Porphyromonadaceae* in males and females obese–aspartame and obese–stevia offspring were found [14]. In contrast, in the study by Suez et al., food intake in mice assigned to a water group with LNCS (aspartame, sucralose, and saccharin) was reduced by up to 50%. It is known that dietary factors are key determinants of the composition of the intestinal microbiota; indeed, differences in both total caloric intake and the type of food consumed can lead to a different microbial composition [15–17].

Thus, the intestinal microbiota might have been altered by a reduced consumption of fiber, protein, fat, and carbohydrates; therefore, it seems uncertain that the reported change in the intestinal microbiota was caused by the LNCS, and the changes that diet per se may provoke in the intestine should be considered. Nonetheless, there are studies that reveal possible modifications of the microbiota due to the use of aspartame. The study by Mahmud et al. analyzes the combined and individual effects of the administration of low concentrations of aspartame and Ace-K. Induction of *Escherichia coli* growth and expression of some important genes which may be related to its colonization in the gut were observed [18]. In another study with human fecal samples, aspartame administration significantly increased *Bifidobacterium* and *Blautia coccoides* growth and decreased the *Bacteroides*/*Prevotella* ratio; nevertheless, the aspartame-based

sweetener used in this study was rich in maltodextrin, thus, the authors did not study the effect of aspartame alone [19].

#### 2.1.2. Potassium Acesulfame (Ace-K)

After its intake, Ace-K is absorbed almost completely in the small intestine as an intact molecule and distributed by the blood to different tissues. Without undergoing any metabolization, more than 99% of Ace-K is excreted in the urinary tract within the first 24 h, with less than 1% being eliminated in the feces [10,20]. The minimum amount of Ace-K ingested, its rapid absorption, and its urinary excretion causes the Ace-K concentration that reaches the fecal or colonic bacteria to be negligible [10,21]. Therefore, it is extremely unlikely that this LNCS could have a direct effect on the colonic microbiota [16]. However, some studies have reported small shifts in the gut microbiota composition following Ace-K intake.

A cross-sectional study was conducted in humans and showed no modifications in the intestinal microbiota nor significant differences by sex, contrary to the study conducted in rats by Bian X et al. [22]. Other studies also indicate that Ace-K causes changes in the microbiota and their metabolites, such as butyrate and pyruvate [22,23]. The study carried out by Uebanso et al. suggests that the daily intake of maximum adequate diary intake (ADI) levels of Ace-K does not affect the relative amount of the *Clostridium* cluster XIVa in the fecal microbiome [24]. In contrast, a study in mice that received 150 mg kg−<sup>1</sup> of Ace-K by free drinking during 8 weeks, showed that lymphocyte recruitment was increased, with augmented expression of inflammatory cytokines and adhesion molecules [25]. Recent studies in rats indicate that administration of a mixture of sucralose and Ace-K at concentrations near the upper limit of ADI for human consumption during mice pregnancy has consequences on the progeny, causing metabolic and microbiome alterations. The authors observed an increase in *Firmicutes* and a depletion of *A. muciniphila*, which is a beneficial bacterium inversely correlated with fat mass gain, type 1 diabetes, and inflammatory bowel syndrome (IBS) [26]. The researchers also indicated an increase in the variety of species in the microbiota; however, *A. muciniphila* was significantly depleted, suggesting that the divergence between mothers' and pups' microbiomes was due to increasing NNS concentrations [27].

Regarding the bacteriostatic effect of Ace-K, this sweetener shows a strong inhibitory effect on the growth of *E. coli* HB101 and *E. coli* K-12 [28]. In contrast, using a concentration of Ace-K of 2.5 mg/mL, the result was an induction in *E. coli* growth, whereas the growth stimulation decreased gradually when higher concentrations of sweetener were used [18].

#### 2.1.3. Cyclamate

Cyclamate is the sodium or calcium salt of cyclamic acid (cyclohexanesulfamic acid), which itself is prepared by the sulfonation of cyclohexylamine and is eliminated in the feces [29]. In a study carried out by Vamanu et al. [30], the authors determined the effect of sweeteners on the microbiota pattern using an in vitro model. In this study, the total quantity of synthesized short-chain fatty acids (SCFA) and the number of microorganisms were decreased, and a negative influence on the fermentative profile was observed, although with an increase of *Bifidobacterium*. The ratio of butyric/propionic acids was also affected, indicating that those SCFA could affect the gut microbiota composition. Cyclamate also exerts a positive effect, producing an inhibitory anaerobic fermentation of glucose in a rat model of intestinal gut microbiota [28].

Cyclamate and sucralose can alter the ratio between butyric and propionic acids [30]. SCFA have multiple effects on human health. Butyric acid has anti-obesogenic effects, reduces insulin resistance, and improves dyslipidemia [31]. Lower concentrations of propionic and butyric acids have been positively correlated with the four subtypes of IBS and can be harmful to people with that disease [32]. Overall, it seems that cyclamate has some effects on gut microbiota composition, but more studies on its possible effect on human health are needed.

#### 2.1.4. Sucralose

Sucralose has a very low level of absorption (less than 15%) and it is practically not metabolized. Therefore, after intake, more than 85% of sucralose reaches the colon unchanged [10]. The small proportion of sucralose that is absorbed is eliminated in the urine mainly unchanged, though two glucuronides of sucralose were also detected in a small proportion (approximately 2%) [33].

Although more than 85% of the ingested sucralose contacts the colonic microbiota, between 94% and 99% of this LNCS is recovered in the feces without any structural change, thus indicating little or no metabolism by the gut microbiota [10]. Thus, sucralose does not appear to be a substrate for the colonic microbiota [16]. Nevertheless, considering the practically null microbial metabolism of sucralose, we must be cautious when interpreting the results of studies that indicate an alteration of the intestinal microbiota after sucralose consumption [17]. In those cases, it will be worth investigating whether pure sucralose or a commercial formulation was used in the research, since these formulations usually contain around 1% of sucralose and 99% of the carriers maltodextrins [16].

On the other hand, it has been shown that sucralose promotes inflammation in a mouse model of human Crohn's disease-like ileitis as well as dysbiosis of the gut microbiota [34]. Furthermore, sucralose causes a decrease in the number of Firmicutes species [35]. This result is the opposite to that reporter by Olivier-Van Stichelen et al., who found that Firmicutes doubled, including the Clostridiales families Lachnospiraceae and Ruminococcaceae (e.g., Oscillospira), in mice's pups [27]. Wang et al. observed an increase of Firmicutes and a tendency to decrease for Bacteroidetes [28]. These authors did not observe changes in Actinobacteria and Proteobacteria phyla in mice fed with a chow diet, but they reported a synergistic effect when sucralose was provided in the context of a high-fat diet. On the other hand, a chow diet might cause a significant increase in *Bifidobacterium* [28]. A study carried out in humans examined the short-term effect of sucralose consumption on glucose homeostasis and gut microbiome in healthy male volunteers. The authors concluded that no changes occurred in the gut microbiome due to sucralose intake [36]. In contrast, another study shows an increase in the abundance of pro-inflammatory bacteria like *Turicibacter*, which was associated with hepatic inflammation, after sucralose administration [37].

Splenda administration in mice was associated with a high presence of Bacteroidetes, an enhanced overgrowth of *E. coli*, and the expansion of Proteobacteria [38]. The effect of sucralose was analyzed in fecal samples from 13 healthy volunteers. The authors found increased abundances of *Escherichia*, *Shigella,* and *Bilophila.* With regard to SCFA, increased production of valeric acid was observed [19].

A recent publication evaluated the short-term effect of sucralose consumption on glycemic control and its interaction with the intestinal microbiota (comparison before/after the intervention by 16S rRNA sequencing) in healthy subjects. This study concluded that consumption of high doses of sucralose (75% of the ADI) for 7 days did not alter glycemic control, insulin resistance, or intestinal microbiome at the phylum level [36].

Although previous human studies showed similar results concerning glycemic control (glycosylated hemoglobin, fasting blood glucose, C-peptide), both in diabetic [39] and in non-diabetic populations [40], this is the first time that a randomized, controlled, double-blind study concomitantly evaluated the composition of the intestinal microbiome in healthy subjects, thus providing a better level of evidence in comparison to other earlier published trials.

#### 2.1.5. Saccharin

After intake, more than 85% of saccharin is absorbed as an intact molecule, since it does not undergo gastrointestinal metabolism. Once absorbed, it binds to plasma proteins and is distributed throughout the body. Finally, it is eliminated by urine through active tubular transport [10,41,42]. The small percentage of non-absorbed saccharin is excreted into the feces, indicating that high concentrations of this LNCS could lead to changes in the composition of the intestinal microbial population [16]. It is important to highlight that one of the main studies that reported an alteration of the intestinal

microbiota with the consumption of saccharin [17] was carried out by administering the full ADI of saccharin, which does not correspond to what happens with habitual human consumption.

In an in vitro model study, saccharin produced an increase in *Bifidobacterium*. Not only saccharin but also sucralose caused a decrease in the number of Firmicutes species, directly correlated with the SCFA level [30].

Some herbicides, which are considered nowadays safe, can change the gut microbiota of animals in the early stages of embryonic development. Indeed, exposure to glyphosate and glyphosate in combination with saccharin contributes to the broader reproduction of pathogenic bacteria such as *Klebsiella*, *Citrobacter*, *Enterobacter,* and *Pseudomonas* [43]. On the other hand, studies show that saccharin administration can also disrupt monolayer integrity and alter paracellular permeability in a Caco-2 cell monolayer model [44].

Overall, saccharin administration also promotes Bacteroidetes, Turicibacter, and Clostridiales and reduces Firmicutes abundances. The Turicibacter bacteria increases have been related to a pro-inflammatory effect of saccharin [37].

The effect of a mixture of fiber–prebiotics and saccharin–eugenol has been evaluated in dogs. Four diets were prepared: control diet, containing 5% of cellulose; diet containing a 5% fiber and prebiotic blend; diet containing 0.02% of saccharin (sweetener SUCRAM) and eugenol and 5% of a fiber and prebiotic blend plus 0.02% of saccharin and eugenol. The use of saccharine did not affect species richness measured by alpha-diversity or alter the proportions of bacterial phyla. No changes were observed in fecal microbial communities [45]. More studies are needed to confirm these saccharin effects using different concentrations and animal models.

#### 2.1.6. Steviol Glycosides

Steviol glycosides can be extracted from the leaves of *Stevia rebaudiana*. They all have a central steviol structure, conjugated with different sugar residues, such as stevioside and rebaudioside A, which all are steviol glycosides. Steviol glycosides are hydrolyzed neither by enzymes nor by the acid present in the upper gastrointestinal tract [46]. Therefore, they pass through the upper portion of the gastrointestinal tract without being absorbed and enter the colon as intact molecules [47]. In the colon, bacteria of the Bacteroidacea family eliminate the sugar residues that are conjugated to steviol [47,48]. While these sugar residues may represent a source of energy for the microbiota [49], it is worth noting that the energy contribution is negligible, given the low total daily intake of steviol glycosides [50]. The resulting steviol is not a substrate for the intestinal microbiota, since it is resistant to bacterial degradation [48]. Hence, steviol is completely absorbed and reaches the liver where it is conjugated with glucuronic acid. Steviol glucuronide is mainly excreted in the urine in humans [51,52].

While steviol glycosides interact with the colonic microbiota, there are no reports indicating that these compounds could affect bacteria negatively [30]. A recent study showed that steviol incubation in the GIS1–phase 2 system, an in vitro system that simulates the human intestinal microbial ecosystem, reduced the ammonium level and *Bifidobacterium* and exerted a negative influence on the fermentative profile, resulting in higher pH and SCFA ratio [30].

*S. rebaudiana* is another natural steviol glycoside 250 times sweeter than sucrose [53]. In Europe, only the purified steviol glycosides are approved for use in food, and the ADI of 4 mg kg−<sup>1</sup> of body weight per day is safe (EU Regulation (EU) 1129/2011) [1,7,54]. Another study recently reported that a low dose of stevia rebaudioside A alters gut microbiota composition and reduces nucleus accumbens tyrosine hydroxylase and dopamine transporter mRNA levels in rebaudioside A-supplemented rats. Nonetheless, the oligofructose-enriched inulin prebiotic, in the presence or absence of rebaudioside A, reduced fat mass, food intake, gut permeability, and cecal SCFA concentration. However, only stevia rebaudioside A increased SCFAs acetate and valerate, which are positively correlated with fat mass and total weight. Hence, stevia rebaudioside A seems to decrease the "healthy" status of the gut microbiota [55].

Chronic stevia consumption has effects on gut microbiota and immunity in the small intestine of young mice. In 21-day-old mice treated with sucrose, Splenda, and stevia, mice preferred the consumption of Splenda and stevia. Besides, those mice showed an increase in CD3<sup>+</sup> lymphocytes in Peyer's patches, but only stevia induced an increase in the lamina propria. Both Splenda and stevia elevated leptin, C-peptide, IL-6, and IL-17 and decreased resistin. Stevia modified the predominantly genera *Bacillus* such as *Bacillus aerius*, *Bacillus circulans*, *Bacillus licheniformis*, and *Bacillus safensis*, although the authors observed effects on *Streptococcus saliviloxodontae*, *Oceanobacillus sojae*, and *Staphylococcus lugdunensis*. Even though the results of this study are significant, they have some limitations. The modifications observed in the immune system of the mucous membranes and in the microbiota of the small intestine in young mice after weaning depend on age and diet. This study used culture media and not metagenomic approaches, and some results might be related to some carriers present in the evaluated products, such as maltodextrins [56].

Recently, by testing stevia glycosides and erythritol, which are often combined in food preparation to minimize changes in the organoleptic profile, in an in vivo *Cebus apella* model, changes in bacteria growth and gut microbial structure and diversity have been observed [57]. Overall, stevia seems to modify the gut microbiota; however, further studies are needed to clarify its specific effects.

Although different changes in the intestinal microbiota have been described in relation to the influence of sweeteners on the immune system, the wide use of aspartame, ace-K, cyclamate, sucralose, saccharin and steviol glycosides makes it necessary to carry out other analyses to complete the picture of the influence that these sweeteners have on the intestinal microbiota.

#### *2.2. E*ff*ects of Nutritive Low-Calorie Sweeteners on the Gut Microbiota*

#### 2.2.1. Polyols

Polyols are a group of compounds used in an increasingly wide variety of commercial foods as additives. They are quite stable at high temperatures and various pH and do not interfere in Maillard reactions, conferring organoleptic characteristics to the foods. Polyols are naturally present in fruits, vegetables, and mushrooms and are used to produce food without added sugar, reducing the sugar content in recipes. In addition, polyols are non-cariogenic, do not induce salivation, and do not interfere with insulin and glucose levels in the blood. Nevertheless, the excessive consumption of polyols causes gastrointestinal symptoms and laxative effects, which can be even worse in patients with IBS. As we described previously, the FDA, the Codex Alimentarius, and the EFSA have approved eight different polyols, i.e., erythritol, hydrogenated starch hydrolysates, isomalt, lactitol, maltitol, mannitol, sorbitol, and xylitol, for use as bulk sweeteners in human foods [1,58]. Indeed, in September 2019, EFSA launched an open consultation on the "Protocol for the assessment of hazard identification and characterization of the sweeteners", which will be used for the evaluation of the safety of sweeteners under the re-evaluation program of food additives. The evaluation should be completed by the end of 2020 [59].

#### 2.2.2. Erythritol

Erythritol (E-968) is a four-carbon sugar alcohol that has a fast absorption through the small intestine with a very low metabolization and it is over 90% excreted unchanged in the urine [58]. Furthermore, an unabsorbed part (~10%) is fermented in the large intestine by the colonic microbiota, which consequently rarely leads to gas production [60]. Hence, the limited amount of erythritol that reaches the colon could be the explanation of the lack of evidence of effects of erythritol on the gut microbiota in humans, based on clinical trials as we previously reported [1]. Nevertheless, a recent in vitro study demonstrated that low doses of erythritol (25 μg mL<sup>−</sup>1, 50 μg mL<sup>−</sup>1, and 100 μg mL<sup>−</sup>1) did not exert any effect on the growth of *Escherichia*, *Enterococcus*, *Lactobacillus*, *Ruminococcus* and Bacteroides in the human gut microbiota. Moreover, erythritol doses did not disrupt alpha and beta diversities or the composition of the human gut microbial community [57]. In contrast, butyric and pentanoic acids were increased significantly after erythritol consumption, indicating that this polyol

may be able to affect the function of the human gut microbiota. Indeed, the authors reported that this change in SCFAs production was due to the 10% of erythritol that reaches the human colon [57].

#### 2.2.3. Isomalt

Hydrogenated isomalt, isomaltitol (E-953), is not absorbed by the small intestine and is easily fermented in the colon by the microbiota [61]. This fermented fraction of ingested isomalt is approximately 90% [62]. Therefore, itis expected thatisomaltis capable of altering the bacterial population. Isomalt has been proposed as a prebiotic carbohydrate that might contribute to a healthy luminal colonic mucosal environment, with bifidogenic properties and high butyrate production [63]. Accordingly, besides evidence reported in Ruiz-Ojeda et al., 2019 [1], a recent study based on the administration of buckwheat honey to human gut microbes cultures reported that the principal constituents of buckwheat honey are oligosaccharides with a low degree of polymerization, including isomalt and isomaltotriose, which may serve as food to promote the growth of indigenous intestinal probiotics such as *Bifidobacterium* [64]. In addition, an increase in the abundance of *Escherichia*/*Shigella* and *Streptococcus* was also reported, while the alpha diversity, as well as the abundance of *Prevotella*, *Faecalibacterium*, and *Lachnospiraceae incertae sedis*, were decreased, thus fostering a reduction of pathogenic bacteria in the gut tract [64]. However, this might be also explained by the polyphenols composition of the buckwheat honey studied, since polyphenols also markedly affect the gut microbiota [64]. Indeed, the authors concluded that phenolic compounds and oligosaccharides in buckwheat honey appear to synergistically impact human intestinal microbes to enhance the growth of probiotics. More efforts, especially in vivo, are required to elucidate the possible specific impact of isomalt on the gut microbiota.

#### 2.2.4. Lactitol

Lactitol (E-966) is a disaccharide normally not absorbed in the small intestine [65] that therefore reaches the lower gut where it is fermented, producing both gases and SCFA [66]. Lactitol mitigates pathogenic translocation in the small intestine by the reduction of permeability and stimulates the growth of bifidobacteria and lactobacilli [67]. Thus, similarly to isomalt, lactitol could act as a prebiotic, enhancing the composition of the intestinal microbiota, even when consumed at low doses as a sweetener, normally 10 grams [68]. Nevertheless, it is important to highlight that lactitol, due to its limited sweetening power, is usually used in combination with other intense sweeteners [69] or a set of prebiotics [70], and this could disturb the results concerning its effect on the intestinal microbiota. Furthermore, it has also been studied as a synbiotic product along with *Lactobacillus acidophilus* NCFM and jointly promoted beneficial changes since it led to a decrease in the abundance of the *Blautia coccoides*–*Eubacterium rectale* bacterial group and *Clostridium* cluster XIVab counts in the elderly population [71]. Since 2018, two studies were identified regarding lactitol and the gut microbiota. One trial was based on the administration of probiotics, synbiotics, probiotics together with lactitol, or only lactitol to mice with acute colitis. The authors found that the lactitol group showed higher levels of *Akkermansia* compared with the control, probiotic (*Bifidobacterium* and *Lactobacillus*), and synbiotic (probiotics and inulin) groups. It is worth highlighting this work, since *Akkermansia* seems to ameliorate the inflammatory response and insulin resistance in obese and diabetic patients [72], protecting the intestinal epithelial cells and enhancing the mucosal barrier function [73]. As the genome of *Akkermansia* was proved to be able to encode a wide variety of secretory proteins such as glycohydrolyzases [74], the authors speculated that *Akkermansia* might be able to decompose lactitol and promote its own proliferation [75]. Furthermore, the supplementation of probiotics and prebiotics, including lactitol, induced an increment of the proportion of helpful bacteria and regulated the balance of the intestinal microbiota [75]. For instance, the abundance of *Bifidobacterium* was increased in all the experimental groups in comparison with the control. However, the observed effect might not be exerted by lactitol itself [75]. Another study was performed in Korean adults to evaluate the efficacy of supplementation with the prebiotic UG1601 (based on inulin (61.5%), lactitol (34.6%), and an aloe vera gel (3.9%)) for 4 weeks to alleviate the symptoms of constipation associated with the gut microbiota [70]. Here, the clinical trial showed that the prebiotic UG1601 in patients with mild constipation resulted in decreased serum concentrations of the bacterial endotoxin lipopolysaccharide and its receptor CD 14. Additionally, it increased the abundance of *Roseburia hominis*, a major butyrate producer, which could be related to the observed reduction of the levels of these endotoxemia markers [70]. In summary, lactitol along with other compounds, may induce changes in the gut microbiota, but further studies are needed to demonstrate whether lactitol itself triggers an effect on the gut microbiota.

#### 2.2.5. Maltitol

Maltitol (E-965) is obtained through the hydrolysis, reduction, and hydrogenation of starch. This polyol has a very slow absorption rate, being fermented in the colon. Thus, as we previously mentioned, it is expected that maltitol is susceptible fermentation by the gut microbiota [1]. To date, only one clinical trial has been reported which studied the effect of maltitol present in experimental chocolate on the gut microbiota. The authors concluded that both maltitol and polydextrose, as well as maltitol alone, increased the amount of fecal bifidobacteria, lactobacilli, and SCFA compared with the control chocolate [26]. Besides evidence reported by Ruiz-Ojeda et al., 2019 [1], there are no additional studies. Although maltitol could be a good alternative with high sweetening capacity (~90%), safe, and non-cariogenic, data to determine the specific effects of maltitol on the gut microbiota are not still sufficient.

#### 2.2.6. Sorbitol

Sorbitol or D-glucitol (E-420) is partially absorbed in the upper gastrointestinal tract, where it undergoes digestion, while the non-absorbed portion is extensively fermented to SCFA and gases by the colonic microbiota [62]. Consumers can suffer slight gastrointestinal symptoms, such as flatulence or bloating, or more severe symptoms when it is ingested at high doses as 20 g d−<sup>1</sup> [76]. Overall, studies on this isomeric polyol and its effect on the gastrointestinal tract are mostly focused on the symptomatology induced by sorbitol than on its possible capacity to alter the gut microbiota. Since the 1930s, it is known that sorbitol can be fermented by bacteria like *E. coli*, *Lactobacillus* spp., and *Streptococcus* spp. [77] which are present in our intestinal microbiota. However, so far, there has been no thorough study and there is not enough evidence to define the specific effects of sorbitol on the gut microbiota.

#### 2.2.7. Mannitol

Mannitol (E-421) is an isomer of sorbitol, and both are listed as hydrogenated monosaccharides. Approximately, 75% of ingested mannitol reaches the large intestine [78]. The intestinal bacteria metabolize D-mannitol to butyrate and propionate in animal models. Indeed, D-mannitol has been suggested as a prebiotic, due to its stimulation of colonic butyrate and propionate production [79]. Although no data are available so far about the effects of mannitol on the gut microbiota, its role as a substrate reflects an interaction between this polyol and the intestinal microbiota that should be studied more deeply.

#### 2.2.8. Xylitol

Xylitol (E-967) is a five-carbon polyol obtained from the hydrogenation of D-xylose, called wood sugar or birch sugar. Xylitol can be directly metabolized mainly in the liver, remaining unchanged in the gastrointestinal tract [80]. Furthermore, only a certain proportion of the ingested xylitol is absorbed slowly from the intestinal lumen and fermented by the intestinal microbiota. Besides minor amounts of gases such as H2, CH4, and CO2, the end products of the bacterial metabolism of xylitol are mainly SCFA, (i.e., acetate, propionate, and butyrate). Xylitol might cause osmotic diarrhea when the amounts consumed are too high [81]. Hence, it is expected that this polyol is capable of altering the intestinal microbiota. Interesting results were reported, as previously mentioned, in our recent review [1], but further studies were not reported since then.

In summary, according to the new findings reported from February 2018, erythritol, lactitol, and maltitol have shown to exert beneficial effects on the gut microbiota by themselves. Nevertheless, because of the promising effect of lactitol to enhance *Akkermansia* proliferation in mice with acute colitis, we encourage corroborating this finding by further studies in humans. Overall, the latest evidence is not still enough to establish firm conclusions in relation to how polyols influence the gut microbiota. In addition, it is necessary to highlight that some polyols could induce laxative effects, and it would be more reliable to evaluate their effects separately. Figure 1 summarizes the effects of different sweeteners on intestinal microbiota.

**Figure 1.** Schematic representation of sweeteners' effects on the gut microbiota. Abbreviations. IgA, Immunoglobulin A; N/A, not available information; SCFA, short-chain fatty acids.

#### **3. Conclusions and Future Perspectives**

The effects of sweeteners on gut microbiota composition are still in discussion. Even though there are some gaps in the evidence related to the health effects of NNS in both healthy and non-healthy populations, authorities such as FDA, EFSA, and Codex Alimentarius consider them safe and well-tolerated, as long as the appropriate ADI is not exceeded. Regarding NNS, neither aspartame nor its degradation products make contact with the colonic microbiota. In contrast, though Ace-K is absorbed and eliminated by urine and almost does not contact the colonic microbiota, surprisingly, it increases Firmicutes and depletes *A. muciniphila.* However, further research is required in order to firmly establish an effect in humans. We previously reported that saccharin and sucralose seem to change the composition of the gut microbiota. However, it is necessary to take account that only 15% of the consumed saccharin contacts the colonic microbiota, so only high doses could alter the intestinal microbiota composition. On the contrary, more than 85% of the consumed sucralose reaches the colon; therefore, sucralose could potentially either alter or change the gut microbiota composition, but it is not practically metabolized by intestinal bacteria. On the other hand, steviol glycosides directly interact with the intestinal microbiota and need bacteria for their metabolization, so they could potentially alter the bacterial population.

In summary, in the absence of biological plausibility, results indicating a possible alteration of the intestinal bacteria population after the consumption of LNCS should be explained by alternative mechanisms, such as alterations in the dietary pattern, administration of exaggerated LNCS doses, and co-administration of carriers.

**Author Contributions:** J.P.-D. and F.J.R.-O. design the work, participated in the bibliographic search, discussion, and writing of the manuscript. B.P.-V. and A.R.-R. participated in the bibliographic search, discussion, and writing of the manuscript. F.A.-M. revised the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** Julio Plaza-Díaz is part of the "UGR Plan Propio de Investigación 2016" and the "Excellence actions: Unit of Excellence on Exercise and Health (UCEES), University of Granada". Francisco J. Ruiz-Ojeda and Belén Pastor-Villaescusa are supported by a grant to postdoctoral researchers at foreign universities and research centers from the "Fundación Alfonso Martín-Escudero", Madrid, Spain. We are grateful to Belen Vazquez-Gonzalez for assistance with the illustration service.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
