**Comparison of Di**ff**erent Dietary Indices as Predictors of Inflammation, Oxidative Stress and Intestinal Microbiota in Middle-Aged and Elderly Subjects**

**Sergio Ruiz-Saavedra 1,2,3, Nuria Salazar 1,3, Ana Suárez 2,3, Clara G. de los Reyes-Gavilán 1,3, Miguel Gueimonde 1,3 and Sonia González 2,3,\***


Received: 12 November 2020; Accepted: 12 December 2020; Published: 15 December 2020

**Abstract:** During the last decades the gut microbiota has been identified as a key mediator in the diet-health interaction. However, our understanding on the impact of general diet upon microbiota is still limited. Dietary indices represent an essential approach for addressing the link between diet and health from a holistic point of view. Our aim was to test the predictive potential of seven dietary ratings on biomarkers of inflammation, oxidative stress and on the composition and metabolic activity of the intestinal microbiota. A cross-sectional descriptive study was conducted on a sample of 73 subjects aged >50 years with non-declared pathologies. Dietary inflammatory index (DII), Empirical Dietary Inflammatory Index (EDII), Healthy Eating Index (HEI), Alternative Healthy Eating Index (AHEI), Mediterranean adapted Diet Quality Index-International (DQI-I), Modified Mediterranean Diet Score (MMDS) and relative Mediterranean Diet Score (rMED) were calculated based on a Food Frequency Questionnaire. Major phylogenetic types of the intestinal microbiota were determined by real time polymerase chain reaction (qPCR) and fecal short chain fatty acids (SCFAs) by gas chromatography. While DII, HEI, DQI-I and MMDS were identified as predictors of *Faecalibacterium prausnitzii* levels, AHEI and MMDS were negatively associated with *Lactobacillus* group. HEI, AHEI and MMDS were positively associated with fecal SCFAs. In addition, DII and EDII explained lipoperoxidation level and Mediterranean scores the serum IL-8 concentrations. The lower detection of IL-8 in individuals with higher scores on Mediterranean indices may be partially explained by the increased levels of the anti-inflammatory bacterium *F. prausnitzii* in such individuals.

**Keywords:** dietary patterns; Mediterranean diet; dietary indices; microbiota; elderly

#### **1. Introduction**

The empirical relationship between diet and health has been recognized since the time of Hippocrates (400 BC). During the last decades, solid scientific evidence has accumulated on the protective role of certain foods, such as fruits and vegetables, on the risk of suffering non-transmissible diseases [1,2]. However, understanding the net impact of diet on health is more complex than studying isolated components. Humans do not consume single foods but a wide variety of combinations of foods forming the so-called dietary pattern. Therefore, from a physiological point of view, the analysis of the eating habits, considering the interactions between different foods and their components, is of paramount interest [3,4]. In this context, dietary indices have been developed as a useful tool for categorizing dietary practices in different populations, encouraging the comparison among different studies [5,6]. The currently available dietary indices could be clustered into three main categories: the inflammatory ones [7,8], those quantifying the adherence to dietary guidelines [9,10] and those evaluating the degree of adaptation to the Mediterranean dietary pattern [11]. Both, the number and type of components included in each index is different, depending on the purpose for which they have been created and the dietary habits of the population for which they have been designed. While inflammatory ratings, such as the Dietary Inflammatory Index (DII) [7] or the Empirical Dietary Inflammatory Index (EDII) [8], have proven to be useful in the prediction of inflammatory parameters such as C-reactive protein (CRP), interleukin 6 (IL-6) or adiponectin levels [12–14], the Healthy Eating Index (HEI) [15] or the Alternative Healthy Eating Index (AHEI) [10], based on diet quality, have been useful for assessing the risk of chronic diseases [10,16]. Among the different indicators, the Mediterranean dietary index is perhaps the one accumulating more scientific evidence about its beneficial impact on morbidity and mortality [17] through the reduction of different parameters related to oxidative stress [18,19].

Therefore, these indices represent a key tool in the assessment of the association between diet and health. Moreover, the inclusion of some novel biological parameters, such as the gut microbiota, in the study of such correlations may broaden their applicability [20,21]. In this regard, HEI and Mediterranean ratings were recently found to be associated with gut microbiota in terms of both microbial composition and diversity [22–24]. In more detail, MedDietScore index has been associated with higher fecal bifidobacteria: *Escherichia coli* ratio, total bacteria and short chain fatty acids (SCFAs) [25]. However, the studies in this area are still limited and there are no studies comparing the different indices for assessing the interaction between regular diet and microbiota.

In view of this evidence, the main objective of this work was to analyze the diet of a group of middle aged and elderly participants, without declared pathologies, through different dietary indices and to examine their predictive potential on parameters related to inflammation, oxidative stress and the composition and metabolic activity of the intestinal microbiota.

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

#### *2.1. Participants*

The sample included 73 volunteers recruited in Asturias (North of Spain), aged between 56 and 95 years and with a Body Mass Index (BMI) [weight (kg)/height (m2)] from 19.9 to 37.5 kg/m2. In an individual interview, volunteers were informed about the objectives of the study and an informed written consent was obtained before enrolment. Exclusion criteria were the presence of diagnosed immune or digestive related pathologies as well as consumption of corticoids, immunosuppressive drugs, monoclonal antibodies, probiotics or antibiotics in the previous month. The study was approved by the Regional Ethics Committee for Clinical Research (Servicio de Salud del Principado de Asturias nº 17/2010).

#### *2.2. Nutritional Assessment*

A previously validated annual semi-quantitative food frequency questionnaire (FFQ) [26,27] was used by trained personnel to assess volunteers' regular food intake in a personal interview of approximately 1 h duration. Methodological issues about dietary assessment were published elsewhere [28]. Food composition tables of CESNID (Centro de Enseñanza Superior de Nutrición Humana y Dietética) were used to transform food consumption into energy and macronutrients intake [29]. The (poly)phenol content in foods was completed using the Phenol Explorer database [30]. Fiber components were determined using the Marlett et al. food composition tables [31]. Glucosinolates levels were obtained from McNaughton et al. [32] and isothiocyanates along with aliphatic glucosinolates food content were derived from glucosinolates levels following European Prospective Investigation into Cancer and Nutrition (EPIC) criteria [33]. Glucosinolate side chains concentrations were ascertained from International

Agency for Research of Cancer (IARC) data [34]. At the time of carrying out the blood extraction, height and weight were taken by standardized protocols [35]. At the same time as the FFQ interview was conducted, a questionnaire on socio-economic factors (such as level of education or type of work) and lifestyle (physical activity, smoking and self-perception of health status) was administered.

#### *2.3. Dietary Indices Calculation*

A calculation of seven dietary indices was carried out, including DII, EDII, HEI, AHEI, Mediterranean adapted Diet Quality Index-International (DQI-I) [36], Modified Mediterranean Diet Score (MMDS) [37] and relative Mediterranean Diet Score (rMED) [38], as shown in Supplementary Table S1. IBM SPSS program version 24.0 (IBM SPSS, Inc., Chicago, IL, USA) was used to design a database for calculating indices scores from our FFQ data.

DII scores were calculated by evaluating 35 parameters (out of 45 possible items). Components such as eugenol, ginger, saffron, turmeric, green/black tea, isoflavones, pepper, thyme/oregano and rosemary were excluded because a lack of information about them in the FFQ recordings. First, the consumption levels of parameters were standardized by subtracting daily global consumption mean and dividing by the global standard deviation. The resulting Z-scores were then converted to percentile scores and centered by doubling and subtracting one. These centered-percentile scores were multiplied by the overall food parameter-specific inflammatory effect score to obtain the 'food parameter-specific score' (FPES). All FPES of an individual were summed to obtain the final DII Score. Values in our sample ranged from −4.62 to 4.45, with negative scores predicting lower inflammation and positive scores higher dietary-derived inflammation.

For EDII scores, 18 components were accounted. The number of servings consumed was calculated for each component. The resulting values were multiplied by the "Weight" of the components and divided by 1000. All weighted components were summed to obtain each EDII Score, with values ranging from −1.56 to 2.21 in our sample. The more positive the result, the more prone to higher concentrations of inflammatory biomarkers.

The HEI is an index comprised of 13 components. For each parameter, the amount of a dietary component, in g, cups or oz equivalents were calculated per 1000 kcal. These densities were scored according to recommended consumption values. For negative scoring components, considered to be consumed in small quantities due to their negative impact on health, the individuals get higher scores when the consumption values are lower than the established threshold. Then, all components were summed to obtain the HEI score. HEI was reflecting a total score from 29.49 to 77.76 in our sample, with higher values reflecting a healthier diet.

AHEI score calculations procedure was quite similar to that of HEI, with 11 components in total and values from 37.34 to 80.79 in our sample. In the same way, higher values represent healthier diets. For each component, the consumption in servings per day was calculated and scored according to AHEI-2010 criteria. All components were summed to obtain the total AHEI score of each individual.

Considered the most suitable index for international analyses, DQI-I accounts for 18 components. Here, we evaluated a Mediterranean adaptation of DQI-I. The amount consumed of each component was rated from 0 to 3, 5 or 6, depending on the component and all the points were summed to obtain DQI-I scores, whose final values were between 33 and 68 (minimum and maximum). Here, a total sum of 100 would show an individual with perfect adherence to the main four categories evaluated in this index, while a score of 0 is reflecting a diet far away from recommended dietary guidelines.

Composed of 9 components, rMED and MMDS were derived from the index originally developed by Trichopoulou et al. [11] to evaluate the degree of adherence to a traditional Mediterranean diet. rMED was calculated from components intake based on the density nutrients model. Once amounts of consumption per 1000 kcal were obtained, the sample was split in tertiles for each of the nine components. According to the tertile position for each component, individuals were rated. Summing all the ratings we obtained the rMED score, ranging from 2 to 12 in our sample. In the case of MMDS, each parameter was scored 0 or 1 according to the cut-off values of the sex specific-medians among the

participants. Final MMDS scores ranged from 1 to 7 in our sample. In both indices, the highest scores are showing a higher adherence to Mediterranean diet patterns.

#### *2.4. Blood Biochemical Analyses*

Fasting blood samples were drawn by venipuncture and centrifuged (1000× *g*, 15 min). Plasma and serum aliquots were kept at −20 ◦C for later analyses. Serum glucose, serum total cholesterol, serum HDL-cholesterol, serum LDL-cholesterol and serum triglycerides were determined by using an automated biochemical autoanalyzer in an independent laboratory. Serum levels of CRP were determined by ELISA (CRP Human Instant ELISA, Ebioscience, San Diego, CA, USA) and malondialdehyde (MDA) by a colorimetric assay of lipid peroxidation (Byoxytech LPO-586 assay, Oxis International S.A., Paris, France) [39]. Serum leptin was determined by ELISA (Human Leptin ELISA Development Kit 900-K90, PeproTech Inc., Rocky Hill, NJ, USA) according to the manufacturer's instructions. Colorimetric assay P40117 (Innoprot, Innovative Technologies in Biological Systems, S.L., Spain) was used to determine total antioxidant capacity (TAC) in serum [40]. A multiplex immunoassay (Cytometric Bead Array, CBA, BD Biosciences) by flow cytometry allowed to quantify levels of serum IL-10, Tumor Necrosis Factor-Alpha (TNF-α), IL-8, IL-17 and IL-12, while the concentration of transforming growth factor (TGF)-β was determined by ELISA (BD OptEIATM, BD Biosciences). The phagocytic capacity was quantified in a FACSCanto II Flow Cytometer (Becton Dickinson, BD Biosciences, San Diego, CA) by using the Phagotest® kit (Orpegen Pharma, Heildelberg, Germany). Natural killer (NK) cell activity was determined by flow cytometry, using the NKtest® kit (Orpegen Pharma).

#### *2.5. Fecal Collection and Microbial Analysis*

Detailed instructions about fecal samples collection were given to participants who also were provided with a sterile container. After deposition, samples were immediately frozen at −20 ◦C and transported to the laboratory. For analyses, samples were thawed at room temperature (24 ± 2 ◦C), weighed, diluted 1/10 in sterile PBS and homogenized using a LabBlender 400 Stomacher (Seward Medical, London, UK) for 4 min; the DNA was extracted using the QIAamp DNA stool mini kit (Qiagen, Hilden, Germany) following previously described procedures [41]. 7500 Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) and SYBR Green PCR Master Mix (Applied Biosystems) were used to achieve the quantification of bacterial populations, including the major bacterial groups present in the human gut (Supplementary Table S2). Procedure instructions were published elsewhere [41]. Fecal DNA extracts were analyzed and the mean quantity per gram of fecal wet weight was calculated for each bacterial group. One milliliter of the homogenized feces were centrifuged and supernatants were analyzed by gas chromatography to determine acetate, propionate and butyrate concentrations, as previously indicated [42]. A chromatograph 6890N (Agilent Technologies Inc., Palo Alto, CA, USA) connected to a mass spectrometry detector (MS) 5973N (Agilent Technologies) and a flame ionization detector (FID) was used for identification and quantification of SCFAs, respectively, as described previously [43].

#### *2.6. Statistical Analyses*

Statistical analysis was performed using the IBM SPSS program version 24.0 (IBM SPSS, Inc., Chicago, IL, USA). Mean dietary scores were analyzed by a Student t-test and Bonferroni multiple comparison according to general, socio-economic and health-related characteristics, such as smoking status, educational level, mood feeling or self-health perception among others. Similar procedure was performed to analyze mean levels of microbiological and blood variables according to age group. Goodness of fit to the normal distribution was analyzed employing the Kolmogorov−Smirnov test. Scores were examined as predictors of gut microbial groups, fecal SCFAs and blood biomarkers by regression analyses controlling by age and energy intake. These variables included *Akkermansia*, *Bacteroides-Prevotella-Porphyromonas*, *Bifidobacterium*, *Clostridia* cluster XIVa, *Lactobacillus* group,

*Faecalibacterium prausnitzii*, acetic acid, propionic acid, butyric acid, glucose, triglycerides, Low density lipoprotein-high density lipoprotein (LDL-HDL) ratio, leptin, serum malondialdehyde (MDA), (total antioxidant capacity) TAC, C-Reactive protein (CRP), TGF-β, IL-10, IL-17, IL-8, IL-12, TNF-α, Phagocytosis granulocytes (%), Phagocytosis granulocytes and monocytes (%) and NK cell activity. When the distribution of variables was skewed (CRP, TGF-β, IL-10, IL-17, IL-8, IL-12, TNF-α) the values were converted to their natural logarithm. Association of microbial groups with food groups, macronutrients and micronutrients were evaluated through linear regression analyses adjusting by age and energy. The resulting data were plotted in a heatmap using the "pheatmat" function of the R program (version 3.5.1 for Windows). A Pearson correlation test was performed to elucidate closeness among dietary indices. This information was introduced in the program R, using the package "pheatmap," to clusterize and plotting indices based on Euclidean distances. To test the association among dietary indices and previously reported health-beneficial dietary compounds, linear regression analyses were performed and plotted as forehead mentioned. To indicate statistical significance in the interpretation of results, the probability value of 0.05 was used.

#### **3. Results**

The average score on dietary indices according to general characteristics and socio-economic status, lifestyle and health-related factors of the studied sample is presented in Table 1. Among all the variables examined, only significant differences were observed for the age and these were found in all the studied dietary indices. Subjects over 65 years presented worse dietary scores than those in the group of 50–65 years. At the time of interpreting the results it should be taken into account that unlike the rest of the indices studied, in DII and EDII a higher score is associated with a more pro-inflammatory diet. Both DII and DQI-I showed better scores in people with energy intake higher than 1994.8 kcal. EDII score was found to be higher (worse dietary quality) as BMI increases. A lower score in the AHEI, DQI-I and Mediterranean dietary indices (rMED, MMDS), associated with worse dietary quality, was found in those subjects with bad self-health perception. Therefore, age and energy intake have been introduced in further analyses carried out as a covariate.

A general description of the variables that are subsequently analyzed in the study according to age groups is shown in Table 2. The levels of the bacterial groups analyzed were in the range of those previously reported in similar populations and demonstrated the large inter-individual variability present in the human adult fecal microbiota. Significant differences were observed in most of the microbiological parameters analyzed according to age. Subjects over 65 years of age presented lower fecal levels of *Bacteroides-Prevotella-Porphyromonas* group, *Clostridia* cluster *XIVa* and *Faecalibacterium,* as well as all the short chain fatty acids determined. Blood parameters are within the normal physiological ranges and were similar between the groups evaluated except for MDA, IL-8, IL-12 and TNF-α, whose concentration is higher in subjects over 65 years of age.

In order to analyze possible linear relationships among the different dietary indices scores and fecal microbial groups, a linear regression analysis was conducted and a heatmap was plotted (Figure 1). Lower scores on indices related to diet quality (HEI, AHEI, DQI-I), suggestive of an unhealthier diet, were associated with increased *Akkermansia* levels. While DII, HEI, DQI-I and MMDS have shown potential as predictors of *F. prausnitzii*, which showed higher levels in those individuals with healthier diets, only AHEI and MMDS were negatively associated with *Lactobacillus* levels. These results were further examined including other health-related parameters in the analysis (Table 3). In relation to the production of SCFAs, higher scores in HEI, AHEI and MMDS indices were positively associated with the formation of acetic, propionic and butyric acids. Furthermore, as expected, inflammatory indices (DII, EDII) were the best determinants of lipoperoxidation blood levels, while Mediterranean ones were the best identifiers of serum IL-8 concentrations.


**Table** 




values showing a higher degree of adherence

 to the

Mediterranean

 diet. **Table 2.** Microbial levels, short chain fatty acid (SCFA) concentration and blood biomarkers according to age groups.


**Table 3.** Results obtained from regression analyses to identify dietary indices as predictors of gut microbiota levels (log10 n◦. cells per gram of feces), fecal short chain fatty acids (mM) and blood biomarkers.


Linear regression analyses are adjusted by age and energy. R2, coefficient of multiple determination; β, standardized regression coefficient for examined variable. *p* ≤ 0.05. Inflammatory indices (DII, Dietary Inflammatory Index. EDII, Empirical Dietary Inflammatory Index): negative values favor non-inflammatory states and positives values enhance inflammation. Dietary Quality indices (HEI, Healthy Eating Index. AHEI, Alternative Healthy Eating Index. DQI-I, Diet Quality Index-International): higher scores are reflecting consumption values similar to those recommended in dietary guidelines. Mediterranean Dietary indices (rMED, Relative Mediterranean Diet Score. MMDS, Modified Mediterranean Diet Score): higher values showing a higher degree of adherence to the Mediterranean diet.

To deepen into how dietary components may modulate gut microbiota and SCFAs, linear regressions models adjusting by age and energy intake were applied and β-coefficient values were plotted in the heatmap of Supplementary Figure S1. Sauces and dips seemed to be the most significant food group for determining *Clostridia* cluster XIVa and *Lactobacillus* groups. However, the consumption levels of this food group were very low in our sample (data not shown) so that this association should be considered with caution. *F. prausnitzii* showed a positive association with fruits, legumes and with fiber, whereas a negative association was found for saturated fats. SCFAs appeared to be mostly related to oils and fats, seafood, total polyphenols and saturated fats.

Heatmap showing Pearson´s correlation and clusterization among the different dietary indices is presented in Figure 2. Here, inflammatory indices revealed to be closer between them, forming the first cluster while the others integrate the second cluster. At the same time, HEI and AHEI grouped more closely with DQI-I and then with rMED and MMDS.

**Figure 2.** Heatmap defined by Pearson's correlations between dietary indices scores. Blue and red colors represent negative and positive association, respectively. The color intensity is proportional to the degree of association between indices. Asterisks indicate correlation significance: \* *p* ≤ 0.05; \*\* *p* ≤ 0.01. Due to the scale of the Dietary inflammatory index (DII) and the Empirical Dietary Inflammatory Index (EDII), they show an inverse relationship with the rest of the indices.

To elucidate the relationship of dietary indices scores with the consumption of dietary compounds with reported anti-inflammatory and health-protective effects in the literature, a new linear regression analysis was performed adjusting by age and energy and a heatmap was plotted with the results (Figure 3). DII showed more negative correlation with vitamins, fiber, glucosinolates and isothiocyanates than EDII. All indices were associated with total polyphenols and ORAC. Except for rMED, indices had a significant correlation with flavonols, DHA, lutein + zeaxanthin, carotenoids, insoluble fiber and pectin. Glucosinolates and isothiocyanates did not show a significant association with the Mediterranean indices. In general, splitting indices into inflammatory, Mediterranean and diet quality ones, heatmap revealed similar patterns of association for the indices within each group. With the exception of flavones, all evaluated compounds revealed a negative association with inflammatory indices but positive with the Mediterranean and diet quality indices.

**Figure 3.** Heatmap showing β-coefficient values resulting from univariate linear regressions adjusting by age and energy among six different dietary indexes and health-related compounds. Rows include compounds with previously reported beneficial effect as (poly)phenols, fatty acids, carotenoids, vitamins, fiber, glucosinolates and isothiocyanates. Blue and red colors denote negative and positive association, respectively. Asterisks indicate the significance of the association degree \* *p* ≤ 0.05; \*\* *p* ≤ 0.01.

#### **4. Discussion**

In recent years, dietary indices have been a major step towards addressing the diet-health binomial from a global perspective. While tailored to different populations and constructed for specific purposes, a high degree of similarity can be observed among some of them. To our knowledge, this is the first study comparing the usefulness of dietary indices as predictors of human gut microbiota, the production of fecal SCFAs and the concentration in blood of different parameters related to the immune and inflammatory status, in a sample of middle-aged and elderly subjects without diagnosed pathology. As it has been previously proposed by other authors, for some of the indices included in the study, our data showed the existence of differences according to age. Furthermore, changes in the levels of some bacterial groups such as *Akkermansia* [44] or butyrate-producing bacteria, mainly *F*. *prausnitzii* [45,46], have been reported in aged individuals. Therefore, the factor "age" has been introduced in the models as a covariate in order to improve the interpretation of results.

Limitations in the use of dietary indices need to be also taken into account. DII encompasses a total of 45 components, of which 35 were evaluated in this sample. The effects of lacking 10 components in the scoring system may be attenuated by the own nature of the index and the fact of having a

sample that does not show extreme values of consumption for any component. Moreover, nutritional supplements, weighted in other versions of EDII to compute the final score, were not included here. Some dietary indices such as EDII, HEI or AHEI were developed in the context of almost fully "westernized" societies, which could drive to an underrating or overrating in diets of populations with mixed diets (Mediterranean, African, etc.). Indeed, in our sample, the component "Trans Fat" that is evaluated in the AHEI, showed very low values while in typical EEUU diet is probably highly present. This may entail a loss of power in the accuracy of the prediction of our scores. Furthermore, DQI-I incorporates dietary variety, adequacy and moderation as a quality criterion. Although all these parameters have been included, the accuracy of FFQ to provide an accurate measurement of variety may be one of the limitations of this work. One of the main difficulties arises in the capacity of the indices to classify the subjects under study. In this sense, the scores from dietary quality indices showed low variability inter subject, with almost the whole sample obtaining scores indicative of poor or average diet quality for DQI-I, HEI and AHEI. Thus, we propose these indices as the worse in differentiating the poor-quality Mediterranean-style diet from a middle-age-elderly sample. On the other hand, indices related to adherence to a Mediterranean-style dietary pattern have presented a wide range of scores in the sample, ranging from 2–12 points and 1–7 for rMED and MMDS, respectively.

Based on the data obtained, the identification of the best tool to predict the composition and metabolic activity of the gut microbiota as a function of diet, is a difficult task. We considered that in the present study, the HEI, AHEI and DQI-I resulted likely inappropriate as predictor variables of differences between different microbiota as poor diets, that are different but score similarly, may mask trends associated to specific dietary constituents [47]. Interestingly, we found decreased levels of the mucin degrading *Akkermansia* in better scoring individuals compared to increased levels of *Akkermansia* in the worse ones, mainly influenced by vegetable consumption (data not shown). When vegetables are included at significant levels in diet, fiber consumption increases, which could promote the rise of some fiber-degrading species at the expense of other microorganisms such *Akkermansia* [48]. This could contrast with the fact that the presence of *Akkermansia* has been associated with healthy intestine and its abundance has been inversely correlated to several disease states [49–54]. We propose Mediterranean indices and more precisely MMDS, as the most accurate and best predictor in our population sample. Probably, socio-geographical reasons do Mediterranean indices the most suitable ones to measure the quality of diet in the sample and therefore, to predict microbiological and immunological variables. Also, some dietary indices related with inflammation (DII), quality of diet (HEI and DQI-I) and adherence to the Mediterranean-diet (MMDS) seem to be predictors of *F. prausnitzii* fecal levels, which were higher in individuals with healthier diets. *F. prausnitzii*, a member of the commensal microbiota, has been related with intestinal health and gut homeostasis [55]. Several studies highlight the anti-inflammatory properties of *F. prausnitzii* and its ability for upregulating T cell production and reducing IL-8 levels by blocking the NF-kB activation [56,57]. *F. prausnitzii* is a member of *Clostridium* cluster IV, one of the main producers of butyrate in the human colon [58] during fermentation of nondigestible polysaccharides such as dietary fiber. Butyrate plays several pleiotropic effects on host physiology and enhances the protection against pathogen invasion [59,60]. Remarkably, *Faecalibacterium* was found to be at high abundance in an Irish elderly sample [61] whereas some studies showed decreased levels of *F. prausnitzii* in centenarians as compared with younger adults [62]. Further studies are needed to determine the role of these bacteria in the intestinal microbiota of elderly populations.

The indices AHEI and MMDS (related to quality of diet and adherence to Mediterranean diet, respectively) were negatively associated with intestinal *Lactobacillus* levels. Increased *Lactobacillus* levels have been correlated with a higher PUFA/SFA ratio intake, probably mediated by changes in bile acid secretion and composition [23,63,64]. Extra virgin olive oil is an important component of Mediterranean diet and is a source of unsaturated fatty acids that can be metabolized by some intestinal *Lactobacillus* species [65–67]. Dietary indices used in the present work add from 0 to different positive numerical values to the formula relating dietary fats, as depending on the type and amount of fat consumed (calculated as a percentage of total energy intake). Therefore, as higher scores in dietary

indices are generally accompanied in the general population by lower consumption of all type of fats (correlation values of "Lipids" with AHEI and MMDS of −0.154 and −0.149, respectively), this could provide a rationale to the inverse association found by us between scores for AHEI and MMDS and fecal levels of *Lactobacillus*. In this regard, we recently reported increased levels of the *Lactobacillus* group in Spanish adults displaying altered profiles of serum free fatty acids, which were accompanied by subclinical metabolic alterations [68].

Regarding the fecal SCFAs evaluated in our sample, acetic and propionic acids, correlated positively with heathier dietary scores for most of the indices. Microorganisms colonizing the gastrointestinal tract can participate in beneficial interactions within the intestinal ecological niche, as modulated by external factors such as diet. This is the case of the increase of the intestinal butyrate production by cross-feeding mechanisms. In cross-feeding, fiber-degrading bacteria can produce acetate as an end-product of fermentation, which is then metabolized by other members of the intestinal microbiota, as those belonging to *Clostridia* clusters XIVa and IV, to produce butyrate as an end-product of fermentation [51]. Both, bacteria and SCFAs contribute to cell expansion, immunosuppressive functions and overall intestinal homeostasis. Therefore, better dietary scores could be related with an enhanced SCFAs production in the gut [23].

Inflammatory indices have been identified as good predictors of inflammation variables such as CRP, IL-6 and TNF-α receptor 2 [12–14,69,70]. MDA is considered an oxidative stress biomarker that reflects levels of lipoperoxidation in blood. However, to the best of our knowledge, there are no previous reports relating dietary inflammatory indices and MDA. The lower detection of IL-8 in those individuals in the present work showing better scores on the indices related with the adherence to the Mediterranean diet may be partially explained by the increased levels of the anti-inflammatory bacterium *F. prausnitzii* [71] in such individuals. This bacterium was able to block the production of the inflammatory interleukin IL-8 in Crohn disease patients and in a murine colitis model [72].

#### **5. Conclusions**

The associations found among intestinal bacterial groups, SCFAs, blood biomarkers and dietary indices are indirectly reflecting how these variables are influenced by the specific components or food groups scoring in each index. When trying to discern differences among indices by clustering them, they are split in 3 main classes: inflammatory, diet quality and adherence to Mediterranean diet. The methodologies followed to construct them, the population they target and the scoring criteria define their nature and the way they correlated with others. The extension of the usefulness of dietary indices may shed some light into how to modulate gut microbiota focusing on dietary patterns.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6643/12/12/3828/s1, Figure S1: Heatmap showing β-coefficient values resulting from univariate linear regressions adjusting by age and energy among microbial groups, SCFAs and food groups and dietary compounds, Table S1: Characteristics of dietary indexes, Table S2: Primers and annealing temperatures used for the quantification of intestinal microbial groups by qPCR.

**Author Contributions:** The authors' responsibilities were as follows: Conceptualization, M.G. and S.G.; Funding acquisition, M.G.; Methodology, S.R.-S., N.S., A.S., C.G.d.l.R.-G. and M.G.; Supervision, C.G.d.l.R.-G. and S.G.; Writing—original draft, S.R.-S. and S.G.; Writing—review & editing, N.S., A.S., C.G.d.l.R.-G. and M.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Plan Estatal de I+D+I through projects AGL2017-83653-R (AEI/FEDER, UE) and RTI2018-098288-B-I00 (MCIU/AEI/FEDER, UE) and by contracts with Biopolis SL (Valencia, Spain), CAUCE Foundation (Oviedo, Spain) and Alimerka Foundation (Llanera, Spain). N.S. is granted a postdoctoral contract awarded by the Fundación para la Investigación Biosanitaria de Asturias (FINBA) and S.R.-S. is the recipient of a Research Training contract awarded under project RTI2018-098288-B-I00.

**Acknowledgments:** We show our greatest gratitude to all the volunteers participating in the study.

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

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 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/).

## *Review* **The Microbiota–Gut–Brain Axis and Alzheimer's Disease: Neuroinflammation Is to Blame?**

**Ashwinipriyadarshini Megur, Daiva Baltriukiene, Virginija Bukelskien ˙ e and Aurelijus Burokas \* ˙**

Department of Biological Models, Institute of Biochemistry, Life Sciences Center, Vilnius University, Sauletekio Ave. 7, LT-10257 Vilnius, Lithuania; avee.megur@gmail.com (A.M.); daiva.baltriukiene@bchi.vu.lt (D.B.); virginija.bukelskiene@bchi.vu.lt (V.B.) **\*** Correspondence: aurelijus.burokas@gmc.vu.lt; Tel.: +370-52234382

**Abstract:** For years, it has been reported that Alzheimer's disease (AD) is the most common cause of dementia. Various external and internal factors may contribute to the early onset of AD. This review highlights a contribution of the disturbances in the microbiota–gut–brain (MGB) axis to the development of AD. Alteration in the gut microbiota composition is determined by increase in the permeability of the gut barrier and immune cell activation, leading to impairment in the blood–brain barrier function that promotes neuroinflammation, neuronal loss, neural injury, and ultimately AD. Numerous studies have shown that the gut microbiota plays a crucial role in brain function and changes in the behavior of individuals and the formation of bacterial amyloids. Lipopolysaccharides and bacterial amyloids synthesized by the gut microbiota can trigger the immune cells residing in the brain and can activate the immune response leading to neuroinflammation. Growing experimental and clinical data indicate the prominent role of gut dysbiosis and microbiota–host interactions in AD. Modulation of the gut microbiota with antibiotics or probiotic supplementation may create new preventive and therapeutic options in AD. Accumulating evidences affirm that research on MGB involvement in AD is necessary for new treatment targets and therapies for AD.

**Keywords:** microbiota; Alzheimer's disease; microbiota–gut–brain axis; neuroinflammation; probiotics

#### **1. Introduction**

Dementia is a non-curable syndrome which over time leads to a progressive decrease in memory, thinking, and the capacity to perform everyday activities [1]. There are alternative forms of dementia which include vascular dementia, dementia with Lewy bodies, and frontotemporal dementia [2], which can be provoked by neurodegenerative disorders, cerebrovascular disease, brain injury [3], and infections [4]. The progression of dementia can result in a lack of consequential speech generation and inability to understand scriptural as well as phonetic language, failure to recognize and identify objects, execution of poor motor skills, and incapability to think abstractly and to execute paradoxical tasks [4,5].

Alzheimer's Disease (AD) is a persistent neurodegenerative (neuronal loss) disorder [6,7] which was first described by Alois Alzheimer in 1906 [8,9] while investigating a female patient Auguste Deter [10]. AD is known to be the major cause of dementia worldwide, mainly observed in the elderly [11], accounting for approximately 60–70% of all dementia cases [12]. The incidence of AD is higher in women than in men. AD is an extremely incapacitating disorder, progressing from slight memory impairments to a complete loss of mental function, and in the long period, resulting in death [13]. AD can affect distinct people in various ways. Most of the common warning signs include depression [14], memory loss, challenge in planning a task and problem-solving skills, confusion in recognizing time, mood swings and personality shifts, poor judgment in motor activities, difficulty in memorizing the literature, etc. [15].

Many factors can contribute to AD, but the greatest risk factors are determined to be exacerbations due to aging [16–18], degradation of anatomical pathways [12], environmen-

**Citation:** Megur, A.; Baltriukiene, D.; ˙ Bukelskiene, V.; Burokas, A. The ˙ Microbiota–Gut–BrainAxis andAlzheimer's Disease: Neuroinflammation Is to Blame? *Nutrients* **2021**, *13*, 37. https://dx.doi.org/ 10.3390/nu13010037

Received: 27 November 2020 Accepted: 22 December 2020 Published: 24 December 2020

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 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 (https://creativecommons.org/ licenses/by/4.0/).

tal factors [19–21], mitochondrial dysfunction [22,23], immune system dysfunction [24,25], and genetic factors including mutations of amyloid precursor proteins (APP) [26,27].

In this review, we will be focusing on the role of the gut microbiota on the brain. We will be discussing the recent findings which show that a disturbance in the microbiota-brain axis can lead to neuroinflammation giving rise to AD. We will be discussing the recent studies which draw attention towards neuroinflammation in the brain, eventually leading to neuronal loss. Finally, we will be focusing on the administration of antibiotics and pre-vand probiotics modulating the brain function and used as a therapeutic agent in curing AD.

#### **2. AD Pathology**

The two major markers contributing to AD progression include amyloid-beta (Aβ) plaques and neurofibrillary tangles (NFTs) [28,29]. It was proposed that Aβ plaques are developed originally in the orbitofrontal, basal, and temporal neocortex regions of the human brain [30,31]. The accumulation of Aβ stimulates NFT formation [32,33]. The main constituent of NFTs is the protein tau in a hyperphosphorylated form. It is a highly soluble protein playing an essential role in maintenance of the stability of microtubules in the axons of neurons [34]. NFTs formed inside the neuron disrupt the microtubule structure and form an insoluble substance, which is detected in the locus coeruleus, and transentorhinal and entorhinal areas of the brain [35]. In the curtailed stage, it can spread to the hippocampus and neocortex [36]. The aggregation of plaques and tangles is followed by microglia recruitment surrounding the plaques [37]. This raises microglial activation and local inflammatory response which advance the neurotoxicity [25]. Aβ has been recognized as an antimicrobial peptide that activates the immune pathways recognized by toll-like receptor 2 (TLR2) leading to neuroinflammation [38].

A recent study has shown that amyloid pathogenesis begins with altered cleavage of APP β-secretase and γ-secretase to produce insoluble Aβ fibrils [22,39] (Figure 1). Aβ then oligomerizes, diffuses into synaptic clefts, and interferes with synaptic signaling [40]. Subsequently, it polymerizes into insoluble amyloid fibrils that aggregate into plaques [31]. This polymerization leads to activation of kinases [30], which can accelerate hyperphosphorylation of the microtubule-associated tau protein and its polymerization into insoluble NFTs [41].

NFTs are fragments of paired and helically wound protein filaments in the cell cytoplasm of neurons [42]. It has the proficiency of stabilizing microtubules and forging interconnections between adjoining microtubules to form a substantial network of microtubules and to hold them together [43]. The hyperphosphorylation of tau protein occurs when it comes into contact with the kinases released due to their abundance in the environment [44]. Its hyperphosphorylation leads to the formation of oligomers [45]. The microtubule becomes highly unstable due to the dissociation of tubule subunits [46] that fall apart and then get converted into enormous chunks of tau filaments, which further aggregate into NFTs [40]. The appearance of NFTs are straight, fibrillary, and highly insoluble patches [27] in the neuronal cytoplasm [47]. The major property known causes an abnormal loss of communication between neurons and signal processing and finally apoptosis of neurons [32]. Phosphorylation of tau is regulated by several kinases, including glycogen synthase kinase-3 (GSK3) and cyclin-dependent kinase 5 activated by extracellular Aβ [48]. Even GSK3 beta and cell division protein kinase 5 are primarily responsible kinases for tau hyperphosphorylation [13], and other kinases like protein kinase C, protein kinase A [49], ERK2, serine/threonine kinase, caspase 3, and caspase 9 also have a prominent role, which may be activated by Aβ [50].

**Figure 1.** Aβ formation: the amyloid precursor protein (APP) is a transmembrane protein of the neuronal cell. In the case when it is cleaved by α-secretase, the formed soluble aggregates can be digested by microglial cells. When APP is cleaved by β-secretase and γ-secretase, it leads to formation of Aβ insoluble aggregates. Such protein aggregation results in amyloid plaques, one of the hallmarks of AD.

#### **3. The Microbiota–Gut–Brain Axis**

A microbiota is an ecological community of commensal microorganisms that live symbiotically and pathogenically in our body [5] and plays a vital role in regulatory functions in health and disease [51,52] (Figure 2). At the level of bacterial strains, the gut microbiota demonstrates tremendous diversity and variation in microorganisms related to the age of the person and can be different in the individuals [53]. To date, it was considered that microbial colonization in the gut was only involved in colon-specific activities, which includes fermentation of carbohydrates, vitamin synthesis, and metabolism of xenobiotics [54,55]. Furthermore, it was also found that the role of the gut microbiota is to act as a barrier for the pathogenic bacteria invading the gastrointestinal tract (GIT) [56].

**Figure 2.** Modulation of the microbiota–gut–brain axis by antibiotics and probiotics. The communication between the gut microbiota and the brain includes neuronal, immune-mediated, and metabolite-mediated pathways. Gut dysbiosis leads to activation of the immune response and alters the production of neurotransmitters as well as bacterial metabolites. These may have a contribution to abnormal signaling through the vagus nerve. Reduction in the integrity of the gastrointestinal barrier causes bacterial migration and inflammation. Pro-inflammatory cytokines induce disruption of the blood–brain barrier permeability. Antibiotics can hinder the growth of certain bacteria, and probiotics have the potential to normalize the gut microbiota in microbiota–gut–brain processes.

> The microbial colonization in humans is estimated to begin at birth. The new born infant is initially colonized by microorganisms common to its mother, which are *Lactobacillus* and *Prevotella* spp. [57]. When compared with healthy and preterm infants, usually delivered by caesarean section, preterm infants seem to have variations in the microbiota [58].

As well, further comparison with elderly people in nursing homes and in the community showed large differences. The individuals in the nursing home had less microbiota attributed to a limited diet [59]. Alterations of the composition of microorganisms due to dietary changes can result in augmentation of several diseases such as obesity, colorectal cancer, inflammatory bowel disease, heart failure, type 2 diabetes, and neurodegenerative disorders (AD, Parkinson's disease, multiple sclerosis, etc.) [52,57,60,61]. Furthermore, antibiotic treatment in early life can modulate the composition of microbiota in the gut later in life and can have a negative impact on the brain functions [62,63].

Numerous studies indicate that gut microbiota can have an influence in synthesizing various neurotransmitters and neuromodulators, which affect gut–brain communication and brain function [64–66]. Signal transduction is complex and can have the propensity to include neural, endocrine, immune, and metabolic pathways. However, its detailed mechanism and signals still have to be elucidated [53,67,68]. Clinical and preclinical studies have shown that gut microorganisms can produce metabolites, which affect brain functioning (Table 1).

**Table 1.** Effect of metabolites on brain produced by gut microbiota.


Bacterial strains such as *Escherichia*, *Lactobacillus*, *Saccharomyces*, and *Bacillus* can synthesize amino acids including gamma-aminobutyric acid, 5-hydroxytryptamine, dopamine, butyrate, histamine, and serotonin, which can play a significant role in emphasizing the brain activity of the individuals [84,85]. These neurotransmitters synthesized can cross the mucosal layer of the intestine and are capable of entering the blood stream [61,86]. It was found that the microbiota of aged individuals with AD have a lower level of bacteria that resulted in decreased butyrate levels [87], which, in turn, could lead to increased inflammation in the brain and the progression of cognitive loss [27,86]. These findings suggest that the microbiota performs numerous vital functions in our body, including releasing biochemical by-products such as SCFA and gases [88]. Moreover, animal studies conducted on pigs and rats showed an effect on memory due to microbiota, *bacillus* and *saccharomyces* [85–87]. Interestingly, a recent study has shown that microbiota transfer from human subjects with obesity led to reduced memory scores in mice, aligning this trait in humans with that of recipient mice [89], where RNA sequencing of the medial prefrontal

cortex of those mice uncovered that short-term memory is associated with aromatic amino acid pathways, inflammatory genes, and clusters of bacterial species [89].

As the GIT of humans are inhabited by numerous microorganisms essential for byproduct formation, it has been recently reevaluated in functional terms and different important mechanisms have been established in the bidirectional connection with the brain [90–92]. This bidirectional connection with the brain is termed as the "microbiota– gut–brain (MGB) axis". MGB refers to a crosstalk between the brain and the gut involving multiple overlapping pathways, including the autonomic, neuroendocrine, vagus nerve, the immune system, or the metabolic processes of gut microorganisms and immune system as well as bacterial metabolites and neuromodulatory molecules [93,94]. The MGB axis mirrors the constant connection between the central nervous system (CNS) and the GIT [95]. A number of rodent studies suggest potential involvement of the gut microbiota in behavioral changes [75,96–98]. The sympathetic and parasympathetic arms of the autonomic nervous system, including the neuroendocrine and neuroimmune systems, are known to be vital pathways in MGB [99]. The precise mechanism that arbitrates gut–brain interplay is not fully comprehended, yet it is suggested that it entails immune, endocrine, and neural pathways, leading to a possible alteration in AD patients or aggravation of inflammation (Table 2). The results from a rat study showed that *Bifidobacterium infantis*, an intestinal resident microorganism, has a link to immune response in the brain [75]. An augmentation in the number of *Lactobacillus casei*, *Bacteroides fragilis*, and *Streptococcus thermophilus* in the rodent intestine showed a positive effect on brain activity and performance [75,98–102]. On the other hand, *Eubacterium rectale*, *Porphyromonas gingivalis*, and *Lactobacillus rhamnosus* can play a vital role in the onset of AD [103–107].

Consideration of the human microbiota as a substantial correspondent to nutrition, health, and disease is a relatively fairly contemporary study, and currently, peer-reviewed studies relating modifications in the microbiota to the etiopathology of human diseases are few [108]. Claims on the potential involvement of the gut microbiota in brain function are made, in part, due to the well-described pathways of communication between the brain and the GIT which has been intensively studied in the area of food intake, satiety, and regulation of the digestive tract [109].


**Table 2.** Roles played by different microorganisms residing in the gut.


**Table 2.** *Cont.*

Incorporation of certain microorganisms, such as probiotics, in diet intake can be used as a therapeutic strategy to reduce neurological disorders. *Bifidobacterium* and *Lactobacillus casei* are two microorganisms which show a beneficial effect on neurological disorders [75,112].

#### **4. Gut Microbiota in AD**

Changes altering the gut microbiota can activate proinflammatory cytokines and increase intestinal permeability, which lead to the development of insulin resistance that is associated with AD [117] (Figure 2). Interestingly, recent work has shown that AD development could start even in the gut and then spread to the brain [118]. In this study, the gastric wall of mice was injected with Aβ1–42 oligomers. Over 1 year, it was observed that the amyloid migrated from the intestine to the brain. Consequently, the translocation of Aβ oligomers from the gut to the brain can have a major contribution in causing AD and neuroinflammation [118].

*Escherichia coli*, *Salmonella enterica*, *Bacillus subtilis*, *Mycobacterium Tuberculosis*, and *Staphylococcus aureus* are some of the bacterial strains that can produce functional extracellular amyloid fibers [107]. These amyloid proteins help the bacterial strains to form biofilms and to strongly bind to each other to resist destruction by physical and immune factors [119]. The amyloids formed by bacteria are different from the CNS amyloids in the primary structure but show resemblance in their tertiary structure [120]. The appearance of bacterial amyloid in the gut can trigger the immune system, which could lead to enhanced immune responses with endogenous formation of neuronal amyloid in the brain [119]. Studies of AD patient's blood and cerebrospinal fluid showed an escalated inflammatory response when compared to healthy adults [107]. In the latter case, the clearance of amyloid is very precise [121].

In a recent study, aged Fischer 344 rats were orally exposed to transgenic *E. coli* producing the extracellular bacterial amyloid protein curli (a type of amyloid fiber protein). The data showed an enhanced alpha-synuclein production in the gut and intensified aggregation of alpha-synuclein in the brain, leading to enhanced microgliosis and astrogliosis. Elevated expressions of TLR2, IL-6, and TNF-α in the brain of animals exposed to curliproducing bacteria were determined. This suggested that bacterial amyloid functions as a trigger initiating alpha-synuclein aggregation through cross-seeding and prime responses of the innate immune system [122].

A profound experiment conducted on the APP transgenic mouse model for AD suggested that variation in the number of microbial strains could lead to amyloid deposition. These APPPS1 mice showed reduced numbers of *Firmicutes* and an increased number of *Bacteroides* in the intestine. The germ-free APP transgenic mice demonstrated a reduction in

cerebral Aβ pathology [123]. This finding strongly points towards the intestinal microbiota forming amyloid-triggering immune responses that can lead to hallmarks of AD.

Clinical studies of the gut microbiota of AD patients as well as microbiota from AD model mice revealed decreased microbial diversity when compared with controls (Table 3). These include decreased levels of *Fusobacteriaceae*, *Firmicutes*, *Actinobacteria*, and *Bifidobacterium* and increased levels of *Bacteroidetes* [54,124]. *Cyanobacteria,* one of the gutresiding bacteria, produces a neurotoxin β-N-methylamino-L-alanine, which interferes with the N-methyl-D-aspartate glutamate receptor and leads to signal dysfunction in AD [125].


**Table 3.** Investigation of microbiota in the gut of human as well as animal models of AD.

Not only the bacterial strains residing in the gut can lead to neurodegeneration but also the invading pathogens, such as *Mycobacterium leprae,* are known to be responsible for demyelination and nerve damage. *M. leprae* assists in initiation of the pathogen by changing the internal environment of Schwann cells and stimulation of apoptotic pathways in cells [131]. *Chlamydia pneumoniae* causing respiratory tract infection has been reported in CNS disorders, including AD [132]. *C. pneumoniae* antigens were also found in the neocortex of AD in association with NFTs and senile plaques [133]. Moreover, *Cladosporium*, *Malassezia*, *Phoma*, *Saccharomyces,* and *Candida* species DNA, polysaccharide, and proteins were observed in the CNS samples of AD patients [134]. Fungal footprints were identified in the cerebrospinal fluid by using PCR and slot bolt assay techniques [135].

Upon infection, various cell signaling pathways can occur in the body, which can activate inflammation. When infectious microorganisms cross the blood–brain barrier, it leads to neuronal death due to inflammation and forms similar hallmarks to AD. Lipopolysaccharide (LPS) is found in many gram-negative bacteria [136], exclusively on the outer membrane [137]. An experiment conducted on animal models has shown that bacterial LPS injection in the fourth ventricle of the brain produced inflammatory and pathological characteristics as observed in AD [138] and the peritoneal cavity led to extended elevation of Aβ in the hippocampal regions of mice resulting in cognitive decline [139]. An in vitro study conducted on *E. coli* confirmed that bacterial LPS advanced amyloid fibrillogenesis [127]. Studies conducted on AD patients confirmed LPS presence in the hippocampus and neocortex brain lysates in which most of the LPS aggregation has been observed in the perinuclear region [129,140]. The LPSs are located near Aβ 1-40/42 in amyloid plaques as well as blood vessels [128], and in AD patients, its levels are slightly higher compared with healthy adults [141]. When microglial cells come in contact with LPS, the TLRs present on the cell membrane of microglia gets activated through interaction

with glycosylphosphatidylinositol-anchored receptor CD14 and MD-2 protein promoting inflammatory responses [110,142]. CD14-activated receptor TLR4 mediates responses to Aβ [143]. This activation affects the immune response and induces neuroinflammation.

#### **5. Neuroinflammation**

Our brain sustains the immune cells that protect against infection and injury, also supporting neurons in plasticity and circuit efficient connectivity. Inflammation is a response necessary for protection and regulation of the process which is associated with managing and reducing damage of the organism: protection against microorganisms, tissue repair, and removal of debris from the body [144]. Various studies currently indicate the involvement of neuroinflammation playing a crucial role in the progression of neuropathological changes that are observed in AD [145] (Figure 2). A broad variety of cellular and molecular mechanisms, assumedly identical in aging and chronic metabolic diseases such as hypertension, diabetes, metabolic syndrome, dementia, depression, or traumatic brain injury, are currently considered silent contributors to neuroinflammation [146]. The key players responsible for induction of neuroinflammation are known to be activated microglia and astrocytes [24,147].

Microglia which originate from myeloids are known as immunocompetent cells in the brain. Microglia cells are considered to be the most important player in the development and progression of neuroinflammation [25]. Microglia are immensely plastic cells that can transform into complex phenotypes depending on specific microenvironmental signals within the brain [148]. On the membrane, these cells express a diverse range of innate immune receptors that belong to the pattern recognition receptors family [147]. When pattern recognition receptors get activated on microglia, activation of the cell and the production of inflammatory mediators occur in the presence of a distinct signaling cascade [149]. Repeatedly activated microglia release a broad range of proinflammatory [150] and toxic products and, among them, reactive oxygen species, nitric oxide, and cytokines. In addition, endothelial cells and perivascular macrophages are also important in interpreting and propagating these inflammatory signals within the CNS [24]. A threat to the CNS, such as invasion, injury, or disease, activates microglia, induces morphological changes, and increases motility of cells.

In AD, there are studies conducted that the primary initiator of activation of microglia is the accumulation of Aβ [151]. The activated microglia respond to Aβ, resulting in migration to the plaques and phagocytosis of Aβ. It initiates a microglial-mediated inflammatory response by binding to various pattern recognition receptors [152], which, in turn, results in cell activation and release of proinflammatory factors (iNOS, TNF-α, IL-1, and IL-6) [152–155]. In the case of AD, the receptors present on the surface of the microglia bind to Aβ oligomers and Aβ fibrils. In the process of phagocytosis, microglia begin to clean up Aβ fibrils; hence, fibrils undergo an endolysosomal pathway.

Other than microglia, astrocytes are also major participants in neuroinflammation [156]. They are fivefold more than neurons in the CNS [157] and are known to have functions in the maintenance of CNS integrity, such as control of blood perfusion in the cerebrum, maintenance of blood–brain barrier stability, and modulation of neuron or nutrient transmission [158]. In AD patient brains, there have been observed alterations in the morphology of astrocytes, their protein composition, gene expression, and function [150]. The accumulation of activated astrocytes is often present in clusters around amyloid plaques. Aβ deposit can activate the astrocytes which lead to overexpression of cytokines, such as IL-1β and IL-6, resulting in oxidative stress [24,159]. It was recently shown that neurodegeneration presumably associates astrocytes, which, by taking on a microglia-induced A1 proinflammatory phenotype, would encourage neuronal cell death, with TNF-α as the most eminent arbitrator [160,161].

On the other hand, the activated microglia lose their phagocytic effect, thus decreasing the degree of Aβ phagocytosis, inevitably developing its accumulation [162]. Moreover, such discoveries are supported by the results of an association between an increase in AD

risk and alterations in genes encoding immune receptors such as TREM2, CD33, and CR1 (myeloid cell surface antigen) [163]. Since they are all expressed on myeloid cells, it is a more convincing demonstration that alterations in microglial biology are linked to AD pathogenesis. Worth mentioning, a variety of transcriptomic and proteomic analysis of inflammatory cells might provide biomarkers for preclinical detection as well as insights on the progression from mild cognitive impairment to AD condition [164–166].

A relatively close connection has also been reported between microglia and cognitive dysfunction [167]. Importantly, in healthy tissue, microglia have a ramified morphology and prolongations that continuously look after the synaptic activity. However, phagocytic microglia have a salient role in synaptic pruning and honing in the developing nervous system [168]. The most fascinating mechanism describing memory dysfunction in AD suggests that Aβ oligomers lead to microglial activation, which, in turn, excessively engulfs and accelerates the termination of synapses through complement factors such as C1q and C3 [169]. It has also been reported that Aβ oligomer arbitrates memory problems which are closely connected with glial activation [100,170].

Recent evidences now shed light on a dangerous dialogue between central immune cells and the gut microbiota, potentially leading to AD in humans.

#### **6. The Link between Microbiota and Neuroinflammation**

The immune system modulates the gut microbiota framework and issuance [171], while in return, the microbial symbionts control immune system maturation and function [172,173]. Numerous rodent studies have affirmed that there is an interaction between the gut microbiota and various immune cell populations [174,175] or the expression of genes related to neuroinflammation [176,177].

The study furnished evidence stating that microbiota residing in the gut predisposes the development of the immune system by administering hematopoiesis of primary immune cells. It was shown that germ-free (GF) mice have a lower ratio and less distinction capability of myeloid cell progenitors of both yolk sac and bone marrow origin. This supports the idea of the widespread effects of gut microbiota on the immune system, microglia included [175]. Microglia from antibiotic-treated mice or GF mice showed an immature profile and impaired immune response. The absence of gut microbiota alters microglial mRNA profiles and suppresses various microglial genes involved in cell activation, pathogen recognition, and host defense. Microglia transcription and survival factors, normally suppressed in mature adult microglia, were increased in GF mice [178]. The experiment was conducted to examine the transcriptional profiles of different microglial development stages, referring to the genes related to the adult phase of microglial maturation and immune response that are abnormally regulated in GF mice [179].

A number of studies have coined a protective association between dietary polyphenols and the prevention of age-related chronic diseases such as diabetes, cancer, and neurodegenerative diseases [180–182]. Dietary flavonoids and nonsteroidal anti-inflammatory agents modulate the nuclear factor-kappa β signaling pathway and therefore are termed as a potential therapeutic target for AD [182–184]. Polyphenols make an impact on microbiotarelated metabolism and have a potential to improve neurological health, including their ability to interact with intracellular neuronal and glial signaling, to modulate peripheral and cerebrovascular blood flow, and to reduce neuronal damage and loss induced by neurotoxins and neuroinflammation [185–187]. Flavonoids, a subclass of polyphenols, are more likely to combat neuronal dysfunction and toxicity by recruiting antiapoptotic pro-survival signaling pathways, increasing antioxidant gene expression and reducing Aβ pathology [182,188,189]. Flavonoids that are not absorbed in the small intestine and other sugars are then broken down by the gut microbiota into phenolic acids and other metabolites that inhibit the growth of *Ruminococcus gauvreauii*, *Bacteroides galacturonicus*, and *Lactobacillus* sp. strains [190] and flavonoids present in berries have also shown inhibitory actions against *Bacillus cereus*, *Campylobacter jejuni*, *Clostridium perfingens*, *Helicobacter pylori*, *Staphylococcus aureus*, *Staphylococcus epidermidis*, and *Candida albicans* [191]. Recently, it was reported that anthocyanins (one of the flavonoids) could significantly ameliorate the expression of proinflammatory cytokines and ROS/JNK, thus preventing neuroinflammation and AD pathology [192–194]. In an experiment conducted on aged rodents, blueberry supplementations have shown improved spatial memory, object recognition memory, and inhibitory fear conditioning learning [195–197]. In another study on blueberry anthocyanins given to adults aged 40–74 years over 3 weeks, plasma concentrations of NF-kB-related proinflammatory cytokines and chemokines (IL-4, IL-13, IL-8, and IFN-α) were significantly reduced [198]. However, a study conducted by Spilsbury et al. did not reveal any remarkable effect of lower concentrations of flavonoids on NF-κB activity in astrocytes [199]. Nevertheless, the literature date supports that the dietary supplementation of flavonoids might be implicated in the regulation of NF-κB in neurons [199].

Flavonoids are important players in the prevention of neuroinflammation via several anti-inflammatory mechanisms, inhibiting the microglial activation of inflammatory cytokines (TNF-α and IL-1β), inhibiting iNOS and ROS generation in activated glia, and downregulating the activity of pro- inflammatory transcription factors such as NF-κB through modulation of glial and neuronal signaling pathways [182].

Chicory root, known for its high content of fibers (galacto-oligosaccharides and fructans, such as inulin) and beneficial for the MGB axis modulation [64,177,200], recently also has received attention due to its sesquiterpene lactones (a class of sesquiterpenoids that contain a lactone ring) [201]. Interestingly, it has been shown that different sesquiterpene lactones from chicory root have the potential to influence anti-inflammatory responses through modulation of the nuclear factor of the activated T-cells pathway [201].

Bacterial metabolites such as SCFAs were considered the key mediators for microbiota– microglia interaction. These compounds have the potential to translocate from the mucosa to systemic circulation and to cross the blood–brain barrier affecting the CNS and their function [68,202]. Oral administration of SCFA for 4 weeks restored many facets of the immature microglial morphology of GF mice. SCFA claimed to reestablish microglial density and normalized CSF1R surface expression [203]. It is crucial to accentuate that the gut microbiota–microglia interaction is extremely dynamic as many of the defects noticed in the microglia of GF mice could be partially restored by recolonization with conventional gut microbiota or SCFA supplementation [203].

#### **7. Role of Antibiotics on Microbiota in AD**

Antibiotics or antimicrobial substances are typically used to remove or prevent bacterial colonization in the human body [204]. These can alter the bacteria without any specific target or type [205]. As a consequence, a broad spectrum of antibiotics can immensely affect the composition of the gut microbiota, lower its biodiversity, and withhold colonization for a long period after administration. Various studies with distinct antibiotic treatments resulted in long-/or short-term changes in the gut microbiota in both animals as well as humans [206]. Numerous studies have demonstrated that the use of antibiotics has an association with changes in behavior and brain chemistry [207–209]. Studies conducted in vivo with long-term broad spectrum antibiotic treatment have shown a decreased Aβ plaque deposition, attenuation of plaque localization in glial reactivity, and alteration in microglial morphology in the APPSWE/PS1<sup>Δ</sup>E9 mouse model of AD [210]. Another study conducted on 68 patients with advanced AD demonstrated a correlation among usage of antibiotics and prolonged survival. Of the patients who survived for more than 6 months, 31% were on antibiotic care and 14% were on palliative care [211]. Another study in humans showed that antibiotics, i.e., cefepime, can cross the blood–brain barrier, causing altered mental status, along altered consciousness and confusion without mediation of the gut microbiota [212]. Below, some of the preclinical studies of antibiotics in animals and humans have been described briefly.

The patients suffering from infection caused by *Helicobacter pylori* were administered with a cocktail of antibiotics consisting of proton pump inhibitor and clarithromycin, along with amoxicillin or metronidazole. The outcome of this treatment showed an association with neurological disorders, including panic attacks due to major depression and anxiety, delirium, and psychosis [213]. On the other hand, the elimination of pathogenic bacteria such as *Helicobacter pylori* in AD patients by the triple eradication antibiotic regimen (clarithromycin, amoxicillin, and omeprazole) led to positive results for cognitive and functional status parameters [214].

Antibiotic administration with rifampicin and minocycline in AD animal models reduced the Aβ levels in the brain and abbreviates inflammation cytokines [215]. Oral administration of rifampicin to three different mouse models of Alzheimer's disease and tauopathy showed that this antibiotic reduced the accumulation of Aβ oligomers and tau oligomers and enhanced the memory of the mice. These results suggested that rifampicin could prevent AD [216]. In 6 months, AD patients' improvement in the Standardized AD Assessment Scale cognitive subscale was observed when treated with a combination of doxycycline and rifampicin [217].

A pilot study conducted on the TgCRND8 transgenic mouse model showed that 3 months of treatment with erythromycin in drinking water at 0.1 g/L reduced the Aβ1-42 levels in the cortex by 54% when compared to vehicle-treated mice [218].

Several studies conducted on minocycline suggested that it has neuroprotective and anti-inflammatory actions in many animal models. In microglial cell cultures, it was remarkably able to reduce the oligomeric Aβ-induced neuroinflammatory response and enhancement of fibrillar Aβ phagocytosis [219]. Minocycline treatment at 50 mg/kg for 4 weeks in a transgenic hAPP mouse model of AD exhibited attenuated behavioral abnormalities, neuroinflammatory markers, and Aβ [220]. In another study, 4 months of treatment with minocycline at 55 mg/kg/day in food in 3×Tg-AD mice showed a reduction in brain levels of insoluble Aβ, decreased neuroinflammatory markers, and reversed cognitive deficit [221].

A contrary effect of antibiotics was also observed after administration of ampicillin in the Sprague–Dawley rats. In this case, an elevated level of corticosterone in serum, intensified anxiety-like behavior, impairment of memory due to elevated glucocorticoids, and reduction in hippocampal brain-derived neurotrophic factor were determined [222]. Distinct studies demonstrated that administration of intracerebroventricular streptozotocin into the brain of wild-type mice and rats can cause learning impairment and memory loss [223–227].

An experiment conducted on APPSWE/PS1<sup>Δ</sup>E9 transgenic mice administered with antibiotics demonstrated that it led to an alteration in several circulating inflammatory cytokines and chemokines in the blood. It also showed an elevated level of CCL11 (which has a link to age-related deficits in hippocampal neurogenesis) [228] in the blood serum of mice [210]. A recent study conducted on APPSWE/PS1L166P mice treated with a cocktail of antibiotics revealed a selective, microbiome-dependent, sex-specific effect on brain Aβ amyloidosis of Aβ and microglial physiology [229]. Interestingly, the transplants of fecal microbiota from age-matched APPSWE/PS1L166P mice into antibiotic-treated APPSWE/PS1L166P mice restores the gut microbiota and partially restores AD pathology along with microglial morphology [229].

#### **8. Role of Probiotics on Microbiota in AD**

Probiotics are defined as living microbial feed supplements which show a beneficial effect on the host, resulting in improved intestinal microbial balance [230]. The most commonly used probiotics are lactic acid bacteria, particularly *Lactobacilli*, *Streptococci*, *Pediococcus*, *Enterococcus*, and *Bifidobacteria* and some yeast like *Saccharomyces boulardii*. However, not all microorganisms can be probiotic, as they need to be strain-specific (Table 4).


**Table 4.** Effects of probiotics on neurological disorders.

A broad range of probiotics have been used in an animal study and in the models of AD. In rats, *Bifidobacterium* and *Lactobacillus* administration have shown a positive effect on AD treatment [235]. In an AD mouse model, *Bifidobacterium breve* strain A1 prevented cognitive function, making it one of the effective treatments for AD [237]. A reduction in neuroinflammation in mouse models due to *Lactobacillus casei* strain *Shirota* can be effective against AD [234]. Despite the fact that there are few human clinical studies compared to animals, there is increasing indication that probiotics can be used for reducing depression and anxiety-like symptoms [241].

A study with thirty-six healthy women assigned to three groups showed the importance of probiotics in the modulation of brain activity [242]. In this experiment, the group which was treated with fermented milk products containing *Bifidobacterium animalis sub. lactis*, *Streptococcus thermophilus*, *Lactobacillus bulgarigaricus*, and *Lactococcus lactis subs. lactis* showed a compelling reduction in the activity of the specific area in the brain. This region of the brain is involved in sensory/affective tasks when compared to the activation of other cortical regulatory brain areas. The experiment confirmed that probiotic supplementation has a major contribution in activating specific areas in the brain involved in the central control of emotion and sensation [242].

In another study conducted to understand the probiotic application in AD, sixty patients with AD were randomly assigned into two groups [243]. The first group received 200 mL/day milk enriched with *Lactobacillus acidophilus*, *Lactobacillus casei*, *Bifidobacterium bifidum,* and *Lactobacillus fermentum* for weeks, whereas the control group received plain milk of the same amount. The subjects, which were on probiotic supplementation showed a significant improvement in the mini-mental state examination test when compared with controls. The study revealed a beneficial effect on cognitive function and metabolic status of patients with AD. However, the treatment with probiotics was ineffective on oxidative stress and inflammation [243].

A study conducted by Leblhuber et al. showed an increased level of serum kynurenine, which was observed after probiotic administration, potentially caused by macrophage activation. The stimulation of immune cells could induce mechanisms that can be helpful in removing amyloid aggregates and damaged cells or on the other perspective. On the other hand, the intensive activating events could negatively affect gut barrier function and further stimulate neurodegenerative events [244].

When taken together, these human and animal studies prove that probiotics can have a major role in the bidirectional communication between the gut microbiota and the brain, modulating brain function. The exact mechanism of probiotics on the MGB axis is not yet well defined. Therefore, the data suggest that the proper dose of probiotics in AD treatment would be a new way to eliminate amyloid deposition in the brain by the MGB axis and to reduce neuroinflammation (Figure 2).

#### **9. Conclusions**

Accumulating all information from the human as well as animal studies, it can be suggested that GIT microbiota has an important role in the bidirectional communication between the brain and the gut. There is increasing evidence stating that the gut microbiota has a contribution to the pathogenesis of AD. As the gut microbiota is known as the source of a large number of amyloid, LPS, and other toxins, it can contribute to systemic inflammation and disruption of physiological barriers. The products formed by bacteria can move from the GIT to the CNS, especially in aging. Bacterial amyloid can trigger misfolding and can enhance native amyloid aggregation. The gut microbiota products can activate microglia, augmenting inflammatory response in the CNS, which in turn results in microglial function. Triggered microglia start neuroinflammation in the brain, causing loss of neurons, a major factor in AD. Modulation of the gut microbiota composition can be used as a therapeutic target in AD. Some antibiotics as well as probiotics can be implemented as a preventive measure that successfully targets ongoing inflammation. The role of antibiotics and probiotics in modulating the microbiota is under intense debate. The certain microbiota profile also strongly depends on the host's genetics and diet. This only confirms that research on MGB involvement in AD is crucial for new treatment targets and therapies for AD.

**Author Contributions:** Conceptualization, A.M. and A.B.; investigation, A.M.; writing—original draft preparation, A.M., D.B., and A.B.; writing—review and editing, A.M., A.B., and V.B.; visualization, A.M.; supervision, A.B.; project administration, A.B.; All authors have read and agreed to the published version of the manuscript.

**Funding:** This project received funding from the European Regional Development Fund (project No. 01.2.2-LMT-K-718-02-0014) under grant agreement with the Research Council of Lithuania (LMTLT).

**Conflicts of Interest:** No conflict of interest declared.

#### **Abbreviations**


#### **References**


## *Article* **Application of** *Ligilactobacillus salivarius* **CECT5713 to Achieve Term Pregnancies in Women with Repetitive Abortion or Infertility of Unknown Origin by Microbiological and Immunological Modulation of the Vaginal Ecosystem**

**Leónides Fernández 1,\*, Irma Castro 2, Rebeca Arroyo 2, Claudio Alba 2, David Beltrán <sup>3</sup> and Juan M. Rodríguez 2,\***


**Abstract:** In this study, the cervicovaginal environment of women with reproductive failure (repetitive abortion, infertility of unknown origin) was assessed and compared to that of healthy fertile women. Subsequently, the ability of *Ligilactobacillus salivarius* CECT5713 to increase pregnancy rates in women with reproductive failure was evaluated. Vaginal pH and Nugent score were higher in women with reproductive failure than in fertile women. The opposite was observed regarding the immune factors TGF-β 1, TFG-β 2, and VEFG. Lactobacilli were detected at a higher frequency and concentration in fertile women than in women with repetitive abortion or infertility. The metataxonomic study revealed that vaginal samples from fertile women were characterized by the high abundance of *Lactobacillus* sequences, while DNA from this genus was practically absent in one third of samples from women with reproductive failure. Daily oral administration of *L. salivarius* CECT5713 (~9 log10 CFU/day) to women with reproductive failure for a maximum of 6 months resulted in an overall successful pregnancy rate of 56%. The probiotic intervention modified key microbiological, biochemical, and immunological parameters in women who got pregnant. In conclusion, *L. salivarius* CECT5713 has proved to be a good candidate to improve reproductive success in women with reproductive failure.

**Keywords:** infertility; repetitive abortion; implantation failure; *Lactobacillus salivarius*; probiotics; vaginal microbiome; TGF-β; VEGF

#### **1. Introduction**

Increasing evidence has highlighted the relevance of the microbiota of the female genital tract for human reproduction [1,2]. Under physiological conditions, and in contrast to the gut, the human vaginal microbiota is usually characterized by a low microbial diversity and the dominance of bacteria from the genus *Lactobacillus* [3,4]. In fact, a low diversity in the gut has been linked to a variety of gastrointestinal processes, including inflammatory bowel disease [5], while a high diversity in the vagina has been associated to vaginosis [6].

The vaginal microbiota in healthy reproductive-age women is mainly composed of one or a few *Lactobacillus* species, which represent more than 90% of the total microbiota [7,8]. In a seminal study, the bacterial communities of 396 asymptomatic women were classified into five distinct vaginotypes; four of them were dominated by *Lactobacillus crispatus*, *Lactobacillus gasseri*, *Lactobacillus iners*, and *Lactobacillus jensenii*, respectively; in contrast, the fifth one had lower proportions of lactobacilli and was predominantly composed of

**Citation:** Fernández, L.; Castro, I.; Arroyo, R.; Alba, C.; Beltrán, D.; Rodríguez, J.M. Application of *Ligilactobacillus salivarius* CECT5713 to Achieve Term Pregnancies in Women with Repetitive Abortion or Infertility of Unknown Origin by Microbiological and Immunological Modulation of the Vaginal Ecosystem. *Nutrients* **2021**, *13*, 162. https://doi. org/10.3390/nu13010162

Received: 27 November 2020 Accepted: 30 December 2020 Published: 6 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

strictly anaerobic bacterial genera, such as *Gardnerella*, *Prevotella*, *Megasphaera*, *Atopobium,* or *Dialister* [3]. This last vaginotype was associated to high Nugent scores, a Gram-staining based technique used for the diagnosis of bacterial vaginosis (BV).

Several factors are known to contribute to interindividual and intraindividual changes in the vaginal microbiota [9]. Although shifts between different vaginotypes may occur naturally, increase of diversity and colonization by strict anaerobes and decrease or depletion of lactobacilli are considered as risk factors for BV. In fact, vaginal microbiota dysbiosis has been associated with higher rates of intra-amniotic infection, premature delivery, spontaneous abortion, and infertility [10–15].

Different studies have shown that infertile women harbor a differential vaginal microbiota when compared to fertile women [16–19]. Therefore, the composition of the vaginal microbiota (and, particularly, any deviation from the *Lactobacillus*-dominated, low-diversity vaginal microbiome) may play a key role in fertility and in the outcomes of assisted reproduction treatments (ARTs) [20–22]. Abundant isolation of enterococci, streptococci, staphylococci, and/or Gram-negative bacteria (*Escherichia coli*, *Klebsiella pneumoniae*) from the tip of the catheter used for embryo transfer has been correlated with lower implantation and pregnancy rates and increased miscarriage rates [23], while abundant isolation of lactobacilli and low density or no isolation of the aforementioned bacteria has been correlated with better reproductive outcomes [24–28]. Metataxonomic studies of endometrial samples have also revealed that an abnormal endometrial bacterial profile (with a low percentage of sequences of the genus *Lactobacillus*) is a common feature in a high percentage of infertile women subjected to ART [21,29]. Although at least a part of the bacterial DNA detected in endometrial samples may arise from vaginal contamination during sampling, these studies suggest that an abundant presence of *Lactobacillus* DNA in such samples may be a predictor of implantation success [29,30].

As a consequence, the assessment of the microbial communities in the reproductive tract should be considered as a relevant part of the evaluation and personalized care in cases of reproductive failure of unknown cause or origin. When this happens, the use of probiotics may be a possible strategy to modulate the reproductive tract microbiome and to increase the success rates [31]. However, such a combined strategy (assessment of vaginal communities together with use of a target-selected probiotic) has not been explored yet, and commercially available probiotics are being empirically prescribed for repopulation of the female reproductive tract with *Lactobacillus* strains [2], without a proper scientific evidence of their actual usefulness.

Lactobacilli may have different biological activities that contribute to fertility and to a healthy pregnancy, including, among others: (a) the inhibition of the colonization and growth of potentially harmful microbes, including viruses, bacteria, yeast, and protozoa that may compromise fertility [32,33]; (b) contribution to angiogenesis and vasculogenesis that may favor the implantation of the embryo [34]; and, (c) induction of immunomodulation activities, such as those involved in implantation and in tolerance towards the embryo, first, and the fetus, later [35,36]. However, those properties might be strain-specific and, therefore, a strain-by-strain evaluation has to be performed for this specific target.

*Lactobacillus salivarius* CECT5713 [37] has been shown to be a probiotic strain suitable for applications in the mother–infant dyad due to a wide repertoire of desirable phenotypic and genotypic properties [38]. This includes a high survival rate when exposed to gastrointestinal tract conditions, a high acidifying activity, and antimicrobial, anti-inflammatory, and immunomodulatory properties, which have been demonstrated *in vitro*, in animal models, and in human clinical trials [38–44]. Therefore, and after evaluating some vaginalrelated properties in this study, it was selected to be administered in a clinical trial in order to assess its efficacy for the infertility target. It must be highlighted that this species has been renamed as *Ligilactobacillus salivarius* in the recent proposal for reclassification of the genus *Lactobacillus* [45].

In this context, the first objective of this study was to assess the differences in several vaginal parameters (pH, Nugent score, microbiota composition as determined through culture and metataxonomic methods, and soluble immune factor levels) between women with reproductive failure (because of repetitive abortion during the first 12 weeks of pregnancy or infertility of unknown origin) and fertile women. The second objective was to evaluate the ability of *L. salivarius* CECT5713 to modulate those vaginal parameters and to increase pregnancy rates (currently ~29% after IVF procedures in this setting) in the group of women with reproductive failure.

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

#### *2.1. Characterization of Vaginal-Related Properties of L. salivarius CECT5713*

An overlay method [46] was used to determine the ability of *L. salivarius* CECT5713 to inhibit the growth of various species of bacteria and yeasts. It was performed as described previously [37]. All indicator strains had been previously isolated from clinical cases of vaginal or cervical infections, and included five strains of *G. vaginalis*, three of *Streptococcus agalactiae* and of *Candida albicans*, and two of *Candida glabrata*, *Candida parapsilosis,* and *Ureaplasma urealyticum* (our own culture collection). All inhibitory activity assays were performed in triplicate.

The ability of *L. salivarius* CECT5713 to aggregate with cells of the indicator strains cited above was investigated following the procedure of Younes et al. [47]. The suspensions were observed under a phase-contrast microscope. Adherence to vaginal epithelial cells collected from healthy premenopausal women was performed and interpreted as described previously [48]. Adherence was measured as the number of lactobacilli adhered to the vaginal cells in 20 random microscopic fields. *L. salivarius* CECT9145 was used as a control strain because of its high adherence to vaginal cells [49]. The assay was performed in triplicate.

Initially, the α-amylase activity of *L. salivarius* CECT5713 was qualitatively assessed using the method described by Padmavathi et al. [50]. Briefly, the strain was inoculated into a modified MRS media containing starch (0.5% peptone, 0.7% yeast extract, 0.2% NaCl, 2% starch, and 1.5% agar). The plates were incubated at 37 ◦C for 48 h in anaerobiosis and, then, the zone of clearance was observed by adding Gram's iodine as detecting agent. Quantitation of the cell-bound α-amylase activity of *L. salivarius* CECT5713 was done with a kit (Kikkoman Co., Tokyo, Japan) using 2-chloro-4-nitrophenyl 65-azido-65-deoxy-βmaltopentaoside as substrate and using conditions described previously [51]. One unit of activity was defined as the amount of enzyme needed to release 1 μmol 2-chloro-4 nitrophenol from 2-chloro-4-nitrophenyl 65-azido-65-deoxy-β-maltopentaoside per min at 37 ◦C.

#### *2.2. Participants, Sampling, and Design of the Human Study*

A total of 58 women, aged 28–45, participated in this study (Table 1). Volunteers were classified into 3 groups. All women in the RA group (*n* = 21) had a history of recurrent miscarriage with three or more pregnancy losses during the first 12 weeks of pregnancy. All women of the INF group (*n* = 23) had a history of infertility (inability to conceive) despite being the recipients of ART for at least three times, including two cycles, at least, of in vitro fertilization (IVF). Finally, the control group (*n* = 14) included fertile women having at least two children after uncomplicated term pregnancies. None of the women of the RA and INF groups received ART during the whole period of the study. None of the RA group components were diagnosed of antiphospholipidic syndrome and, therefore, they did not receive either heparin and/or salicylic acid during the study. None of the participants had received hormonal therapy, antibiotics or probiotics in the 4 weeks previous to sampling. Vaginal samples were taken at least 7 days after coitus to avoid or minimize the impact of the partner's semen on the vaginal pH, microbiota composition or immunoprofile (in the latter case, particularly in relation to the concentration of the two isoforms of the transforming growth factors beta 1 and 2 (TGF-β 1 and TGF-β 2)). Women with lactose intolerance or cow's milk protein allergy were excluded because of the excipient used

to administer the strain in the subsequent pilot trial (see below). Informed consent was obtained from all subjects involved in the study.

**Table 1.** Characteristics of the participants (*N* = 58) which included fertile women (Control group), women with a history of repetitive abortion (RA group), and women with infertility of unknown origin (INF group).


<sup>1</sup> Antibiotic usage means ≥4 annual treatments due to recurrent infections. <sup>2</sup> One-way ANOVA tests were used to evaluate differences in mean values of women age, weight, and height and duration of the menstrual cycle between groups. Values followed by different superscript letters within the same row indicate statistically significant differences between groups according to Scheffé post hoc comparison tests. <sup>3</sup> Freeman–Halton extension of the Fisher exact probability tests for a 2 × 3 contingency table were used to compute the (two-tailed) probability of obtaining a distribution of values of categorical variables (regularity of the menstrual cycle, history of infections, antibiotic usage and history of other conditions).

> At recruitment (within the first three days post-ovulation; day 0), two samples were collected: A vaginal swab specimen for in fresh determination of the Nugent score, and a cervicovaginal lavage (CVL) of the cervical and the vaginal walls with 10 mL of sterile normal saline for all the other analysis. Aliquots of the CVL samples were used for culturebased analysis. Subsequently, CVL samples were clarified by centrifugation at 800× *g* for 10 min at 4 ◦C. Aliquots of CVL supernatants and cell pellets were stored at −80 ◦C until the immunological and metataxonomic analyses were performed. Demographic, anthropometric, and health data (including a past or present history of recurrent infections at different body locations and use of antibiotics) were recorded at recruitment (Supplementary Figure S1). High use of antibiotics was defined as receiving ≥4 antibiotic treatments per year because of recurrent infections while a range between 0 and 2 annual treatments was considered as a low use of antibiotics.

> Starting at day 0, women of the RA and INF groups consumed (oral route) a daily sachet with ~50 mg of freeze-dried probiotic (~9 log10 CFU of *L. salivarius* CECT5713) for 6 months or until a diagnosis of pregnancy (whatever happened first). At that point,

the same two samples described above were collected from each woman. After a diagnosis of pregnancy, oral administration of the probiotic strain was maintained until the 15th week of pregnancy. All the spontaneous pregnancies that occurred within the first year after day 0 were recorded in this study.

Probiotic-containing sachets were kept at 4–8 ◦C throughout the study. All volunteers signed a written consent and were provided with diaries to record compliance with the study product intake. Minimum compliance rate (% of the total treatment doses) was set at 86%. This study was conducted according to the guidelines laid down in the Declaration of Helsinki and it was approved by the Ethical Committee of Biomedical Research of Consejería de Salud y Familias (Junta de Andalucía, Granada, Spain) (P050/19, Act 11/19). The study was registered in the ClinicalTrials.gov database (NCT04446572).

#### *2.3. Measurement of Vaginal pH and Nugent Score*

At each of the two study visits, the pH of the lateral vaginal wall was measured (Whatman pH paper, pH 3.8–5.5 and pH 6.0–8.1). Nugent scoring was performed as described previously [52]. Briefly, the swab material was transferred to a glass slide, heat fixed, and Gram stained. Gram-positive, Gram-negative, and Gram-variable bacterial morphotypes were quantified. A Nugent score of 0–3 was considered normal, 4–6 was considered intermediate, and 7–10 was considered consistent with bacterial vaginosis [52].

#### *2.4. Culture-Dependent Analysis*

CVL samples collected during the trial were serially diluted and plated onto Columbia Nalidixic Acid (CNA), Gardnerella (GAR), CHROMagar StrepB (CHR), Mac Conkey (MCK), Mycoplasma (MYC), and Sabouraud Dextrose Chloramphenicol (SDC) agar plates (BioMerieux, Marcy l'Etoile, France) for selective isolation and quantification of the main cultivable non-*Lactobacillus* bacteria and yeasts that may be found in the vagina, including the agents most frequently involved in vaginal infections. They were also inoculated onto agar plates of MRS (Oxoid, Basingstoke, UK) supplemented with either L-cysteine (2.5 g/L) (MRS-C) or horse blood (5%) (MRS-B) for isolation of lactobacilli, including *L. iners* (MRS-B). All media were incubated for 48 h at 37 ◦C under aerobic conditions, with the exception of the MRS-C and MRS-B plates, which were incubated anaerobically (85% nitrogen, 10% hydrogen, 5% carbon dioxide) in an anaerobic workstation (DW Scientific, Shipley, UK) for up to 72 h. After incubation, the colonies were recorded and at least one representative of each colony morphology was selected from the agar plates. The isolates were identified by Matrix Assisted Laser Desorption Ionization-Time of Flight (MALDI-TOF) mass spectrometry (Bruker, Bremen, Germany). When the identification by MALDI-TOF was not possible at the species level (particularly in the case of lactobacilli isolates), the identification was carried out by 16S ribosomal RNA (rRNA) gene sequencing as described by Mediano et al. [53].

#### *2.5. DNA Extraction from the Samples*

Approximately 1 mL of each CVL sample was used for DNA extraction following a method described previously [54]. Extracted DNA was eluted in 22 μL of nucleasefree water and stored at −20 ◦C until further analysis. Purity and concentration of each extracted DNA was initially estimated using a NanoDrop 1000 spectrophotometer (NanoDrop Technologies, Inc., Rockland, DE, USA). Negative controls (blanks) were processed in parallel.

#### *2.6. Real-Time Quantitative PCR (qPCR) Assay for the Detection and Quantification of L. salivarius DNA*

Primers and conditions for quantification of *L. salivarius* DNA have been described previously [55]. The DNA concentration of all samples was adjusted to 5 ng μL<sup>−</sup>1. A commercial real-time PCR thermocycler (CFX96™, Biorad Laboratories, Hercules, CA, USA) was used for all experiments. Standard curves using 1:10 DNA dilutions (ranging from 2 ng to 0.2 pg) from *L. salivarius* CECT5713 were used to calculate the concentrations of the

unknown bacterial genomic targets. Threshold cycle (Ct) values between 15.29 and 20.07 were obtained for this range of *L. salivarius* DNA (R2 = 0.9915). The Ct values measured for DNA extracted from non-target species (*L. reuteri* MP07 and *Lactobacillus plantarum* MP02; our own collection) were ≥39.27 ± 0.64. These two control strains were selected because they belong to the *L. salivarius* taxonomically closest species [56]. All samples and standards were run in triplicate.

#### *2.7. Metataxonomic Analysis*

The V3-V4 hypervariable region of the 16S rDNA was amplified by PCR using the universal primers S-D-Bact-0341-b-S-17 (CCTACGGGNGGCWGCAG) and S-D-Bact-129 0785-a-A-21 (GACTACHVGGGTATCTAATCC) [57] and sequenced in the MiSeq system of Illumina at the facilities of Parque Científico de Madrid (Tres Cantos, Spain). Barcodes appended to 3 and 5 terminal ends of the PCR amplicons allowed separation of forward and reverse sequences in a second PCR-reaction. DNA concentration of the PCR products was quantified in a 2100 Bioanalyzer system (Agilent, Santa Clara, CA, USA). After pooling the PCR products at about equal molar ratios, DNA amplicons were purified by using a QIAEX II Gel Extraction Kit (Qiagen, Hilden, Germany) from the excised band having the correct size after running on an agarose gel. DNA concentration was then quantified with PicoGreen (BMG Labtech, Jena, Germany). The pooled, purified and barcoded DNA amplicons were sequenced using the Illumina MiSeq pair-end protocol (Illumina Inc., San Diego, CA, USA) following the manufacturer's protocols.

#### *2.8. Bioinformatic Analysis*

Raw sequence data were demultiplexed and quality filtered using Illumina MiSeq Reporter analysis software. Microbiome bioinformatics was done with QIIME 2 2019.1 [58]. Denoising was performed with DADA2 [59]. Taxonomy was assigned to ASVs using the q2-feature-classifier [60] and the naïve Bayes classifier *classify-sklearn* against the SILVA database version 132 [61]. Posterior bioinformatic analysis was conducted using the R version 3.5.1 (https://www.R-project.org) [62]. A table of Operational Taxonomic Units (OTUs) counts per sample was generated, and bacterial taxa abundances were normalized to the total number of sequences in each sample. The relative abundance values of the different bacterial taxa in the three groups of CVL samples (control, RA and INF) were analyzed using the linear discriminant analysis (LDA) effect size (LEfSe) algorithm [63] in an online version (http://huttenhower.sph.harvard.edu/galaxy/). Alpha diversity was studied with the Shannon and Simpson diversity indexes with the R Vegan package (Version 2.5.6) (https://github.com/vegandevs/vegan/). Beta diversity was studied using principal coordinates analysis (PCoA) to visually display patterns of bacterial profiles at the genus level through a distance matrix containing a dissimilarity value for each pairwise sample comparison. The Bray–Curtis and binary Jaccard indices were used for quantitative (relative abundance) and qualitative analyses (presence/absence), respectively. Analysis of variance of the distance matrices was performed with the "nonparametric *MANOVA* test" Adonis with 999 or permutational multivariate ANOVA (PERMANOVA) with 999 permutations with the R Vegan package. The heatmap graph was generated by using *gplots* package. Dendogram linkages were based on the relative abundance of the 20 most abundant bacterial genera within the samples and on the complete linkage method for hierarchical clustering (*hclust* function).

#### *2.9. Immunological Analysis*

The concentrations of several soluble immune factors (IL1β, IL1ra, IL2, IL4, IL5, IL6, IL7, IL8, IL9, IL10, IL12, IL13, IL15, IL17, IL6, basic FGF, eotaxin, GCSF, GMCSF, IFNγ, MCP1, MIP1α, MIP1β, PDGF-BB, RANTES, TNFα, VEGF) were determined by magnetic bead-based multiplex immunoassays, using a Bioplex 200 instrument (Bio-Rad, Hercules, CA, USA) and the Bio-Plex Pro™ Human Cytokine 27-plex Assay (ref. M500KCAF0Y, Bio-Rad). In parallel, the levels of TGF-β 1 and TGF-β 2 were measured by ELISA with

the RayBio® Human TGF-β 1 and Human TGF-β 2 ELISA kits, respectively (RayBiotech, Norcross, GA, USA). All determinations were carried out following the manufacturer's protocols and standard curves were performed for each analyte.

#### *2.10. Statistical Analysis*

Microbiological data were recorded as CFU/mL and transformed to logarithmic values before statistical analysis. The normality of data distribution was analyzed using the Shapiro–Wilks test. Then, the quantitative variables were expressed as means and 95% confidence intervals (CI) or standard deviations (SD) when normally distributed and as medians and interquartile ranges (IQR) if they did not follow a normal distribution. The qualitative values were presented as total number of events and percentages. One-way ANOVA tests were used to compare the means of the experimental groups and Scheffé post hoc tests were used to identify which pairs of means were statistically different. The effect of the probiotic intervention on several vaginal parameters in each group of women with reproductive failure was analyzed using one-way ANOVA repeated measures tests. The Fisher's exact probability test, or the Freeman–Halton extension of the Fisher exact probability test for a 2 × 3 contingency table, was used for comparison of proportions and frequencies. For non-parametric analyses, differences between groups were assessed using Kruskal–Wallis tests and Wilcoxon–Mann–Whitney tests to identify which pair of groups were different, with Bonferroni correction for multiple comparisons when indicated. Correlations between the 20-major relative abundant bacterial genera were visualized using R package *qgraph* [64]. Statistical analysis and plotting were performed either using Statgraphics Centurion XVIII version 18.1.06 (Statgraphics Technologies, Inc., The Plains, VA, USA) or in the R environment (version 3.5.1; R-project, http://www.r-project.org) and *ggplot2* [Wickham, 2016]. Differences were considered statistically significant at *p* < 0.05.

#### **3. Results**

#### *3.1. Characterization of Vaginal-Relevant Properties of L. salivarius CECT5713*

*L. salivarius* CECT5713 showed inhibitory antimicrobial activity (inhibition zone > 2 mm around the streak) against all the *G. vaginalis*, *S. agalactiae, C. albicans*, *C. glabrata*, *C. parapsilosis,* and *U. urealyticum* strains used as indicators in this study. The strain was able to form large, well defined co-aggregates with all the selected vaginal and cervical pathogens. Co-aggregation was particularly intense with *G. vaginalis*, *S. agalactiae,* and *C. albicans* strains. In this study, the strain tested was strongly adhesive to vaginal epithelial cells, a mean (±SD) of 329 (±46) adherent lactobacilli in 20 random microscopic fields. The mean (±SD) value for *L. salivarius* CECT9145, a control strain with a high adherence to vaginal cells, was 336 (±52) adherent lactobacilli in 20 microscopic fields. Extracellular amylase production by *L. salivarius* CECT5713 was observed by the zone of clearance around the colonies (~2.0 mm) when flooded with iodine solution. Later, when the α-amylase activity was measured, this strain showed a high level of α-amylase activity (0.83 U/mL) at 16 h (concentration of *L. salivarius* CECT5713: ~8.6 log10 CFU/mL), and could be detected in supernatants at a similar level for up to 48 h (when the assay was finished).

#### *3.2. Demographic, Anthropometric, and Clinical Characteristics of the Participants in the Human Study*

The characteristics of the 58 women that participated in this study are presented in Table 1. The mean (95% CI) age in the control group was 34.6 years (33.5–35.8), while in those of repetitive abortions (RA) and with infertility of unknown origin (INF) was 39.4 (38.5–40.4) and 38.0 (37.1–38.9) years, respectively (Table 1). Women in the control group were significantly younger than other participants (*p* < 0.001; one-way ANOVA), but there were no differences in mean values of body weight and height between the three groups of women (Table 1).

About 71% of the women in the control group had a regular menstrual cycle, while in the other two (RA and INF) this percentage was 48%, although this difference was not statistically significant (*p* = 0.337; Fisher exact probability tests). No differences were observed in the mean duration of the menstrual cycle that was 28, 27.4, and 27.5 days for women in the control, RA, and INF groups, respectively (Table 1).

Interestingly, statistically significant differences were found between the control women and those in the other two groups regarding a history of recurrent vaginal and urinary tract infections (*p* = 0.017 and *p* = 0.006, respectively; Fisher exact probability tests) and the use of antibiotics both during infancy (*p* < 0.001) and adulthood (*p* = 0.003), which were higher in the last two groups (Table 1; Supplementary Figure S1). A trend to a higher rate of ORL infections (pharyngitis, otitis) among women with repetitive abortion or infertility was also observed but it did not reach statistical significance (*p* = 0.057). In contrast, no differences were observed among the three groups in relation to the rates of skin, lower respiratory tract and gastrointestinal infections (Table 1).

#### *3.3. Baseline Vaginal Health Parameters*

The vaginal pH values of the control group (4.53; range 4.38–4.68) were statistically different from those of the two study groups: 5.67 (5.55–5.79) and 5.96 (5.84–6.07) for RA and INF, respectively (*p* = 0.000; one-vay ANOVA). Similarly, the Nugent scores of the two study groups were significantly higher (5.95 (5.54–6.37) and 6.30 (5.91–6.70), respectively), than those from controls (1.79 (1.27–2.30); *p* = 0.000; one-way ANOVA) (Table 2). The CVL concentrations of the growth factors TGF-β 1, TFG-β 2 and VEFG of the control group were 4.83 (4.65–5.01) pg/mL, 3.22 (3.10–3.34) pg/mL, and 406.0 (322.0–490.0) pg/mL, respectively, while they appeared to be halved in both study groups (RA and INF), the differences being statistically significant (Table 2). No differences were observed among the three groups in relation to the remaining soluble immune factors analyzed in this work, which showed a high degree of interindividual variability (data not shown).

All women of the control group harbored lactobacilli in their vaginas (*n* = 14), the mean (95% CI) value being 7.24 (6.89–7.60) log10 CFU/mL using culture-dependent assessment. The frequency of lactobacilli detection was lower in the RA and INF groups: 57% and 26%, respectively (*p* < 0.001; Fisher exact probability tests). In addition, mean lactobacilli concentrations were 2.20 and 1.46 log10 units lower in CVL samples from lactobacilli-positive women in the RA and INF groups, respectively. The lactobacilli profile was also different (Figure 1). Seven species were identified in the samples from women of the control group, including *L. crispatus* (the dominant species), *L. jensenii*, *L. gasseri*, *L. iners*, *Limosilactobacillus* (formely *Lactobacillus*) *fermentum*, *L. salivarius*, and *Limosilactobacillus vaginalis.* However, the lactobacilli species profiles in the study groups (RA and INF) were narrower than in controls and *L. fermentum*, *L. salivarius*, and *L. vaginalis* were not detected. *L. crispatus* was the dominant species in 6 samples (43%) from fertile women, 5 samples (24%) from women with repetitive abortion and only 1 sample (4%) from infertile women. It is interesting to note that *L. iners* was isolated only from one CVL sample of the control group while it was isolated from about one-third (5 out of a total of 18 lactobacilli positive samples) from samples of RA and INF groups. *L. salivarius* was detected in the sample of a unique woman from the control group as determined by species-specific qPCR (7.29 log10 copies/mL) and culture (7.3 log10 CFU/mL) (Table 2). The strain was genetically different from *L. salivarius* CECT5713 (results not shown).


**Table 2.** Comparison of baseline vaginal parameters (pH, Nugent score, cytokines, and microbiology) of the participants (*n* = 58) which included fertile women (Control group), women with a history of repetitive abortion (RA group), and women with infertility of unknown origin (INF group).

<sup>1</sup> One-way ANOVA tests were used to evaluate differences in mean values between groups. Values followed by different superscript letters within the same row indicate statistically significant differences between groups according to Scheff<sup>é</sup> post hoc comparison tests. <sup>2</sup> Mean (95% CI) and range (min–max) values in lactobacilli-positive women. <sup>3</sup> Freeman–Halton extension of the Fisher exact probability test for a 2 × 3 contingency table were used to compute the (two-tailed) probability of obtaining a distribution of values of lactobacilli positive women. Abbreviations: TGF-β 1, transforming growth factor β 1; TGF-β 2, transforming growth factor β 2; VEGF, vascular endothelial growth factor.

Globally, the comparison of RA and INF groups at the beginning of the study revealed some statistically relevant differences (Figure 2). The mean of the vaginal pH values was 0.29 units higher in the INF group, but the opposite was observed for TGF-β 1 and VEGF, which had mean concentrations 0.43 pg/mL and 94 pg/mL higher, respectively, in the RA group. No differences were observed regarding other characteristics, including age, weight, height, Nugent score, TGF-β 2, and lactobacilli viable counts (Figure 2).

**Figure 2.** Comparison of selected baseline (**A**) demographic characteristics (age, weight and height) and (**B**) vaginal parameters (pH, Nugent score, TGF-β 1, TGF-β 2, and VEGF concentrations, and viable *Lactobacillus* counts) in CVL samples of women with repetitive abortion (RA, purple) and women with infertility of unknown origin (INF, red) at recruitment. For each boxplot, the line and the cross within the box represent the median and mean, respectively. The bottom and top boundaries of each box indicate the first and third quartiles (the 25th and 75th percentiles), respectively. The whiskers represent the lowest and highest values within the 1.5 interquartile range (IQR) and the dots outside the rectangles are suspected outliers (>1.5 × IQR). One-way ANOVA tests were used to compare both groups.

The 16S rRNA gene sequencing analysis of the CVL samples (*n* = 58) yielded 4,363,364 high quality filtered sequences, ranging from 33,160 to 139,044 per sample (median [IQR] = 73,383 [66,587–82,821] sequences per sample). Sequences were assigned to a total of 23 phyla and 453 genera, and Figure 3 shows the 5 most abundant phyla and the 20 most abundant genera in CVL samples from the fertile control group and from the RA and INF groups. The comparison of the relative abundance (% of total) of sequences at the phylum level from the three groups revealed statistically significant differences with regard to the 4 dominant phyla: *Firmicutes*, *Actinobacteria*, *Proteobacteria,* and *Bacteroidetes* (Table 3). The most frequent (present in all samples) and abundant phylum was *Firmicutes* (Figure 3). The relative abundance of *Firmicutes* in samples provided by fertile controls (median [IQR] = 99.60% [99.18–99.80%]) was higher than in samples from women of RA and INF groups (median [IQR] = 97.29% [72.34–99.35%] and 89.96% [52.46–98.85%], respectively) (*p* < 0.001; Kruskal–Wallis rank test with Bonferroni correction) (Table 3). In contrast, the median (IQR) values of the relative abundance of *Actinobacteria, Proteobacteria,* and *Bacteroidetes* were higher in women of the RA and INF groups (*p* < 0.012, *p* < 0.003, and *p* < 0.006, respectively; Kruskal–Wallis rank tests with Bonferroni correction) (Table 3).

**Figure 3.** Pie charts showing the percentages of the relative abundances of the 5 most abundant phyla (**A**) and the 20 most abundant genera (**B**) in the CVL samples from healthy fertile women (inner pie charts; C group), women with a history of repetitive abortion (middle pie charts; RA group), and women with infertility of unknown origin (outer pie charts; INF group).

> The only bacterial genus that was detected in all samples was *Lactobacillus*, but there were significant differences in its relative abundance in samples from the three groups (Table 3; Figure 3). The median [IQR] relative abundance of *Lactobacillus* in CVL samples from women of RA and INF groups (93.49% [67.18–97.53%] and 71.95% [0.76–94.09%], respectively) was lower than in samples from fertile control women (97.88% [96.92–99.31%]) (*p* = 0.001; Kruskal–Wallis rank test with Bonferroni correction) (Table 3). In fact, the only bacterial genus that characterized and differentially explained the greatest difference between the microbial communities in CVL samples between fertile control women and women of RA and INF groups was *Lactobacillus*, according to the LEfSe analysis (Figure 4).

**Figure 4.** LEfSe analysis identifying taxonomic differences in the microbiota of CVL samples from healthy fertile women (C, bluish green) and women with repetitive abortion (RA) and with infertility of unknown origin (INF). Differentially abundant bacterial taxa were identified using linear discriminant analysis (LDA) and the effect size (LEfSe) algorithm. (**A**) Histogram of LDA scores (absolute LDA (log10) score > 2.0, *p* < 0.05) showing the substantial enrichment of *Lactobacillus* in the microbiota profile of the CVL samples from healthy fertile women. (**B**) Cladogram showing LEfSe comparison of differential bacterial taxa in CVL samples. The central point represents the root of the bacterial tree and each ring the next lower taxonomic level from phylum to genus (from the inner to the outer ring: phylum, class, order, family, and genus). The color node (other than yellow) indicates which taxa are significantly higher in relative abundance, and the diameter of the node is proportional to the relative abundance of the taxon.


**Table 3.** Relative frequencies, medians and interquartile range (IQR) of the most abundant bacterial phyla and genera detected in CVL samples from fertile women (Control group), women with a history of repetitive abortion (RA group), and women with infertility of unknown origin (INF group).

<sup>1</sup> *n* (%): Number of samples in which the phylum/genus was detected (relative frequency of detection). <sup>2</sup> Kruskal–Wallis rank tests with Bonferroni correction.

> Other genera were present in a variable number of samples, ranging from 96% (*Staphylococcus* in the INF group) to 7% (*Escherichia*/*Shigella* in the control group), but the median relative abundance of any of these genera was <1% (Table 3). The bacterial profile at the genus level in some individual samples from women in the RA and INF groups did not differ from that of samples from women from the fertile control group, which were highly homogenous (Figure 5). However, aberrant profiles with reduced content or even complete

absence of *Lactobacillus* were registered in some samples from women of the RA and INF groups (Figure 5).

**Figure 5.** Relative abundance of the predominant bacterial genera in CVL samples of healthy fertile women (C), women with repetitive abortion (RA) and women with infertility of unknown origin (INF). In women with a history of reproductive failure, because either of recurrent miscarriage (RA group) or infertility (INF groups), P indicates the group of women who got pregnant after the probiotic intervention with *L. salivarius* CECT5713 and NP those women who did not.

> The analysis of alpha diversity at the genus level, calculated either by the Shannon or the Simpson's indices, revealed significant differences between the vaginal microbiota of women in the fertile and INF groups (*p* < 0.001; Kruskal–Wallis tests with Bonferroni correction) (Figure 6A,B).

> The analysis of the beta diversity, calculated according to the relative abundance of bacterial genera (Bray–Curtis distance) and the presence/absence of bacterial genera (Binnary Jaccard distance matrix), indicated that the profiles of bacterial genera of CVL samples of the 3 groups clustered apart (*p* = 0.004 and *p* = 0.002, respectively; PERMANOVA) (Figure 6C,D). In addition, samples from fertile controls clustered closer (shorter distance to centroid) according to the relative abundance of bacterial genera (Bray–Curtis distance) than those from RA and INF groups, indicating that the bacterial profiles in CVL samples from controls were highly uniform (Figure 6E,F).

> An initial assessment of potentially dominant patterns in the bacteriological profile of the CVL samples is shown in the heatmap plot presented in Figure 6G. There was a clear separation of samples based on the presence of *Lactobacillus*. One cluster was characterized by the marked and almost exclusively presence of *Lactobacillus* in CVL samples. This cluster comprised all the samples from fertile women although not exclusively, because it included also some samples from the RA and INF groups. The second cluster was characterized by the absence or reduced presence of *Lactobacillus* and the presence of multiple bacterial genera, such as *Gardenella* and *Bifidobacterium*. This second cluster contained exclusively CVL samples from the RA and INF groups. Although globally there was no clear separation between the CVL samples from the three groups, it was perceived a higher similarity between samples from the fertile control group and women with a history of repetitive abortion than between the fertile control group and women with infertility of unknown origin (Figure 6G).

**Figure 6.** Metataxonomic profiles of CVL samples of healthy fertile women (C; bluish green), women with repetitive abortion (RA; purple) and women with infertility of unknown origin (INF; red). (**A**) Comparison of alpha diversity at genus level calculated using the Shannon index between the three groups of women. (**B**) Comparison of alpha diversity at genus level calculated using the Simpson index between the three groups of women. (**C**) Principal coordinate analysis (PCoA)

plots of bacterial profiles at the genus level based on the Bray–Curtis dissimilarity analysis (relative abundance). (**D**) Principal coordinate analysis (PCoA) plots of bacterial profiles at the genus level based on the Jaccard's coefficient for binary data (presence or absence). The values on each axis label in graphs C and D represent the percentage of the total variance explained by that axis. The differences between groups of CVL samples were analyzed using the PERMANOVA test with 999 permutations. (**E**) Comparison of the mean distances of samples to the centroids in the PCoA plots based on the Bray–Curtis dissimilarity index in each group. (**F**) Comparison of the mean distances of samples to the centroids in the PCoA plots based on the Jaccard's coefficient (graph D) in each group. (**G**) Heatmap showing the relative abundance of the 20 most abundant bacterial genera (x axis) detected in CVL samples. The relative abundance of each bacterial genus within each sample is indicated by the color of the scale ranging from white (high relative abundance) to green (low relative abundance) as indicated in the scale shown at the left down corner. Dendrogram linkages are based upon relative abundance of the genus within the samples and *hclust* was used as the clustering algorithm. The column between the dendrogram of the vaginal samples and the individual values of the relative abundance of bacterial genera indicates the study group (control fertile women: C, in bluish green; women with repetitive abortion: RA, in purple; women with infertility of unknown origin: INF, in red). The differences between groups (C, healthy fertile women; RA, women with repetitive abortion; INF, women with infertility of unknown origin) were analyzed using Kruskal–Wallis tests with Bonferroni correction for data in panels A and B, and with one-way ANOVA tests for data in panels E and F.

#### *3.4. Main Outcome of the Clinical Trial: Pregnancies and Successful Pregnancies*

Administration of *L. salivarius* CECT5713 (~9 log10 CFU/day) for 6 months (or until a diagnosis of pregnancy if this happened first) to the women of the RA and INF groups led to 29 pregnancies out of the 44 participating patients. This means a pregnancy effectiveness of 66% with a 95% CI of 52–80% (Table 4). Among them, there were 25 successful pregnancies and 4 abortions. This means an effectiveness for reproductive success of 57% with a 95% CI of 42–72% (Table 4). Interestingly, all successful pregnancies led to full-term singletons (gestational age ≥ 38 weeks).

**Table 4.** Main outcomes after the probiotic treatment with *L. salivarius* CECT5713 in women with repetitive abortion (RA) and women with infertility of unknown origin (INF).


<sup>1</sup> Two women in each group end up in abortion.

Women of the RA group had the highest rate of reproductive success (15 full term pregnancies and 2 abortions out of 21 participants) (Table 4). The rate in the INF group was lower although still noticeable: 12 pregnancies (10 full term and 2 abortions) out of 23 enrolled. Therefore, the pregnancy effectiveness and successful pregnancy rates (95% CI) tended to be higher in RA group that in INF group (RR [95% CI] = 1.55 [1.00–2.42] and 1.64 [0.96–2.82], respectively), although the difference between both groups did not reach statistical significance (Table 4). It must be highlighted that all women of these groups had been unsuccessfully subjected to ART interventions in previous attempts to avoid spontaneous miscarriage (RA group) or to get pregnant (INF group).

#### *3.5. Secondary Outcomes Associated with the Probiotic Treatment: RA Group*

There were no differences in age, weight, or height between women in the RA group that ended up having a successful pregnancy (*n* = 15) and those who did not (*n* = 6) after the probiotic intervention. However, differential changes in their vaginal parameters were observed (Table 5). The vaginal pH of women who delivered was about 0.9 units lower than in those who did not (*p* < 0.001; one-way ANOVA). Similar results were noted for the Nugent score (a mean [95% CI]) reduction of 3.33 [3.73–2.93] units in women who got pregnant after the probiotic intervention versus a mean [95% CI] reduction of 0.67 [1.29–0.04] units in those who did not complete a full-term pregnancy; *p* = 0.000 one-way ANOVA) (Table 5, Supplementary Figure S2). In fact, the probiotic treatment did not modify the Nugent score in those women that did not get pregnant (*p* = 0.102; one-way repeated measures ANOVA) (Table 5).


**Table 5.** Effect of the probiotic intervention with *L. salivarius* CECT5713 on the vaginal parameters of women who were able to complete a full-term pregnancy (*n* = 15) and of those who did not (*n* = 6) among the women that had a history of repetitive abortion (RA group; *n* = 21).


**Table 5.** *Cont.*

<sup>1</sup> One-way ANOVA tests were used to evaluate differences in mean values between groups, except for lactobacilli presence. <sup>2</sup> Fisher exact probability test for a 2 × 2 contingency table. <sup>3</sup> One-way repeated measures ANOVA tests were used to determine whether there was a change in each group of participants when comparing the baseline and post-intervention parameters. <sup>4</sup> Mean (95% CI) of *L. salivarius* qPCR (copies/mL) in positive samples.

> The vaginal cytokine concentrations also differed in both subgroups of women (with successful pregnancy or not) in the RA group after the probiotic treatment. There was no modification in the vaginal TGF-β 1, TGF-β 2, and VEGF concentrations with respect to the baseline in the women who did not become pregnant, but there was a mean (95% CI) significant increase of 1.40 (1.18–1.62) pg/mL, 1.25 (1.12–1.38) pg/mL, and 402 (319–485) pg/mL, respectively, in those who did (*p* = 0.000; one-way repeated measures ANOVA) (Table 5, Supplementary Figure S2). In addition, it should be noted that there were already differences in the concentration of these cytokines even before starting the treatment between those that became and those that did not become pregnant (Table 5).

> On the other hand, the probiotic treatment resulted in a mean (95% CI) increase in lactobacilli counts of 2.12 (1.66–2.59) log10 CFU/mL in women that finally got pregnant, but there was no change in those that did not (Table 5, Supplementary Figure S2). The presence of *L. salivarius* (mean [95% CI] = 6.85 [6.58–7.12] log10 copies/mL) was confirmed by qPCR in all women that got pregnant, but only in 50% of the women with unsuccessful pregnancies, their concentration being significantly lower (mean [95% CI] = 2.63 [0.41–3.24] copies/mL) (Table 5). The lactobacilli profile in CVL samples obtained at the beginning of the probiotic treatment and after 6 months or until a diagnosis of pregnancy is presented in Figure 7. The most noticeable difference was the presence of viable *L. salivarius* in most women (17/21) after the probiotic treatment. In addition, *L. iners*, which was present in 3 women at the beginning of the study, was isolated at the end of the treatment only from 2 women who did not end up in pregnancy. There were no differences in the metataxonomic profile at the genus level of CVL samples from women of the RA group regarding the pregnancy outcome (Figure 5; Supplementary Table S1).

**Figure 7.** Changes in the profile of dominant *Lactobacillus* species in CVL samples from women with a history of repetitive abortion (RA group) and women with infertility of unknown origin (INF group) after the probiotic intervention with *L. salivarius* CECT5713. The outcome is indicated in the last file: +, successful full-term pregnancy and -, no pregnancy. The presence of isolates from a given species is indicated by a colored square.

#### *3.6. Secondary Outcomes Associated with the Probiotic Treatment: INF Group*

The women in the INF group that got pregnant after the probiotic intervention (*n* = 10) and those who did not (*n* = 13) did not differ in age, weight and height. The CVL pH and the Nugent score decreased significantly in all members of the INF group after the probiotic treatment (*p* < 0.05; one-way repeated measures ANOVA), although the magnitude of the change was smaller in the women that did not get pregnant when compared to those that got pregnant (Table 6; Supplementary Figure S2). Specifically, the mean (95% CI) reductions in CVL pH and Nugent score in women that got pregnant were −1.32 (−1.43–−1.21) and −3.90 (−4.25–−3.55), respectively, and in women that did not get pregnancy these reductions were only −0.19 (−0.29–−0.09) and−0.54 (−0.85–−0.23), respectively (Table 6; Supplementary Figure S2).

**Table 6.** Effect of the probiotic intervention with *L. salivarius* CECT5713 on the vaginal parameters of women who were able to complete a full-term pregnancy (*n* = 15) and of those who did not (*n* = 6) among the women with infertility of unknown origin (INF group; *n* = 23).




<sup>1</sup> One-way ANOVA tests were used to evaluate differences in mean values between groups, except for lactobacilli presence. <sup>2</sup> Fisher exact probability test for a 2 × 2 contingency table. <sup>3</sup> One-way repeated measures ANOVA tests were used to determine whether there was a change in each group of participants when comparing the baseline and post-intervention parameters. <sup>4</sup> Mean (95% CI) of *L. salivarius* qPCR (copies/mL) in positive samples.

> The change in the vaginal cytokine concentrations after the probiotic treatment was similar to that described in the RA group: There was no modification in the vaginal TGF-β 1, TGF-β 2, and VEGF levels of women who did not become pregnant, but there was a mean (95% CI) significant increase of 2.29 (2.16–2.42) pg/mL, 1.25 (1.13–1.37) pg/mL, and 462 (411–513) pg/mL, respectively, in those who did (Table 6; Supplementary Figure S2). In this INF group, there were already differences in the concentrations of TGF-β 2 and VEGF, but not in that of TGF-β 1, between those that became and those that did not became pregnant even before starting the treatment (Table 6).

> The probiotic intervention resulted in a high degree of vaginal colonization by lactobacilli (6.46 [5.94–6.98] log10 CFU/mL) of all women that got pregnant, while this only happened in 46% of those that experienced a treatment failure, the density of lactobacilli reached being significantly lower (4.95 [4.28–5.62] log10 CFU/mL) (Table 6). Similarly to the RA group, the presence of *L. salivarius* (mean [95% CI] = 6.48 [6.28–6.68] copies/mL) was confirmed by qPCR in all women that got pregnant, but only in 31% of the women with unsuccessful pregnancies and, then, at a lower concentration (mean [95% CI] = 3.55 [3.24–3.86] copies/mL) (Table 6). The main difference in the lactobacilli profile of CVL samples of women in the INF group registered after the probiotic intervention was the detection of viable *L. salivarius* in all women who got pregnant, but only in 4 out of 13 of those women that failed to get pregnant. There were no differences in the metataxonomic profile at the genus level of CVL samples from women of the RA group regarding the pregnancy outcome (Figure 5; Supplementary Table S2).

#### *3.7. Comparison of Vaginal Parameters between Women Who Became Pregnant and Those Who Did Not from Both the RA and INF Groups*

The mean [95% CI] pH value in CVL samples was slightly but significantly more acidic in the women who become pregnant (5.69 [5.57–5.81] units) than in those who did not (5.99 [5.85–6.13] units) (*p* = 0.024; one-way ANOVA) (Figure 8; Supplementary Table S3). There were also differences in the concentration of vaginal cytokines TGF-β 2 and VEFG at the beginning of the study according to the final pregnancy outcome, but the differences were similar to those described already separately for RA and INF groups (Figure 8; Supplementary Table S3). The only parameters that did not differed initially between both groups were the Nugent score, TGF-β 1 concentration, and the frequency of detection and counts of lactobacilli (Figure 8; Supplementary Table S3). Globally, *Lactobacillus* was detected in all women who became pregnant, but only in half of those that did not (*p* < 0.001; Fisher exact probability test).

**Figure 8.** Changes in vaginal parameters (pH, Nugent score, TGF-β 1, TGF-β 2, and VEGF concentrations, viable *Lactobacillus* counts and *L. salivarius* copies in CVL samples) in women with a history of reproductive failure, because either of recurrent miscarriage (RA group) or infertility (INF groups), after the probiotic intervention with *L. salivarius* CECT5713 according to their outcome (pregnancy versus no pregnancy).

The probiotic intervention resulted in differential and remarkable changes in the vaginal parameters in those women who became pregnant but not in those who did not (Figure 8; Supplementary Table S3). First, the probiotic administration of *L. salivarius* CECT5713 resulted to be more effective regarding the change in the vaginal pH and Nugent score in women who got pregnant, which recorded mean (95% CI) decreases of −1.20 (−1.29–−1.12) and −3.56 (−3.82–−3.30) units, respectively (*p* = 0.000; one-way repeated measures ANOVA). In contrast, the change in these two parameters was smaller (−0.21 (−0.31–−0.10) and −0.58 (−0.88–−0.28) units, respectively) in the group of women who did not get pregnant (Figure 8; Supplementary Table S3). Second, the probiotic intervention led to a significant increase in the concentrations of vaginal cytokines TGF-β 1, TGFβ 2 and VEFG (mean [95% CI] increase of 1.76 [1.60–1.91] pg/mL, 1.25 [1.17–1.33] pg/mL, and 426 [378–473] pg/mL, respectively) in women who got pregnant but no change was registered in the group that did not (Figure 8; Supplementary Table S3). Third, regarding the lactobacilli profile of CVL samples, there was a mean (95% CI) increase of 2.67 (2.26–3.08) log10 CFU/mL units in viable *Lactobacillus* counts after the probiotic treatment in the group of women who became pregnant as opposed to those that did not. Differences were also noted on the *L. salivarius* content in CVL samples. This lactobacilli species was detected, and at a high concentration (mean [95% CI] = 6.70 [6.52–6.89] log10 copies/mL), in CVL samples from all women having a successful pregnancy unlike women who did not become pregnant (Figure 8; Supplementary Table S3). The metataxonomic profile at the genus level of CVL samples from women of the INF group was equal in women that did or did not become pregnant, except for a slightly higher relative frequency of *Escherichia/Shighella* in women that got pregnant (Figure 5; Supplementary Table S4).

#### *3.8. Comparison of Vaginal Parameters between Control Women, All Women Who Became Pregnant and Those Who Did Not from Both RA and INF Groups*

The analysis of post-intervention vaginal parameters (pH, Nugent score, TGF-β 1, TGF-β 2, VEGF, lactobacilli counts) revealed that the pH value of CVL samples and Nugent score in women who became pregnant after the probiotic intervention were similar to those of fertile control women (Table 7; Supplementary Figure S3). The concentrations of TGF-β 1, TGF-β 2, and VEGF in post- intervention CVL samples of women who became pregnant were closer to those found in fertile control women, although statistically significant differences were found between them (Table 7; Supplementary Figure S3). Besides, it is remarkable to note that the post-intervention concentration of VEGF in women that became pregnant was about twice that registered in fertile control women (mean [95% CI] = 755.0 [637.1–872.5] pg/mL and 406.0 [322.0–490.0] pg/mL, respectively). There was a high interindividual variation in lactobacilli counts varying from undetectable (in 57% of the women who did not become pregnant) to 7.5 log10 CFU/mL in CVL samples of women who did not become pregnant after the probiotic intervention, but the mean [95% CI] value (4.87 [3.83–5.90] log10 CFU/mL) was lower than in samples of the other participants (Table 7; Supplementary Figure S3). There was less than 1 log10 CFU/mL difference between the lactobacilli viable counts in CVL samples of women who enjoyed a full term pregnancy after the probiotic intervention and those of fertile controls (mean [95% CI] = 6.47 [6.22–6.72] log10 CFU/mL and 7.24 [6.89–7.60] log10 CFU/mL, respectively) (Table 7; Supplementary Figure S3).

**Table 7.** Comparison of vaginal parameters (pH, Nugent score, TGF-β 1, TGF-β 2, and VEGF concentrations, and *Lactobacillus* counts) of all women who were able to complete a full-term pregnancy (*n* = 25) and of those who did not (*n* = 19) among all women with a history of repetitive abortion and with infertility of unknown origin (RA and INF groups) after the probiotic intervention with *L. salivarius* CECT5713 and vaginal parameters of fertile women (Control group; *n* = 14).


<sup>1</sup> One-way ANOVA tests were used to evaluate differences in mean values between groups. Values followed by different superscript letters within the same row indicate statistically significant differences between groups according to Scheff<sup>é</sup> post hoc comparison tests. <sup>2</sup> Freeman–Halton extension of the Fisher exact probability tests for a 2 <sup>×</sup> 3 contingency table were used to compute the (two-tailed) probability of obtaining a distribution of values of lactobacilli positive women. <sup>3</sup> Mean (95% CI) of *L. salivarius* qPCR (copies/mL) in lactobacilli-positive women. TGF-β 1, transforming growth factor-β 1; TGF-β 2, transforming growth factor-β 2; VEGF, vascular endothelial growth factor.

> Additionally, a network structure of the baseline vaginal bacterial genera communities on the three different groups of women (fertile controls, women who got pregnant after the probiotic intervention and women who did not get pregnant after the probiotic intervention) was constructed based on the genus-genus correlations (Figure 9). In the group of fertile women, the strongest correlation was observed between two minority genera, *Escherichia/Shigella* and *Enterococcus*; the most abundant genera, *Lactobacillus*, established negative and weak relationship with other Firmicutes (*Finegoldia* and *Peptoniphilus*) and *Prevotella.* In contrast, in the group of women with either repeated abortions or infertility of unknown origin, *Lactobacillus* showed strong negative association with two genera of the *Actinobacteria*, *Gardenella,* and *Bifidobacterium*. However, in the group of women that responded to the probiotic intervention and ended up in a successful pregnancy, the strongest negative association was between *Lactobacillus* and *Gardenella*, while in those women that did not get pregnant this negative association was weaker than that registered between *Lactobacillus* and *Bifidobacterium,* indicating that indeed the bacterial profile in CVL samples may indicate different fertility problems (Figure 9).

**Figure 9.** Estimated network structures based on a sample of 58 vaginal samples: 14 from healthy fertile women (**A**, Control group), 25 from women with a successful reproductive outcome after the probiotic intervention *L. salivarius* CECT5713 (**B**, Pregnancy) and 19 from women who did not have a successful pregnancy after the probiotic intervention with *L. salivarius* CECT5713 (**C**, No pregnancy). The 14 most abundant genera were represented. Red lines indicate negative correlation and green lines indicates positive correlation. The thickness and the intensity of the line reflects the intensity of the correlation.

#### **4. Discussion**

In this study, the comparison between the vaginal microbiota of women with a history of reproductive failure, due to recurrent miscarriage or infertility, and healthy fertile women confirmed that dominance of specific species of *Lactobacillus* in the vaginal microbiota plays a determinant role in the success of human reproduction. Overall, the lowest vaginal pH values and Nugent scores were associated with vaginal communities dominated by lactobacilli, while those with the highest pH values and Nugent scores were associated with a depletion of lactobacilli. Close associations between low pH, low Nugent score and a high concentration and dominance of lactobacilli in the human vagina has been repeatedly reported [3,4,65]. In this study, the frequency of detection of lactobacilli in the vaginal samples was much higher in fertile women (100%) than in women with repetitive miscarriage (57%). Interestingly, infertile women showed the lowest percentage of women from whom lactobacilli could be isolated (26%). Use of antibiotics in both infancy and adulthood was significantly higher among women of the RA and INF groups than among women of the control group. It has been long known that opportunistic vaginal infections may arise as an adverse effect to the use of antibiotics because of their negative effect on the lactobacilli population [66]. The results obtained in this study suggest, for the first time, that an antibiotic-associated depletion of vaginal lactobacilli may have long-term health consequences by impairing fertility or embryo implantation and that such effect may be contrasted reversed by microbiological modulation of the vaginal ecosystem.

The species most frequently isolated from vaginal samples in this study belonged to *L. crispatus*, *L. gasseri*, *L. iners,* and *L. jensenii*, which are particularly common and abundant in the human vagina and absent or infrequently found in other human habitats [3,32,67]. Stable codominance of multiple *Lactobacillus* species is rarely observed in the same vaginal community [67]. Initial presence of *L. crispatus* seemed to be positively correlated with a successful reproductive outcome after the intervention with the probiotic assayed in this study. In contrast, initial presence of *L. iners* and *L. gasseri* seemed to be negatively correlated with a successful reproductive outcome after the probiotic intervention unless the *L. salivarius* strain provided in the trial was able to become dominant in the vaginal samples. *L. crispatus* and *L. iners* are probably the most common inhabitants of the healthy human vagina and are able to perform relevant ecological functions in the vaginal environment. Transitions from a vaginal community dominated by *L. iners* to one dominated by *L. crispatus*, and viceversa, seems to be relatively frequent [68]. The relationships between these two species and their potential functions have received an increasing scientific interest in the last years [67–71]. However, while there is a general agreement that a *L. crispatus*-dominated vaginotype promotes vaginal and reproductive health [72–74], the role of *L. iners* is very controversial since this peculiar species has been associated to beneficial roles for vaginal health [8,75] but, also, to dysbiosis, vaginal infections and a variety of gynecological conditions, including adverse pregnant outcomes [69,71,76–78]. Functional studies are required to investigate its roles in vaginal bacterial communities and whether, under certain circumstances, it can be used as a biomarker of reproductive failure.

A characterization of some properties of *L. salivarius* CECT5713 that may be relevant for vaginal and reproductive health showed that this strain was able to inhibit all the clinical isolates of *G. vaginalis*, *S. agalactiae*, *C. albicans*, *C. glabrata*, *C. parapsilosis*, and *U. urealyticum* tested in this study. This antimicrobial activity is relevant since vaginal infections are associated with an increased risk of adverse urogenital and reproductive health outcomes [79]. *L. salivarius* CECT5713 has a high acidifying ability by producing high amounts of L-lactic acid and small amounts of acetic acid [37]. Eubiosis and dysbiosis in the vaginal communities are distinguished by the high concentration of lactic acid and the high acidity that characterize the eubiosis state [79–81], as a direct result of the metabolic activity of the local lactobacilli, which is enough to inactivate reproductive tract pathogens, including viruses, bacteria and yeasts [49,82–87]. The capability and rate of production of lactic acid by lactobacilli is strain-specific and only high levels of lactic acid and a concomitant very low pH can inhibit microbial growth efficiently in the local vaginal biofilm [88,89]. From this point of view, *L. salivarius* CECT5713 seems a suitable candidate as a probiotic for the cervicovaginal target. In addition, this strain encodes an α-amylase in its genome (GenBank: ADJ79335.1), which is fully functional as revealed in the activity assays performed in this work. This enzyme might contribute, together with host α-amylase, to degradation of vaginal glycogen and, therefore, to increase lactic acid production and to maintain the vaginal pH at ≤4.5, promoting the desired lactobacilli dominance in the vaginal ecosystem [90].

Other properties of *L. salivarius* CECT5713 that are interesting in relation to the control of harmful vaginal microbes include a high rate of adhesion to vaginal cells and coaggregation with the vaginal pathogens used in this study. High adherence of *L. salivarius* strains to vaginal cells has been previously observed and related to the prevention of vaginal colonization by *S. agalactiae* [49]. Both adhesion and co-aggregation activities seem to be highly strain-specific traits [48,49,91,92]. Cell-dependent reduction of *Candida* spp. adhesion by *Lactobacillus* species has been related to co-aggregation and competition for binding sites [93,94]. Overall, *L. salivarius* CECT5713 seems to be a strain suitable for applications involving vaginal homeostasis. This strain was isolated from human milk and infant feces of a healthy mother–child pair [37], and has been shown to be a good probiotic strain due to its extensive repertoire of desirable properties and safety, being particularly suited for application in the mother–infant dyad [38].

In this work, oral administration of *L. salivarius* CECT5713 to women of the RA and INF groups led to a relevant number of pregnancies. Women of such groups who had term pregnancies experienced significant changes in some key microbiological, biochemical and immunological parameters in the vaginal samples, such as concentration of cultivable lactobacilli, concentration of *L. salivarius* specific DNA, pH, Nugent score, and concentrations of VEGF, TGF-β 1 and TGF-β 2. The fact that all of them had high concentrations of *L. salivarius* in the vaginal samples and that DNA from this species was also detected by the qPCR assay reveals that the strain was able to reach and colonize the vaginal mucosa. The significant reductions of the pH values after the treatment indicate that the strain was metabolically active and suggests a good agreement between the in vitro potential of the strain and its in vivo capabilities.

The changes induced by *L. salivarius* CECT5713 in the concentrations of the growth factors VEGF, TGF-β 1 and TGF-β 2 seem to be particularly relevant and can be considered as biomarkers of the efficacy of the strain for the target pursued in the clinical trial. VEGF is a 45-kDa homodimeric heparin-binding glycoprotein with angiogenic activity that plays a key role as regulator of vasculogenesis, angiogenesis and vascular function in the human endometrium [95,96]. Vasculogenesis and angiogenesis are crucial steps for embryogenesis and particularly for embryo implantation (vessel formation and trophoblastic invasion) and both processes have been correlated with an increased expression of VEGF and VEGF receptors [97–101]; otherwise, endometrial angiogenesis may be impaired and result in a lethal phenotype, ranging from failed implantation to first-trimester miscarriage [95,102–105].

TGF-β 1 and TGF-β 2 also promote angiogenesis in vivo [106], and participate in implantation, trophoblast differentiation, and immunoregulation at the maternal-fetal interface [100,107]. Transcription of TGF-β 1 increases notably in human uterine endometrium during the first trimester of pregnancy [108], while recurrent pregnancy is associated with a decrease in the decidual TGF-β [109–111]. Expression of both VEGF and TGF-β 1 is highly regulated in a temporal and spatial manner during the early stages of implantation, a fact that underlines their critical role in the evolving pregnancy [109–111]. In addition, TGF-β 1 increases expression of VEGF in the trophoblast [111–115] suggesting a link between the action of both growth factors.

TGF-β 1 and TGF-β 2 are also of particular interest in this field because of their well-known roles in regulating the inflammatory response and inducing active immune tolerance in mucosal tissues [116,117]. Interestingly, both are present at very high concentrations in human seminal fluid [118,119], acting as male-female signaling agents that regulate the female immune response to sperm after coitus and promote maternal immune

tolerance for embryo implantation and subsequent pregnancy [120–123]. Although studies in mouse models have shown that exposure to the high concentrations of TGF-β present in seminal fluid is absolutely required to boost uterine Treg cells prior to embryo implantation [124–129], this fact is not taken into account in many current ARTs, including IVF techniques, where such exposure is absent. Most TGF-β present in human semen is latent and requires activation to bind to receptors on cervical cells [130,131]. Interestingly, activation after coitus is facilitated by the acid pH of the vaginal environment [123] and, in this study, administration of *L. salivarius* CECT5713 led to an increase of TGF-β 1 and TGF-β 2 concentrations and, concomitantly, to a significant decrease in the vaginal pH values.

Our study has some limitations. First, the microbiota of the genitourinary tract of the partner was not evaluated and some studies have shown that male microbiota may also play a fundamental role in reproductive outcomes [132,133]. In fact, the couple (when applicable) should be considered as a single entity to achieve the best reproductive outcomes [134]. This approach will be taken into account in our future studies in this field. In addition, the metataxomomic analysis included in this study was carried at the genus level since the 16S rRNA gene approach has poor discriminatory power at the species level [135,136]. Other approaches, such as shotgun sequencing, should be used in the future to solve such limitation and to have a broader view of the vaginal microbiome.

Although our knowledge of the mechanisms that these early embryo–maternal interactions has increased in recent years, implantation remains as a rate-limiting step in human ART and the currently available treatments for infertility or recurrent pregnancy loss of unknown etiology have a rather limited efficacy [137,138]. Therefore, the possibility of enhancing angiogenic and tolerance activities in the endometrium by modifying the reproductive microbiota using bacterial strains specifically tailored for these targets provides a novel strategy to improve reproductive functions and deserves future basic and clinical research efforts.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2072-664 3/13/1/162/s1, Supplementary Figure S1. History of recurrent infections and use of antibiotics in infancy and adulthood among the women recruited in this study. Supplementary Figure S2. Changes in vaginal parameters (pH, Nugent score, TGF-β 1, TGF-β 2, and VEGF concentrations, viable *Lactobacillus* counts in CVL samples) in women with a history of reproductive failure, because either of recurrent miscarriage (RA group) or infertility of unknown origin (INF groups), after the probiotic intervention with *L. salivarius* CECT5713 according to their outcome (pregnancy versus no pregnancy). Supplementary Figure S3. Comparison of vaginal parameters (pH, Nugent score, TGF-β 1, TGF-β 2, and VEGF concentrations, viable *Lactobacillus* counts in CVL samples) in healthy fertile women (C, control group) and those of women with a history of reproductive failure, because either of recurrent miscarriage (RA group) or infertility of unknown origin (INF groups), after the probiotic intervention with *L. salivarius* CECT5713. One-way ANOVA tests followed by Scheffé *post hoc* comparison tests were used to compare the groups; different letters above the boxplots indicate significant differences. Supplementary Table S1. Relative frequencies, medians and interquartile ranges (IQR) of the most abundant bacterial phyla (grey shadow) and genera detected in CVL samples from women who were able to complete a full-term pregnancy (*n* = 15) and of those who did not (*n* = 6) among the women that had a history of repetitive abortion (RA group; *n* = 21). Supplementary Table S2. Relative frequencies, medians and interquartile ranges (IQR) of the most abundant bacterial phyla (grey shadow) and genera detected in CVL samples from women who were able to complete a full-term pregnancy (*n* = 10) and of those who did not (*n* = 13) among the women with infertility of unknown origin (INF group; *n* = 23). Supplementary Table S3. Differences in the baseline characteristics and effect of the probiotic intervention with *L. salivarius* CECT5713 on the vaginal parameters of all women who were able to complete a full-term pregnancy (*n* = 25) and of those who did not (*n* = 19) among all participants from both RA and INF groups (*n* = 44). Supplementary Table S4. Relative frequencies, medians and interquartile ranges (IQR) of the most abundant bacterial phyla (grey shadow) and genera detected in CVL samples from women who were able to complete a full-term pregnancy (*n* = 25) and of those who did not (*n* = 19) among women with a history of reproductive failure, because either of recurrent miscarriage (RA group) or infertility of unknown origin (INF groups) (*n* = 44).

**Author Contributions:** D.B., L.F., and J.M.R. designed and coordinated the study. D.B. recruited participants and recorded samples-associated metadata. I.C. and R.A. processed the samples and performed the microbiological and immunological analyses. C.A. executed bioinformatic analysis. L.F. and J.M.R. drafted the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was partly funded by contract 291-2018 established between Complutense University of Madrid and Biosearch Life S. A. (Granada, Spain). Irma Castro is the recipient of a predoctoral contract (BES-2017-080713) from the Ministerio de Ciencia, Innovación y Universidades (Spain).

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Ethics Committee of Consejería de Salud y Familias (Junta de Andalucía, Granada, Spain) (protocol code P050/19, Act 11/19, 10th December 2019).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author. The data are not publicly available due to ethical restrictions.

**Acknowledgments:** We sincerely thank all the women that participated in the assay. We also thank Evaristo Suárez (University of Oviedo, Spain) for critical reading of the manuscript and fruitful discussions.

**Conflicts of Interest:** Biosearch Life S.A., the company that partly funded the study, is the proprietary of the strain *L. salivarius* CECT 5713.

#### **References**


## *Review* **Prevention and Management with Pro-, Pre and Synbiotics in Children with Asthma and Allergic Rhinitis: A Narrative Review**

**Lien Meirlaen †, Elvira Ingrid Levy † and Yvan Vandenplas \***

KidZ Health Castle, UZ Brussel, Vrije Universiteit Brussel, 1090 Brussels, Belgium; lienmeirlaen@hotmail.com (L.M.); elvira.levy9@gmail.com (E.I.L.)

**\*** Correspondence: yvan.vandenplas@uzbrussel.be; Tel.: +32-475748794

† These authors contributed equally to this work.

**Abstract:** Allergic diseases including allergic rhinitis and asthma are increasing in the developing world, related to a westernizing lifestyle, while the prevalence is stable and decreasing in the industrialized world. This paper aims to answer the question if prevention and/or treatment of allergic rhinitis and asthma can be achieved by administrating pro-, pre- and/or synbiotics that might contribute to stabilizing the disturbed microbiome that influences the immune system through the gut–lung axis. We searched for relevant English articles in PubMed and Google Scholar. Articles interesting for the topic were selected using subject heading and key words. Interesting references in included articles were also considered. While there is substantial evidence from animal studies in well controlled conditions that selected probiotic strains may offer benefits in the prevention of wheezing and asthma, outcomes from clinical studies in infants (including as well pre- and postnatal administration) are disappointing. The latter may be related to the multiple confounding factors such as environment, strain selection and dosage, moment of administration and genetic background. There is little evidence to recommend administration of pro, pre- or synbiotics in the prevention of asthma and allergic rhinitis in children.

**Keywords:** probiotics; prebiotics; synbiotics; microbiome; children; allergic rhinitis; asthma

#### **1. Introduction**

#### *1.1. Prevalence of Asthma and Allergic Rhinitis*

The global prevalence of atopic diseases such as asthma, allergic rhinitis and atopic dermatitis is remarkable and has been expanding over the years [1]. Allergic rhinitis occurs in 10 to 30% of adults and up to 40% in children and its prevalence is increasing [2]. With around 339 million people affected globally, asthma is one of the most common long-term non-transmissible diseases [3]. The worldwide prevalence of doctor-diagnosed asthma in adults is 4.3% (95% confidence interval (CI) 4.2–4.4), with a wide variation between countries: the highest occurrence is found in developed countries such as Australia (21%) and the lowest in third world countries such as Ethiopia (2%) [4]. In children, asthma is more frequent in boys than in girls due to their smaller airways relative to their lung size, with a turnaround during puberty, as the prevalence in women is 20% higher than in men [5]. Asthma prevalence is steady or even shrinking in many developed countries, but as lifestyles become more westernized in developing countries, there is a fast increase in its prevalence in these parts of the world [6]. The interaction between the genomic background, changing environmental conditions such as more pollution [7], increasing obesity, the "hygiene hypothesis" and less breastfeeding [8] is likely to play a crucial part. Parental reduction in smoking has proven to reduce asthma [9]. Important to mention is that in less developed countries, the detection rate of allergic disease is likely to be lower, which may result in an underestimation of its prevalence [10]. By identifying and

**Citation:** Meirlaen, L.; Levy, E.I.; Vandenplas, Y. Prevention and Management with Pro-, Pre and Synbiotics in Children with Asthma and Allergic Rhinitis: A Narrative Review. *Nutrients* **2021**, *13*, 934. https://doi.org/10.3390/nu13030934

Academic Editors: Sonia González, Nuria Salazar and Silvia Arboleya

Received: 16 January 2021 Accepted: 11 March 2021 Published: 14 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

characterizing more of these conditions and the involved lifestyle factors, epidemiologic studies try to deduce potential approaches for prevention of allergic diseases [11]. Asthma causes impaired life quality, substantial disability and preventable deaths in children and adolescents, combined with important health care costs [6]. As a consequence, the increased social and economic burden of asthma makes asthma prevention an important public health goal [12].

#### *1.2. Pathophysiology Asthma and Allergic rhinitis*

Atopic diseases like asthma and allergic rhinitis are complex multifactorial conditions of which the outcome is strongly influenced by a complex interplay between genetic background, the state of the body's defenses, gut microbiota and the environment. There are different mechanisms and typical pathological characteristics of asthma immunopathology, which can be divided in three groups: non-eosinophilic (neutrophilic type 1 and type 17 and pauci-granulocytic), eosinophilic (allergic and non-allergic), and mixed granulocytic inflammation [6]. The eosinophilic group represents 50% of all asthma patients. In this process, allergen or trigger factor exposure stimulates local inflammatory responses mediated by immunoglobulin E (IgE) release. This leads to allergen sensitization and the forming of an atopic response. Type 2 T helper (Th2) cells play a crucial part in this inflammatory process by producing cytokines that control fabrication of allergen-specific immunoglobulin E and inflammation of tissue characterized by the invasion of eosinophils, mast cells and activated CD4+ T-cells. Regulatory T-cells (Treg) are involved in preventing the sensitization to allergens by the production of anti-inflammatory cytokines such as IL-10, by secreting transforming growth factor B, and by possibly suppressing the production of immunoglobulin E and proliferation of Type 1 T helper (Th1)/Type 2 T helper (Th2) balance. The mechanisms of tolerance induction are complex [13]. The intestinal microbiome contributes to the pathological process of allergic diseases because of its notable effect on mucosal immunity. A healthy microbiome at a young age changes the balance between T helper 1 T helper 2, shifting towards a T helper 1 cell response. About 60–70% of the immune cells are located within the gastrointestinal tract. On the other hand, atopic diseases involve Type 2 T helper reactions to allergens. Unusual allergic responses are believed to occur in cases of intestinal dysbiosis during the development of the immune system, causing a shift of the Th1/Th2 cytokine balance towards a Th2 response, a consequent activation of Th2 cytokines and increased production of IgE [14]. Additionally, there is increasing evidence that a balanced gut microbiome is needed for the proper formation of T-regulatory cells, which are important for tolerance induction [13].

#### *1.3. Definitions Pro-, Pre- and Synbiotics*

Probiotics are live microorganisms that, when administered in sufficient quantities, give a health improvement of the host. Probiotics induce immunomodulatory mechanisms in many different ways, including skewing of the Th1/Th2 balance towards Th1 by inhibiting Th2 cytokines or indirectly expanding IL-10 and Treg formation via either dendritic cell development or Toll-like receptors, although the exact mechanism remains to be clarified [15]. Prebiotics are substrates that are selectively utilized by host microorganisms conferring a health benefit. Synbiotics are defined as a mixture comprising live microorganisms and substrate(s) selectively utilized by host microorganisms that confer a health benefit on the host.

#### *1.4. Rationale for Using Pro-, Pre- and Synbiotics in Atopic Diseases*

Living circumstances in the industrialized world such as a decreased fermented food consumption, increased intake of antibiotics and other drugs, and improved hygiene are according to data from epidemiologic studies associated to the increase in allergic diseases. More or less exposure to microbial stimuli during infancy is associated to more or less allergic disease. The association has been described as the "hygiene hypothesis". A lack of exposure to microbial stimuli early in childhood is a major factor involved in

the steep increase in allergy [16]. In those who spend their childhood on a farm, allergic diseases are less common [16]. The comparison between the composition of microbiota of farm children and the microbiota of children with other lifestyles shows a significant difference [16]. Children living on farms are exposed to a wider range of microbes than children not living on a farm, and this exposure explains a substantial fraction of the inverse relation between asthma and growing up on a farm [17]. The gastrointestinal microbiota composition differs between allergic and healthy infants, independent of the prevalence of allergic disease in the region [13]. In contrary to what has been believed for a long time, an amniotic microbiome has been reported, and as a consequence, the fetal intestine may not be sterile since there is the presence of microbial deoxyribonucleic acid in meconium [18]. Early life is characterized by a rapid change in gastrointestinal microbiota composition. The first altering factor of the neonatal microbiome is the contact with vaginal, fecal and skin bacteria of the mother. In caesarean section-born babies, a less diversified microbiome is observed. The second altering factor is feeding. Human milk is rich in oligosaccharides which have prebiotic properties (a substrate that is selectively utilized by host microorganisms conferring a health benefit [19]) and promote the growth of selected species of bacteria. Human milk is also a natural bacterial inoculum. The third altering factor is environmental influenced alterations, which may undo the first two beneficial gut alterations: environments like neonatal intensive care units and medication such as antibiotics or proton pump inhibitors administered perinatal or during early life [20,21].

During early life, a balanced gastrointestinal microbiota is of major importance for the balanced skewing of the developing of the immune system and also determines the gut– lung communication of the gut–lung axis. Therefore, dysbiosis of the intestinal microbiome during early life will contribute to immune-mediated diseases later in life [14]. However, these associations between gut microbiota and allergic disease cannot provide a satisfactory explanation for all observations and does not result in evidence to decrease the rise in allergic disorders. However, the microbiota hypothesis does provide a rationale for using pro-, pre- and synbiotics, to alter the microbiota composition in the intestine to result in a more balanced development of the immune system [13]. Since a child's microbiota does not reflect adult patterns until they are two years old, the infant microbiota may be more susceptible to manipulation [22].

More knowledge is needed on the mechanisms behind dysbiosis, translocation of microbiota from the intestine to the respiratory tract through various mechanisms and for a better evaluation of the therapeutic possibilities to correct this dysbiosis, which in turn can be used to manage various respiratory diseases [23].

In this paper, we will try to answer the question if probiotics or prebiotics and/or synbiotic supplementation can alter the microbiome sufficiently to have an efficacious prevention and/or management of allergic rhinitis and asthma.

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

A search was performed in PubMed, EMBASE, Google Scholar, Web of Science and Cochrane Library. We included preferably meta-analyses, systematic reviews and clinical trials from 1990 up until October 2020 published in the English language. The following keywords in the respective language were used: "asthma", "wheezing", "respiratory disease", "allergic rhinitis", "allergic coryza", "probiotics", "prebiotics", "synbiotics", "prevention", "therapy", "therapeutics", "child". These keywords were combined with the Boolean command "OR" and were linked by the Boolean command "AND". Records were screened based on the titles and abstracts. Articles were extracted using subject heading and key words of interest to the topic. A second selection was made by reading the abstract. Interesting references in included articles were also considered. Records were excluded if the abstract or full text was not available, if the topic was not relevant, if non-English or if the study design was not adequate. Duplicates were removed.

Search strategy for human studies in the results section: In PubMed, the following search string was used: ("Asthma"[MeSH Terms] OR "respiratory disease"[Title/Abstract] OR "wheezing"[Title/Abstract] OR "recurrent wheeze"[Title/Abstract] OR "rhinitis, allergic, seasonal"[MeSH Terms] OR "allergic coryza"[Title/Abstract]) AND ("Probiotics"[MeSH Terms] OR "Prebiotics"[MeSH Terms] OR "Synbiotics"[MeSH Terms]) AND "Child"[MeSH Terms].

#### **3. Results**

#### *3.1. Probiotics for Prevention of Asthma*

#### 3.1.1. Animal Studies

A beneficial effect of the administration of probiotics was suggested by showing that oral administration of *Lactococcus lactis* NZ9000 to rats resulted in a decrease in infiltration of pro-inflammatory leucocytes, mainly eosinophils and decreased lung IL-4 and IL-5 expression in the broncho-alveolar lavage and a reduced level of serum allergen-specific IgE [24]. Another study conducted in mice using *Lactobacillus rhamnosus* GR-1 significantly prevented airway hyperreactivity development and prevented microbiome disturbance in the asthmatic animals, supporting the existence of the gut–lung axis [25]. An interesting aspect is that most probiotics are given orally; however, a new approach was tested by giving probiotics (*Lactobacillus paracasei* NCC2461 [26] and *Lactobacillus rhamnosus* GG [27] in mice through the nose and showed benefits in reducing inflammation of the lungs [28]. The probiotic *Bifidobacterium breve* M-16V administered to pregnant mice was shown to be effective in lowering eosinophils in the broncho-alveolar lavage fluid of neonatal mice and reduced allergic lung inflammation in mice exposed to air pollution [29]. In another animal study, the intranasal administration of *Lactobacillus rhamnosus* GG (LGG), but not *Lactobacillus rhamnosus* GR-1, suppressed airway hyper-reactivity and reduced the counts of eosinophils, IL-13 and IL-5 in broncho-alveolar fluid [27]. In addition to inhibiting inflammatory cell infiltration in lung tissue, *Lactobacillus* GG was shown to decrease MMP9 expression, a class of enzymes that are involved in the degradation of the extracellular matrix and of which levels were significantly increased in asthma [30]. *Lactobacillus* GG and *Bifidobacterium lactis* were shown to increase natural regulatory T cells in the lungs of asthmatic mice in another animal study [31]. Lee et al. mentioned that four *Lactobacillus* species used in animal studies had different immunomodulatory effects [32] against allergy *Lactobacillus planetarum* had shown some beneficial effect, but this was not the case for *Lactobacillus salivarius* and *fermentum* [33]. Probiotic strain-specific induction of Foxp3þ T regulatory cells was found in mouse allergy models [34].

#### 3.1.2. Human Studies

In humans, evidence of the use of probiotics as a preventive agent for respiratory allergies in children was reported to be low [35] (Table 1). A meta-analysis of 2013 showed that by giving the most frequently used probiotics (*Lactobacillus* spp. and/or *Bifidobacteria* spp.) to prenatal mothers plus continued after birth versus only postnatally, no difference in IgE levels were seen. Less atopy was seen if the probiotics were given to pregnant women and continued after birth. Probiotics given after birth only decreases the risk of atopic sensitization in young children but not of asthma or wheeze [14]. This supports the theory that probiotics that have colonized the mothers' intestine will be transferred at birth during vaginal delivery. Further administration of pro- and prebiotics to the pregnant mother results in the potential transmission of tolerogenic mediators such as regulatory cytokines, antibodies and growth factors across the placenta, stimulating the development of the fetal immune system [36]. This could help to prevent asthma or allergic rhinitis. Like mentioned above, the findings in pregnant mice are of human interest since up to now, knowledge was restricted to the fact that *Bifidobacterium breve* M-16V in infants can suppress T-helper type 2 immune responses and modulate the systemic Type 1 T helper/Type 2 T helper balance. Exposure of the pregnant mother to air pollution increases asthma susceptibility of the newborn and later on. Therefore, *Bifidobacterium breve* M-16V might contribute to reducing asthma in a population living in highly polluted areas [29]. In 2014, the Panda Study showed that giving a probiotic mixture postnatally (two *Bifidobactera* spp. and *Lactococcus lactis*) for one year does not have a beneficial effect

on the development of allergic diseases after six years [37]. After five years follow-up, the negative outcome persisted [38]. Furthermore, no association (relative risk (RR) 0.59, 95% CI 0.36–0.96, *p* = 0.059) was found in a study with a follow-up of 11 years. This study was a two-center RCT using *Lactobacillus rhamnosus* HN001 or *Bifidobacterium lactis* HN019 daily taken from 35-week gestation to six months postpartum in mothers while breastfeeding and from birth to the age of two years in infants [39]. Consistent with the previously mentioned studies, a more recent meta-analysis including 19 RCTs involving 5157 children showed no association as well in lowering the incidence of asthma and wheezing if probiotics were given to pregnant mothers or postnatally. However, in infants with atopic diseases, probiotics seem to reduce the wheezing incidence significantly (RR 0.61, 95% CI 0.42– 0.90; *p* < 0.05). No association was found between probiotics and a subgroup analysis of asthma (RR 0.94, 95% CI 0.82–1.09). Important to mention is that due to the small sample size in the subgroup analysis, the information should be interpreted carefully. The question "Do infants with atopic disease benefit from probiotics (Lactobacillus spp. and/or Bifidobacteria spp., Propionibacterium freudenreichii ssp. shermanii JS)?" should be tested in more heterogenetic, well-designed RCTs. Beneficial effects of specific strains might become lost by pooling probiotic strains together, since the effects are strain-specific. As a consequence, meta-analysis should be strain-specific. Due to the wide heterogeneity of strains, mixture and doses administered, the efficacy of specific probiotic strains has been difficult to analyze. Therefore, further research is needed to optimize the selection of probiotic strains and the configuration of intervention regimens [12].

#### *3.2. Probiotics for the Treatment of Asthma*

The curative effects of probiotics in asthmatics are not well established [40] (Table 2). A recent study in baby mice indicated that *Bifidobacterium infantis* could reduce the infiltration of inflammatory cells by promoting Th1 immune responses and oppositely suppressing Th2 immune responses [41]. In a 2008 systematic review, probiotic administration showed no positive effect in the treatment of asthma [42]. A later meta-analysis from Das et al., which included 12 studies, showed no enhancement in quality-of-life scores in asthmatic patients. However, probiotics were found to be efficacious in diminishing the amount of asthma attacks [43]. Altogether, the present evidence does not support use of probiotics in the treatment of asthma, although some studies suggest some benefit while harm was not reported [40].

#### *3.3. Probiotics for Prevention of Allergic Rhinitis*

The occurrence of perennial allergic rhinitis and seasonal allergic rhinitis has been rising globally and their management is costly [44] (Table 3). Currently, there is no strong proof that probiotics are successful in preventing allergic rhinitis [45]. Surprisingly, some studies suggest that there may even be an increased prevalence of allergic rhino-conjunctivitis in patients taking probiotics in the perinatal period and in childhood [46]. In a systematic review published in 2014, five RCTs that have studied the preventive role of probiotics in allergic rhinitis were assessed. Combining data from adults and children, no difference in incidence of allergic rhinitis between the probiotic and control groups (odds ratio (OR) 1.07, 95% CI, 0.81–1.42, *p* = 0.64, fixed-effects model), and no significant difference in the prevention of allergic rhinitis have been found [47]. A 2019 meta-analysis of seventeen RCTs including 5264 children could not identify a clear advantage of probiotic supplementation during pre- and postnatal periods in the prevention of allergic rhinitis [48]. In follow-up research of a previous study investigating the pre- and postnatal usage of probiotics in high-risk children between five and ten years of age, Peldan et al. sent surveys to their parents to investigate if atopic diseases, including allergic rhinitis, were present. The lifetime prevalence of allergic rhinitis was equal in both probiotic and placebo groups (35.2% vs. 41.7%, adjusted OR 0.74, 95% CI 0.55–1.00, *p* < 0.05); nevertheless, the prevalence of allergic rhino-conjunctivitis at five to ten years of age was greater in the probiotic than in the placebo group (36.5% vs. 29.0%, OR 1.43, 95% CI 1.06–1.94, *p* = 0.03) [46]. Following the

authors of this study, the question form may be biased since manifestations of viral rhinitis may be mistaken for allergic rhinitis [46]. After a follow-up of 11 years, the same negative outcome of no association (RR 0.85, 95% CI 0.65–1.1, *p* = 0.24) was found for probiotics *Lactobacillus rhamnosus* HN001 and *Bifidobacterium lactis HN019* taken by mothers every day from 35-week gestation to six months postnatally while breastfeeding and by infants from birth to two years of age [39]. However, similar to the prevention of asthma, the absence of evidence for a potential benefit may be due to shortcomings in study designs and the presence of multiple confounding variables. Probiotic intervention may have a favorable role in the prevention and additional treatment of allergic rhinitis, although results up to now are disappointing [49].

#### *3.4. Probiotics for Treatment of Allergic Rhinitis*

Avoidance of contact with allergens, medications to reduce symptoms to decrease inflammation and immunotherapy are standard approaches in the management of allergic rhinitis [50]. The question raised is if oral probiotics might modulate the microbiome in such a way that they result in an alteration of the immune system which would contribute to the treatment of allergic rhinitis [51] (Table 4). The development of allergic inflammation in a murine house dust mite asthma model is suppressed by synbiotic mixtures of nondigestible oligosaccharides and *Bifidobacterium breve* M-16V [52].

A review from 2010 (including seven trials, n = 616, children and adults mixed) suggested that probiotics (*Lactobacillus* spp. and *Bifidobacterium* spp.) contribute to a decrease in allergic rhinitis symptoms, quality of life and decrease the need for drug intake (standard mean difference (SMD) −1.17, 95% CI −1.47–0.86; *p* < 0.00001) [53]. Another metaanalysis performed in 2014 including 11 RCTs reported similar conclusions, as probiotics significantly improved both quality of life and nasal symptom scores (SMD −2.97, 95% CI, −4.77–1.16, *p* = 0.001). However, this was not associated with an improvement in immunologic variables [47]. This meta-analysis was criticized for its methodology [47,54]. A 2016 meta-analysis of 22 RCTs also came up with evidence of a potential benefit of probiotics, once more demonstrating improvement in quality of life. A clinically significant benefit was reported for at least one outcome in 17 studies, while no benefit could be shown in six trials. Improvement was mainly regarding quality of life (SMD −2.30, 85% CI −3.93 to −0.67, *p* = 0.006), while no effect was shown on rhinitis symptoms (SMD −0.34, 95% CI −0.62–0.07; *p* = 0.13) or total IgE levels (SMD 0.01, 95% CI −0.17–0.19, *p* = 0.88), and for antigen-specific IgE (SMD 0.09, 95% CI −0.44–0.62, *p* = 0.74) in the placebo group compared to the probiotic. Studies are characterized by a high degree of heterogeneity in probiotic strains tested, inclusion criteria and outcomes [55].

In 212 children under five-years-old from Pakistan, a probiotic product administered as a chewable tablet, containing two x 10<sup>9</sup> CFU of *Lactobacillus* Paracasei (LP-33), was administered once a day for six weeks while the control group was treated with cetirizine tablet 2.5 mg (<two years) or 5 mg (two-five years) once daily. Significant improvement from baseline symptoms (rhinorrhea, sneezing, nasal blocking, coughing, feeding difficulties and sleeping difficulties) was reported equally in both groups in almost all children [56]. Although the title of the paper mentions probiotics, the study was in fact performed with postbiotics since it was lyophilized extracts of bifidobacteria which were shown to suppress allergic rhinitis in mice via inducing IL-10-producing B cells [57]. Another study (with mice) showed that Clostridium butyricum extracts—again, postbiotics—can efficiently inhibit experimental allergic rhinitis by increasing IL-10 expression in B cells [58].

A pilot study in only 20 adult (18–65-years-old) patients with allergic rhinitis caused by house dust mite allergy suggests that probiotics-impregnated bed linen with five natural genetically unmodified bacterial probiotic strains of Bacillus species (strains of Bacillus subtilis, Bacillus amyloliquefaciens and Bacillus pumilus) reduces symptoms and increases quality of life [59]. A large-scale study is recommended to further investigate all these findings [59].


**Table 1.** Studies examining probiotics for prevention of asthma in humans.




*Nutrients* **2021**, *13*, 934

**Table 3.**

Studies

examining

 probiotics

 for

prevention

 of allergic rhinitis.

Lactobacillus;

 B:

Bifidobacterium


#### *3.5. Prebiotics for Prevention/Treatment of Asthma or Allergic Rhinitis*

Inulin, fructo-oligosaccharides and galacto-oligosaccharides are well known examples of prebiotics. Table 5 provides an overview of the literature. These substrates will contribute to the growth of two common bacteria in the gut-bifidobacteria and lactobacilli [60]. Some of the substrates interacting with the infant's gut microbiome are human milk oligosaccharides (HMOs) [61], which form the third biggest fraction in human milk [36]. In a mouse model, 2'-fucosyllactose and 6'-sialyllactose decrease the symptoms of food allergy due to the induction of IL-10(+) T regulatory cells and indirect stabilization of mast cells [62]. Prebiotics such as non-human galacto- and fructo-oligosaccharides have been added to infant formula to try to mimic the results of HMOs. However, these non-human prebiotics are less structurally diverse than HMOs [47]. An 18-year follow-up of high-allergy-risk breastfed infants was conducted to evaluate the relation between HMO profiles of the mother and the risk of developing asthma, eczema and sensitization. One HMO profile, namely the acidic Lewis HMOs, showed an increased risk of developing allergic disease and asthma in youth (OR 5.82, 95% CI 1.59–21.23) compared to the neutral Lewis HMO profile. Another finding of the study is that the acidic-predominant profile was associated with a lower risk of food sensitization (OR 0.08, 95% CI 0.01–0.67, *p* < 0.05). HMOs have only been recently available on the market; nevertheless, there are some studies investigating their effect on allergies [63]. A meta-analysis with two studies reporting early respiratory symptoms as outcome (n = 249) has examined if these non-human oligosaccharides have effects on allergy. The study found that infants who received prebiotics (non-human oligosaccharides) had reduced asthma or recurrent wheezing (RR 0.37, 95% CI 0.17–0.80, *p* < 0.01) [64]. Another double blinded RCT (n = 461) compared Chinese toddlers drinking standard milk formula with those drinking a formula containing bioactive proteins and/or the HMO 2 -fucosyllactose and/or milk fat, for a period of six months. In this study, however, no difference was found in the occurrence of upper respiratory infections. No analysis for allergy was conducted [65]. Concluding, there is still little evidence to use prebiotics for the prevention of asthma and none for allergic rhinitis to our knowledge on rhinitis. No studies have been conducted to analyze the effects of prebiotics as a treatment for asthma or allergic rhinitis.

#### *3.6. Synbiotics for Prevention/Treatment of Asthma or Allergic Rhinitis* 3.6.1. Asthma

The literature on synbiotics regarding prevention and/or treatment of allergic manifestations is still limited (Table 6). Some analyses do not differentiate between pre-, pro- and synbiotics. [66]. Ninety infants with atopic dermatitis were managed with a formula with extensively hydrolyzed protein and were included in a double-blind, placebo controlled multicenter trial for 12 weeks, randomized to the formula with or without synbiotics over a period of seven months. One year later, information regarding respiratory symptoms and asthma medication was collected with a questionnaire. The significant reduced prevalence of "frequent wheezing" and "wheezing and/or noisy breathing apart from colds" was observed in the synbiotic group (13.9% vs. 34.2%, absolute risk reduction (ARR) −20.3%, 95% CI −39.2% to −1.5%, and 2.8% vs. 30.8%, ARR −28.0%, 95% CI −43.3% to −12.5%, respectively). Additionally, the use of asthma medication was significantly lower (5.6% vs. 25.6%, ARR −20.1%, 95% CI −35.7% to −4.5%). However, total IgE levels did not differ. Increased specific cat-IgE levels were noticed in five children (15.2%) in the placebo group versus none in the synbiotic group (ARR −15.2%, 95% CI −27.4% to −2.9%). The outcome of this trial suggests that synbiotics may prevent asthma in infants presenting with atopic dermatitis [67]. However, the limited number of children included in this trial is a major limitation. Cabana et al. [68] performed an RCT in 92 infants with a mixture of LGG and inulin as synbiotic (in the study mentioned as probiotics) between birth and the age of six months of life in infants with mixed breast and formula feeding [68]. Asthma at the age of five years was a secondary outcome, but was not statistically different in both groups with an incidence of 17.4% in the control prebiotic and 9.7% in the symbiotic [68].


*Nutrients* **2021**, *13*, 934

mentioned

 or not applicable.


*Nutrients* **2021**, *13*, 934

**Table 6.**

Studies

examining

 synbiotics

 for

prevention/treatment

 of asthma.

FOS:

fructo-oligosaccharide.

A double-blinded, placebo-controlled RCT performed in Iranian children younger than 12 years tested the efficacy of synbiotic (Kidilact®: Streptococcus thermophilus, *Bifidobacterium* spp., *Lactobacillus* spp. zinc and fructo-oligosaccharide) asthma management. Multiple outcomes did not show a difference between both groups; the number of outpatient visits, 19 in the synbiotic versus 55 in the control arm (*p* = 0.001), was the only statistically significant difference [69].

#### 3.6.2. Allergic Rhinitis

The effect of synbiotics on prevention of allergic rhinitis will remain unanswered because no RCTs have been conducted yet to our knowledge (Table 7). Clinical symptoms and quality of life improve with immunotherapy, but synbiotics do not contribute to this improvement.

**Table 7.** Studies examining synbiotics for prevention/treatment of allergic rhinitis.


Legend. #: number; RCT, randomized controlled trial; spp., species; cfu, colony-forming unit; CI, confidence interval; RR, relative risk; L: Lactobacillus; B: Bifidobacterium; S; Stretococcus; FOS; fructo-oligosaccharide.

> Synbiotics in the treatment of allergic rhinitis are also poorly studied, although some of the trials reporting on the efficacy of probiotics, in fact, concern synbiotics [37]. A placebo-controlled, double-blind RCT in a small number of children and adults (*n* = 20, age nine-53 years) in Iran showed that immunotherapy and a synbiotic (Streptococcus thermophilus, *Bifidobacterium* spp., *Lactobacillus* spp., fructo-oligosaccharide) reduced the gene expression of IL-17 after two and six months (*p* = 0.001, *p* = 0.0001) more compared to the group receiving immunotherapy and a placebo [70]. Other probiotics [71] were also shown to reduce cytokine IL-17 by directly and indirectly downregulating and suppressing the T helper 17 subset. A 2019 crossover RCT (*n* = 152 subjects (30.1 ± 7.6 years) in adults in Iran showed that adding synbiotics (however, in the study, mentioned as probiotics) to budesonide significantly ameliorated quality of life in persistent allergic rhinitis patients (*p* < 0.05 for social functioning and *p* < 0.001 for mental health and vitalism) [37]. The patient population used in this study may not be representative for allergic rhinitis patients in the

overall population, since symptoms did not respond to usual therapy with antihistamines, antileukotrienes, decongestants and nasal steroids [51].

More well-designed studies, investigating only the effects of synbiotics for allergy prevention and/or treatment, are needed [36].

#### **4. Conclusions**

Meta-analyses have showed marked heterogeneity as well in inclusion criteria, studied products and primary outcomes between studies, making direct comparison hazardous. Today, the American Academy of Pediatrics, the European Academy of Allergy and Clinical Immunology, the National Institute of Allergy and Infectious Disease, and the European Society for Pediatric Gastroenterology, Hepatology and Nutrition do not recommend the use of probiotics for primary prevention of allergic disease [13]. The lack of evidence is the consequence of large heterogeneities between study designs, differences in strains, and dosages and duration of probiotics administered. Future research may clarify these issues [35]. Data from laboratory research in well-controlled conditions demonstrate an important role for gastrointestinal microbiota composition on the development of allergic disease in the respiratory tract, suggesting even a causal relation. Data from clinical human studies remain disappointing. The multiple confounding variables in the clinical situation, therefore, illustrate the impact of environmental and other variables on the development of allergic disease. Overall, we have to conclude that the evidence is insufficient to recommend administration of pro-, pre- or synbiotics in the prevention or treatment of respiratory tract allergies. However, adverse effects are not reported. Additionally, data obtained in controlled situations suggest benefits. Future research requires thoughtful development of appropriate study design according to internationally set standards to ensure uniformity [72]. The modes of action of pro-, pre- and synbiotics need to be further clarified in health and disease [40].

**Author Contributions:** Conceptualization, L.M., E.I.L. and Y.V.; methodology, L.M., E.I.L.; software, L.M., E.I.L.; validation, L.M., E.I.L., Y.V.; formal analysis, L.M., E.I.L.; investigation, L.M., E.I.L.; resources, L.M., E.I.L.; data curation, L.M., E.I.L.; writing—original draft preparation, L.M., E.I.L.; writing—review and editing, L.M. E.I.L.; visualization, L.M., E.I.L.; supervision, E.I.L, Y.V.; project administration, L.M., E.I.L.; funding acquisition, none. Please turn to the CRediT taxonomy for the term explanation. Authorship must be limited to those who have contributed substantially to the work reported. All authors have read and agreed to the published version of the manuscript.

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

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** L.M. and E.I.L. declare no conflict of interest. Y.V. has participated as a clinical investigator, and/or advisory board member, and/or consultant, and/or speaker for Abbott Nutrition" ByHeart, CHR Hansen, Danone, ELSE Nutrition, Friesland Campina, Nestle Health Science, Nestle Nutrition Institute, Nutricia, Mead Johnson Nutrition, Phathom Pharmaceuticals, United Pharmaceuticals, Wyeth.

#### **References**

	- Asthma and Immunology and the European Academy of Allergy and Clinical Immunology. *J. Allergy Clin. Immunol.* **2017**, *139*, 1099–1110. [CrossRef] [PubMed]

## *Article* **Levels of Predominant Intestinal Microorganisms in 1 Month-Old Full-Term Babies and Weight Gain during the First Year of Life**

**Sonia González 1,2, Marta Selma-Royo 3, Silvia Arboleya 2,4, Cecilia Martínez-Costa 5,6, Gonzalo Solís 7,8, Marta Suárez 7,8, Nuria Fernández 2,9, Clara G. de los Reyes-Gavilán 2,4, Susana Díaz-Coto 10, Pablo Martínez-Camblor 11, Maria Carmen Collado 3,\* and Miguel Gueimonde 2,4,\***

	- greyes\_gavilan@ipla.csic.es (C.G.d.l.R.-G.)
	- <sup>3</sup> Institute of Agrochemistry and Food Technology (IATA-CSIC), 46980 Paterna, Spain; mselma@iata.csic.es
	- <sup>4</sup> Department of Microbiology and Biochemistry of Dairy Products, Instituto de Productos Lácteos de Asturias (IPLA-CSIC), 33300 Asturias, Spain
	- <sup>5</sup> Department of Pediatrics, School of Medicine, University of Valencia, 46010 Valencia, Spain; cecilia.martinez@uv.es
	- <sup>6</sup> Pediatric Gastroenterology and Nutrition Section, INCLIVA Research Center, Hospital Clínico Universitario Valencia, 46010 Valencia, Spain
	- <sup>7</sup> Pediatrics Service, Hospital Universitario Central de Asturias, SESPA, 33004 Oviedo, Spain; gsolis@telefonica.net (G.S.); msr1070@hotmail.com (M.S.)
	- <sup>8</sup> Pediatrics Research Group, Instituto de Investigación Sanitaria del Principado de Asturias (ISPA), 33004 Oviedo, Spain
	- <sup>9</sup> Pediatrics Service, Hospital de Cabueñes, SESPA, 33201 Gijón, Spain
	- <sup>10</sup> Department of Statistics, University of Oviedo, 33004 Oviedo, Spain; UO266718@uniovi.es
	- <sup>11</sup> Department of Anesthesiology, Geisel School of Medicine at Dartmouth, Dartmouth, NH 03756, USA; Pablo.Martinez-Camblor@hitchcock.org

**Abstract:** The early life gut microbiota has been reported to be involved in neonatal weight gain and later infant growth. Therefore, this early microbiota may constitute a target for the promotion of healthy neonatal growth and development with potential consequences for later life. Unfortunately, we are still far from understanding the association between neonatal microbiota and weight gain and growth. In this context, we evaluated the relationship between early microbiota and weight in a cohort of full-term infants. The absolute levels of specific fecal microorganisms were determined in 88 vaginally delivered and 36 C-section-delivered full-term newborns at 1 month of age and their growth up to 12 months of age. We observed statistically significant associations between the levels of some early life gut microbes and infant weight gain during the first year of life. Classifying the infants into tertiles according to their *Staphylococcus* levels at 1 month of age allowed us to observe a significantly lower weight at 12 months of life in the C-section-delivered infants from the highest tertile. Univariate and multivariate models pointed out associations between the levels of some fecal microorganisms at 1 month of age and weight gain at 6 and 12 months. Interestingly, these associations were different in vaginally and C-section-delivered babies. A significant direct association between *Staphylococcus* and weight gain at 1 month of life was observed in vaginally delivered babies, whereas in C-section-delivered infants, lower *Bacteroides* levels at 1 month were associated with higher later weight gain (at 6 and 12 months). Our results indicate an association between the gut microbiota and weight gain in early life and highlight potential microbial predictors for later weight gain.

**Keywords:** infants; microbiota; *Staphylococcus*; *Enterococcus*; *Bifidobacterium*; weight gain

**Citation:** González, S.; Selma-Royo, M.; Arboleya, S.; Martínez-Costa, C.; Solís, G.; Suárez, M.; Fernández, N.; de los Reyes-Gavilán, C.G.; Díaz-Coto, S.; Martínez-Camblor, P.; et al. Levels of Predominant Intestinal Microorganisms in 1 Month-Old Full-Term Babies and Weight Gain during the First Year of Life. *Nutrients* **2021**, *13*, 2412. https://doi.org/ 10.3390/nu13072412

Academic Editor: Nadja Haiden

Received: 15 June 2021 Accepted: 13 July 2021 Published: 14 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

#### **1. Introduction**

From birth, and initially depending on the delivery mode, the neonatal gut is colonized by a rapidly diversifying microbiota, reaching an adult-type microbiota around 3–5 years of life. During early life, other perinatal factors, such as feeding practices, environment or antibiotic treatments, also contribute to shaping the microbiota development [1]. Current evidence supports the role of this early microbiota in promoting and maintaining a balanced immune response and adequate brain development and, subsequently, in the future health of the infant [2,3]. Induction of early microbiota alterations by antibiotics use has been linked to allergic diseases [4], obesity [5], risk of colorectal cancer [6] and other potential noncommunicable diseases (NCDs) later in life [7,8]. These studies underline the importance of the early life microbiota as a key driver for adequate infant development and later health. Moreover, recent evidence indicates that altering this early microbiota may also have long-lasting effects on body weight and weight gain in childhood and on the later risk of obesity during adulthood [9–13]. Indeed, higher birth weight and rapid growth during early life have been linked to increased risk of overweight and obesity during childhood and adulthood [14–18]. Interestingly, a recent study reported the very early life microbiota which is present in meconium or first-pass neonatal samples as a predictor of infant overweight by the age of 2 years [19]. Early microbiota composition has also been linked to overweight and obesity at later infancy [20,21]. On the other side, other recent studies have highlighted the effect of antibiotic treatment on infant growth and development during the first 6 years of life [12].

In this context, the potential relationship between early microbiota and weight gain is of great interest since this relationship offers opportunities for the microbiota-mediated modulation of weight gain [22] and/or the prevention of growth impairment [23]. Recently, some studies have assessed the potential association between early microbiota and weight gain in preterm infants [24,25]; however, data on full-term babies are still scarce.

In this study, we evaluated the association between the early life microbiota and the later weight gain in both vaginally and C-section-delivered healthy full-term infants. We aimed at identifying if the levels of selected microorganisms at 1 month of age were related to later weight gain during the first year of life in these two groups of infants. With this goal, we used quantitative PCR (qPCR) for assessing specific microbial groups in the infant's feces at the age of 1 month and monitored weight gain during the first year of life.

#### **2. Subjects and Methods**

#### *2.1. Recruitment and Sampling*

A total of 124 infants born after a full-term, uncomplicated pregnancy by vaginal delivery (*n* = 88) or by C-section (*n* = 36) were recruited at the neonatology units of the University Central Hospital of Asturias (Oviedo, Northern area, Spain) and the University Clinic Hospital of Valencia (Valencia, Mediterranean area, Spain). Inclusion criteria were no metabolic (obesity, diabetes) or chronic diseases and no probiotics consumption by mothers during late pregnancy or infants' early life and no antibiotics administration to the infants. Clinical data such as gestational age or perinatal maternal antibiotics were recorded, as well as neonatal weight and height at birth, at 1, 6 and 12 months of life, the change in weight (weight gain), Z-scores for weight, height and weight-for-height were calculated for each infant at each time point. A fresh fecal sample was collected at 1 month of age and immediately frozen for later microbiota analyses.

Families received detailed study information and signed an informed consent form. The study recruitment and sampling have been approved by the Regional Ethics Committee of Clinical Research of Asturias (Ref. 12/16. 3 February 2016), the Ethics Committee of the Hospital Clínico Universitario de Valencia INCLIVA (Ref. 9 January 2015) and the Committee on Bioethics of CSIC. The procedures were performed in accordance with the fundamental principles set out in the Declaration of Helsinki, the Oviedo Bioethics Convention, the Council of Europe Convention on Human Rights and Biomedicine and the Spanish legislation on bioethics. The Directive 95/46/EC of the European Parliament and

the Council of 24 October 1995, on the protection of individuals regarding the processing of personal data and on the free movement of such data were (and will be) strictly followed.

#### *2.2. Fecal Microbiota Analyses*

Total DNA was isolated from all fecal pellets as described previously by using the QiAGEN Stool Kit (QIAgen. Hilden. Germany) [26]. The extracted DNA was then used for quantifying fecal levels of the Enterobacteriaceae family, the *Bacteroides*-group and the genera *Bifidobacterium*, *Enterococcus*, *Staphylococcus* and *Lactobacillus* group by quantitative PCR using previously described primers, conditions and standard cultures (Table 1). PCRs were performed either in a LightCycler® 480 Real-Time PCR System (Roche®) or a 7500 Fast Real-Time PCR system (Applied Biosystems) by using the SYBR Green. A subgroup of samples (*n* = 33) was analyzed in both machines to ensure comparability, without detecting statistically significant differences between the data obtained in each of them (data not shown), with both machines showing high correlation (Pearson's correlation coefficients ranging between 0.785 and 0.942 depending in the primer pair used).


**Table 1.** Primers and PCR conditions for the different primer pairs used in this study.

#### *2.3. Anthropometrical Determinations*

Child height and weight were recorded to the nearest 0.1 cm and 0.1 kg, respectively, through standardized procedures by pediatric nurse and registered at birth, 1, 6 and 12 months. With this information and the date of birth, Z-score was calculated by using WHO ANTHRO, Software for Calculating anthropometry, Version 3.2.2 (https://www.who.int/ childgrowth/software/es/; accessed on 14 July 2021). The WHO Child Growth Standards provide child growth measures standardized by age and sex using Z-score.

#### *2.4. Statistical Analyses*

For statistical analysis, the free software R (https://www.r-project.org; accessed on 7 June 2021) was used. Variables are described by mean and standard deviations, median and percentiles or by counts and frequencies. Student–Welch and Chi-square tests were used for checking the mean and distributions equalities, respectively. Pearson correlation coefficients were used for studying the association between continuous variables. A heatmap, generated in R package using the heatmap.2 application in ggplots package [33] was employed for summarizing the analyses. Comparisons on bacterial levels among the different infants' groups were achieved by using a *t*-test with Bonferroni's correction. Multiple mixed linear models were used for studying the effect of the microbial levels at one month and the weight gain, weight, height and Z-scores at 1, 6 and 12 months. These models were used unadjusted and after adjusting for potential confounders in both groups of infants (vaginally or C-section-delivered). Backward stepwise analyses based on the Aikaike Information Criterion (AIC) were employed to determine whether the variables were included in a potential predictive model. A forest plot was used to show the effect

sizes (with 95% confidence intervals) in both the so-labeled univariate and the multivariate models in both groups of infants (adjusting by infant gender and feeding type). Time variation in weight was determined according to the tertile classification of each of the microbial groups analyzed in this study. For this purpose, the cut-off points established were: 1) for vaginally delivered babies; *Bacteroides* group (T1 < 6.72; T2 6.72–8.59; T3 > 8.59); *Bifidobacterium* (T1 < 8.37; T2 8.37–8.98; T3 > 8.98); Enterobacteriaceae (T1 < 7.79; T2 7.79–8.60; T3 > 8.60); *Enterococcus* (T1 < 6.52, T2 6.52–7.60, T3 >7.60); *Lactobacillus* group (T1 < 5.60, T2 5.60–6.75, T3 > 6.75); *Staphylococcus* (T1 < 5.94, T2 5.94–6.75, T3 > 6.75) and 2) for C-section-delivered babies; *Bacteroides* group (T1 < 6.48; T2 6.48–7.30; T3 > 7.30); *Bifidobacterium* (T1 < 7.78; T2 7.78–8.79; T3 > 8.79); Enterobacteriaceae (T1 < 6.95; T2 6.95– 8.42; T3 > 8.42); *Enterococcus* (T1 < 6.58, T2 6.58−7.92, T3 > 7.92); *Lactobacillus* group (T1 < 5.23, T2 5.23−6.54, T3 > 6.54); *Staphylococcus* (T1 < 5.40, T2 5.40–6.77, T3 > 6.77). *p*-values below 0.05 were considered statistically significant.

#### **3. Results**

#### *3.1. General Description of the Population*

The 124 full-term babies (55 males/69 females) included in this study were born at gestational ages ranging from 37 to 41 weeks (mean 39.6). Of these, 88 babies were delivered vaginally (birth weights between 2135 and 4800 g) and 36 by C-section (birth weights between 2215 and 4690 g). There were no statistically significant differences in mean weight according to delivery mode (mean weight of 3189 vs. 3215 for vaginal and C-section babies, respectively). Fifty-six of the infants born vaginally were exclusively breastfed, whereas 31 babies received formula or mixed feeding at the age of 1 month. In C-section babies, the proportion of children receiving each of these feeding types was 50 percent. Female babies showed a significantly higher rate of vaginal delivery than males (80% vs. 58%, *p* = 0.019), whereas no differences in feeding habits were observed between boys and girls.

#### *3.2. Gut Microbiota Composition and Weight Gain Are Affected by Different Variables*

In this study, the main microbial phyla representatives were quantified by using groupspecific qPCR methods. As expected, the microbiota of vaginally delivered babies was different from that of C-section ones, with significantly (*P* < 0.05) higher levels of *Bacteroides*group of microorganisms (7.65 ± 1.42 vs. 6.74 ± 0.98 Log nº cells/g, respectively) and *Bifidobacterium* (8.52 ± 0.76 vs. 8.05 ± 1.02) in the former group. No differences between both groups of infants were observed for any of the other microbial groups analyzed (Enterobacteriaceae, 8.07 ± 1.11 vs. 7.66 ± 1.38; *Enterococcus*, 6.96 ± 1.38 vs. 6.97 ± 1.56; *Lactobacillus*, 6.08 ± 1.52 vs. 5.79 ± 1.58; *Staphylococcus*, 5.94 ± 1.56 vs. 5.84 ± 1.60). These differences in the levels of some of the microbial groups analyzed between both delivery mode groups prompted us to consider them as two different groups and analyze them separately.

In both groups of 1 month-old infants, the genus *Bifidobacterium* was the bacterial group showing the highest levels, followed by members of the Enterobacteriaceae family and *Bacteroides*-group (Table 2). Interestingly, no differences in bacterial levels were observed between 1 month-old males and females neither in vaginally delivered nor in C-section-delivered babies. Concerning infant feeding practices, exclusive breastfeeding was found to be associated with reduced levels of enterococci at 1 month of age compared to formula/mixed feeding; the differences reaching statistical significance (*P* < 0.05) in vaginally delivered babies (Table 2).


**Table 2.** Levels (Log nº cells/g) of some relevant bacterial groups in fecal samples of the vaginally delivered or C-sectiondelivered infant population included in this study, categorized by feeding type, gender and mode of delivery.

All values are shown as mean ± standard deviation. EBF, exclusive breastfeeding; MF, formula/mixed feeding. There is a missing value in feeding type (*n* = 87). \* Denotes statistically significant differences (*p* ≤ 0.05) between genders or feeding types within the same delivery group. \$ Denotes statistically significant differences (*<sup>p</sup>* ≤ 0.05) for infants from the same gender or feeding type between the two delivery groups (vaginally delivered or C-section-delivered).

> As expected, when analyzing the anthropometric parameters in the sample (Table 3), statistically significant differences were found between both genders, with body weight and height being higher in males. Moreover, C-section-delivered infants on formula/mixedfeeding showed a significantly lower birth weight and weight and height by the age of 1 month (*P* = 0.022) without noticing statistically significant differences at a later age. Zscores showed statistically significant differences in weight for height at 1 and 6 months but not at 12 months of age, and no other statistically significant differences in Z-scores were obtained between vaginally delivered and C-section babies (Supplementary Table S1).

**Table 3.** Weight and weight gain during the first year of life in the infants included in this study as categorized by feeding type, gender and delivery mode.


All values are shown as mean ± standard deviation. EBF, exclusive breastfeeding; MF, formula/mixed feeding. \* Denotes statistically significant differences (*<sup>p</sup>* ≤ 0.05) between gender or feeding types within the same delivery group. \$ Denotes statistically significant differences (*p* ≤ 0.05) for infants from the same gender or feeding type between the two delivery groups (vaginally delivered or C-sectiondelivered).

#### *3.3. Gut Microbial Groups at 1 Month Are Associated with Weight Gain*

The analysis of Pearson correlation coefficients pointed out different associations between microbes and infant growth depending on the mode of delivery (vaginal and C-section-delivered babies). In vaginally delivered infants, the family Enterobacteriaceae was the microbial group showing more correlations with the infant's growth variables (Figure 1). A significant positive association was observed between the levels of these microorganisms at 1 month and Z-score birth weight, weight at 1 month, Z-score weight at 1 month, Z-score weight for height at 1 month and Z-score weight at 6 months (Figure 1). Similarly, in this group of infants, the levels of *Staphylococcus* showed a significant positive

association with weight and Z-score for weight at 1 month of age. In C-section-delivered babies, the only significant correlations observed were the negative association between the levels of *Bacteroides* at 1 month and weight and height (as raw measures and as Z-scores) at the age of 6 months, and the direct association between levels of enterocci and weight gain at 6 months (Figure 1). Although no other statistically significant differences were obtained, the data indicate different interactions between bacteria and infant development depending on the delivery mode; the levels of some microorganisms at 1 month of age, such as *Bacteroides* or *Staphylococcus*, showed a clearly different pattern in vaginally and C-section-delivered infants.

**Figure 1.** Heatmap showing the pairwise Pearson correlation coefficients between microbial groups at 1 month of age and the analyzed growth variables for both vaginally delivered and C-section-delivered babies. \* *p* < 0.05.

> To gain further insight into these associations, infants were classified according to the tertiles of the levels of the different microorganisms analyzed, and the variations on body weight, along the first year of life, were compared among these tertiles (Figure 2). No statistically significant differences were observed on the evolution of weight during the first 12 months of life among the tertiles for the fecal levels of *Bacteroides*, *Bifidobacterium*, Enterobacteriaceae, *Enterococcus* and *Lactobacillus*, neither in vaginally delivered nor in C-section babies. However, C-section children classified according to the tertiles obtained for the levels of *Staphylococcus* showed statistically significant differences in their weight trajectory (Figure 2). C-section infants harboring high levels of staphylococci at 1 month

of age (upper tertile) displayed a significantly lower weight at 1 year of age, with this phenomenon not being observed in vaginally delivered babies.

**Figure 2.** Long-term variations in body weight for the tertiles according to fecal levels of the *Staphylococcus* at 1 month of age in vaginally delivered or C-section-delivered babies (*n* = 122) (T1, tertile 1; T2, tertile 2; T3, tertile 3). \* *p* ≤ 0.05.

Hereafter, uni- and multivariate regression models were performed for a deeper assessment of the association between early microbiota and weigh-gain in both groups of infants (Figure 3). To take into consideration the potential effects of gender and feeding type, the models were controlled for these two variables, and the relationship between microbiota at 1 month of age and infant weight gain at 1, 6 and 12 months of age was assessed (Figure 3). Different effects were observed between both groups of infants. A significant positive effect of the levels of *Staphylococcus* at 1 month of age on weight gain at 1 month was obtained in both the unadjusted (*p* = 0.016) and adjusted (*p* = 0.036) models in vaginally delivered babies, but these do not reach significance in C-section-delivered infants. On the contrary, a negative association of *Bacteroides* levels at 1 month of age with weight gain at 6 and 12 months was observed in the C-section group (*p* = 0.007 in unadjusted and *p* = 0.014 in the adjusted model at 6 months of age, and *p* = 0.031 and *p* = 0.052, respectively, at 12 months of age). The other microbial groups analyzed did not show any statistically significant effect.

**Figure 3.** Forrest plots showing the effect sizes (with 95% confidence interval) of the association of microbiota-related variables with infant weight gain at 1, 6 and 12 months of age according to delivery mode (vaginal delivery (**A**) and C-section (**B**)). Results from unadjusted model and adjusted by gender and type of feeding. Dotted lines represent the zero value, with values on the left side indicating negative associations and those on the right side indicating associations with positive sign.

#### **4. Discussion**

The levels of the different microbial groups analyzed in this study were in line with those previously described for 1 month-old full-term infants [26,34,35]. Additionally, in accordance with previous studies, bifidobacteria was the bacterial group showing the highest levels, followed by enterobacteria, which is another of the dominant microbial groups in un-weaned infants [26,34]. Additionally, as expected [1], differences in the microbial levels were observed between vaginal and C-section-delivered babies.

Gender-associated differences in the infant microbiota composition and diversity have been previously reported [36,37]; however, in the present work, we did not notice any significant differences in the levels of the analyzed microbial groups between males and females, neither in vaginally nor in C-section-delivered babies. This is one aspect that deserves further attention since understanding the potential gender differences in the microbiome, the so-called microgenderome [38], and the role that these play in the risk of disease is of utmost importance for developing microbiota modulation strategies in early life. It must be taken into consideration that early life constitutes a critical moment. Some studies have reported an association between antibiotic treatments during the postnatal period microbiota [34,39] and an increased risk of obesity and related metabolic disorders [40,41]. This suggests a possible influence of microbiota alterations during this period in obesity risk later in life, as it has been demonstrated in animal models [42].

Moreover, some studies have also reported associations between early microbiota and weight gain [24,37,43]. However, due to the growing evidence linking the microbiota in early life to obesity risk, we consider that studies focused on full-term infants, as the present one, are especially relevant in this context. In this regard, previous studies demonstrated an altered microbiota during the first year of life in infants developing obesity later on [20], pointing out at the first months of life as the key moment for later metabolic homeostasis.

Interestingly, some microorganisms, such as *Staphylococcus* or *Enterococcus*, have been previously reported to be negatively associated with infant weight/weight gain in preterm infants during very early life [24]. The levels of these microorganisms were also found to be lower, at 5 and 9 months of life, in excessive weight gaining full-term infants than in those showing an appropriate weight gain [43]. However, some differences among studies are also present, likely due to the different methodologies and experimental designs used; for instance, we analyzed the microbiota at 1 month of age, whereas others analyzed it at a later stage (5 and 9 months of age) [43], and we segregated the analyses by delivery mode whilst other authors did not. Actually, our results indicate the existence of different interactions in vaginally delivered and in C-section-delivered babies. We observed that changes in the sign of the microbe-weigh association might occur along different sampling times, as evidenced by our data on staphylococci, showing a positive association with weight gain at 1 month of age but not at later ages when the interaction seems to be even negative. Interestingly, in C-section-delivered babies, but not in vaginal infants, the levels of the *Bacteroides*-group at 1 month of age correlated negatively with later weight, even when the model was adjusted by feeding mode. Delayed colonization by this microorganism has been often reported in C-section-delivered babies [44,45] and C-section delivery has been linked to increased risk of childhood obesity [46]. These observations point out at the levels of *Bacteroides* during early life as a potential early marker for the later risk of excessive weight gain in C-section-delivered babies, an aspect that should be the subject of further studies.

It is important to point out that different factors may influence infant growth trajectories. Among these, infant feeding habits may be of importance. Our results showed that exclusive breastfeeding was associated with significantly lower levels of *Enterococcus* in vaginally delivered babies and with a trend (non-statistically significant) also observed in the C-section-delivered group. In the former group, a trend towards higher *Staphylococcus* levels was also observed. These two microorganisms have been linked to the feeding pattern of the infant. Breastmilk has been previously described as a source of *Staphylococcus*, with increased levels of this microorganism being found in breastfed babies [47]. Other studies, in accordance with our results, reported lower levels of *Enterococcus* in breastfed infants [48]. Altogether, these results suggest that the observed differences in microbial groups and weight gain may be partly related to the feeding habit of the infant. However, although the feeding habit is likely an important factor, our multivariate models were corrected for this variable and some of the effects still remained significant, indicating a microbiota–weight association independent of the feeding type. Therefore, the microbiota– host relation needs to be considered in the analyses focused on infant growth trajectories in

order to shed light on the influence of this relationship for child development. Once this relationship is fully understood, it may be possible to develop nutritional strategies, such as dietary probiotics or prebiotics targeting the infant, or perhaps the pregnant or lactating mother, for modulating early life microbiota and the later infant weight gain.

It is also important to underline that our sample size is still limited for establishing strong general conclusions, especially in a context where several potential confounder factors may be present, as is the case in infant microbiota studies. However, it is also true that the infants included originated not just from a unique hospital and geographical location, which could be a source of bias, but from two distant locations. It is worth pointing out as well that our microbiota data are restricted to defined microbial groups for which absolute levels were determined and the potential influence of other microorganisms may have been overseen.

#### **5. Conclusions**

This work is among the first ones assessing the relationship between the absolute levels of relevant early life intestinal microorganisms, such as bifidobacteria, enterobacteria, lactobacilli, enterococci or staphylococci, and the later weight gain in either vaginally or C-section-delivered full-term infants. The data point out the relationship between specific infant gut microbes and healthy infant development. Our results underline the interest in exploring the intestinal microbiota as a potential target for favoring proper growth and weight gain in infants with potential consequences in later health.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/nu13072412/s1, Table S1: General description of the WHO Z-scores in the sample across time by type of partum and feeding.

**Author Contributions:** S.G., M.S.-R., S.A., C.M.-C., G.S., M.S., N.F., C.G.d.l.R.-G., S.D.-C., P.M.-C., M.C.C. and M.G. wrote sections of the first draft, thoroughly edited the manuscript and approved the final draft. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by the EU Joint Programming Initiative—A Healthy Diet for a Healthy Life (JPI HDHL. http://www.healthydietforhealthylife.eu/; accessed on 14 July 2021. Project EarlyMicroHealth) and the Project AGL2017-83653R funded by the Spanish "Ministerio de Ciencia, Innovación y Universidades (MCIU), Agencia Estatal de Investigación (AEI) and FEDER" and by the European Research Council under the European Union's Horizon 2020 research and innovation program (ERC starting grant, no. 639226). Silvia Arboleya is the recipient of a Juan de la Cierva Postdoctoral Contract from the Spanish Ministry of Science and Innovation (Ref. IJCI-2017-32156) funded by the Spanish Ministry of Science, Innovation and Universities.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board by the Regional Ethics Committee of Clinical Research of Asturias (Ref. 12/16. 3 February 2016), the Ethics Committee of the Hospital Clínico Universitario de Valencia INCLIVA (Ref. 9 January 2015) and the Committee on Bioethics of CSIC.

**Informed Consent Statement:** Informed consent was obtained from the paernts of all the infants involved in the study.

**Data Availability Statement:** The data are available upon reasoned request to the authors, under the restrictions established by the ethical approval of the study.

**Acknowledgments:** We thank all the families involved in the study.

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

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Nutrients* Editorial Office E-mail: nutrients@mdpi.com www.mdpi.com/journal/nutrients

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34

www.mdpi.com

ISBN 978-3-0365-5364-1