**E**ff**ects of Selenium- and Zinc-Enriched** *Lactobacillus plantarum* **SeZi on Antioxidant Capacities and Gut Microbiome in an ICR Mouse Model**

## **Sini Kang 1, Rui Li 1, Hui Jin 1, Hyun Ju You 2,\* and Geun Eog Ji 1,3,\***


Received: 28 September 2020; Accepted: 19 October 2020; Published: 21 October 2020

**Abstract:** Selenium and zinc are essential trace minerals for humans with various biological functions. In this study, selenium- and zinc-tolerant lactic acid bacteria (LAB) isolates were screened out from human fecal samples. Amongst three hundred LAB isolates, the *Lactobacillus plantarum* SeZi strain displayed the tolerance against selenium and zinc with the greatest biomass production and bioaccumulation of selenium and zinc. To further assess the characteristics of this strain, the lyophilized *L. plantarum* SeZi were prepared and administered to Institute of Cancer Research (ICR) mice. The mice were divided into four groups, provided with normal chow (Con), or normal chow supplemented with Na2SeO3 and ZnSO4·7H2O (SZ), *L. plantarum* SeZi (Lp), or selenium- and zinc-enriched *L. plantarum* SeZi (SZ + Lp), respectively. After 4 weeks of oral administration, the concentrations of selenium and zinc in blood were significantly increased in the SZ + Lp group when compared to the control or SZ group (*p* < 0.05). The increased selenium level led to an enhanced glutathione peroxidase activity and decreased blood malondialdehyde level in the SZ + Lp group (*p* < 0.05). Meanwhile, the results of bacterial community and microbial metabolic pathway analysis via 16S rRNA gene amplicon sequencing showed that *L. plantarum* SeZi significantly promoted the utilization of selenocysteine, seleno-cystathionine and seleno-methionine in the selenocompounds metabolism. Here, the in vivo antioxidant capacities of the selenium- and zinc-enriched lactobacillus strain showed us the utilization of a unique probiotic as a Se/Zn supplement with high availability, low toxicity, and additional probiotic advantages.

**Keywords:** selenium; zinc; bioaccumulation; antioxidant capacities; gut microbiota

## **1. Introduction**

Micronutrient deficiencies, known as "hidden hunger", have affected more than 50% of the world's population [1]. Selenium (Se) is a vital trace element, contributing to modulation of growth, regulation of antiviral capacity, and prevention of disease, especially cancer and cardiovascular disease [2–4]. The antioxidant activity of selenium is exhibited as a form of selenoenzymes, including glutathione peroxidase (GSH-Px), selenoprotein P, thioredoxin reductase, and methionine sulfoxide reductase [5]. Selenium deficiency can trigger serious health issues such as poor growth, muscle pain, decreased immune responses, and hypofunction of glandula thyreoidea [6,7]. Besides, the deficit of selenium is associated with a cardiomyopathy named as Keshan disease (KD) and an osteoarthropathy

named as Kashin-Beck disease (KBD) [8]. Although KD and KBD are just local problems primarily in China and east Serbia due to environmentally low selenium [9], the hypothyreose or insuline secretion impairments associated with the lack of selenium have a global impact. Zinc (Zn) is another essential micronutrient for humans. It is a key component of many metalloenzymes (i.e., superoxide dismutase (SOD), carbonic anhydrase, alcohol dehydrogenase) associated with human growth, immunity, fertility, and reproduction [10]. Additionally, zinc is significant for the correct secretion of hormone isoline by pancreas [11]. The chronic deficiency of zinc can lead to glucose intolerance and pre-diabetic syndromes [12,13]. On the other hand, zinc chronic overdose, which can be caused by some nutraceuticals, might be responsible for some neurodegenerations such as nervus opticus inflammation [14,15].

Inorganic selenium, such as selenate (SeO4 <sup>2</sup>−) and selenite (SeO3 <sup>2</sup>−), are toxic and poorly bioavailable [16]. The reduction of selenium oxyanions largely depends on biotic processes by microorganisms [17]. The utilization of microorganisms as the natural adsorbent for metal ions (i.e., selenium and zinc) is eco-friendly and cost-effective [18]. The bioabsorption capacities are attributed to their intrinsic biochemical and structural properties of the cellular membrane [19]. Lactic acid bacteria (LAB), as important food-grade bacteria with probiotic advantages, have been extensively studied in this field. The selenium concentration in the medium is highly linked to bacterial selenium level, but the growth of most bacterial isolates from the human gut can be inhibited by the addition of inorganic selenium into the medium [20]. However, some LAB strains have been reported to be capable of resisting selenium oxyanions at high concentrations during cultivation [21–23]. Especially, *Lactobacillus plantarum* has been suggested as Se-enriched lactobacilli for food applications [24]. Few studies about Zn-enriched LAB have been conducted, but it has been found that the bacterial growth and probiotic effect of *L. plantarum* can be enhanced by zinc in the gut [25].

Although the resistances of LAB to selenium and zinc have been reported, the in vivo antioxidant capacities of the SeZn-enriched probiotic products have not been reported. In this study, SeZn-tolerant LAB strains isolated from human feces were screened out to further investigate the effects on selenium and zinc bioaccumulation and related metabolism, antioxidant activities, and compositional changes of intestinal microbiota in vivo in an Institute of Cancer Research (ICR) mouse model.

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

### *2.1. Isolation of Probiotic Strains from Human Feces*

According to the protocol approved by the Institutional Review Board of Seoul National University (IRB No. 1702/002-013), fresh fecal samples were obtained from five children (1–6 years old) in Korea and stored at 4 ◦C during transportation. Each fecal sample (1 g) was serially diluted with a sterilized phosphate buffered saline solution (pH 7.4). The suspension was plated on *Lactobacillus* Selection (LBS) agar (Difco, Sparks, MD, USA) to isolate *Lactobacillus* spp. The plates were incubated anaerobically at 37 ◦C for 48 h [26]. Three hundreds of morphologically different microbial colonies were collected and cultured for further tests. The isolated LAB strains were then cultured in De Man, Rogosa and Sharpe (MRS) medium (Becton Dickinson, Cockeysville, MD, USA) containing 0.05% L-cysteine hydrochloride anaerobically at 37 ◦C. The bacterial stocks were stored at −80 ◦C with 17% glycerol utilized as a cryoprotectant.

### *2.2. Screening of SeZn-Tolerant LAB Isolates from Human Feces*

To identify selenium-tolerant strains, the isolates were plated on the MRS agar in the presence of 60 mM Na2SeO3 (Sigma-Aldrich, St. Louis, MO, USA) at 37 ◦C for 24 h under anaerobic condition. When the concentration of selenium in the medium is high, strains convert inorganic selenium into element of selenium (red color) in the medium [27]. Thus, the strains with selenium resistance were selected based on the results of bacteria growth and color changes.

Thereafter, the screened strains were further tested for zinc-tolerant abilities by culturing in the MRS agar with 100 mM ZnSO4·7H2O (Sigma-Aldrich) at 37 ◦C for 24 h under anaerobic condition. Strains with strong zinc tolerance were selected by observing the bacterial growth. The final screened SeZn-tolerant bacteria were identified by phylogenetic analysis of 16S rRNA gene sequence.

### *2.3. Assessing Bioaccumulation of Selenium and Zinc in LAB Strains During the Cultivation*

Considering the application for food and feed additive, the initial concentrations of Na2SeO3 and ZnSO4·7H2O were set at 0.01 mM and 3.5 mM, respectively. After 24 h anaerobic culture in MRS broth, the LAB strains were centrifuged (15,600× *g*, 5 min) to gain pellets. The bacterial pellets were washed three times with phosphate buffered saline (PBS) and frozen at −80 ◦C for lyophilization. One liter of PBS buffer (pH 7.4) was prepared by dissolving NaCl (8 g), KCl (200 mg), Na2HPO4 (1.44 g) and KH2PO4 (245 mg) in the distilled water and autoclaved at 121 ◦C for 15 min.

The concentrations of zinc and selenium in the bacterial biomass were measured using an inductively coupled plasma-atomic emission spectrometer (ICP-AES, Optima-4300 DV, Perkin Elmer, Waltham, MA, USA). The lyophilized sample (400 mg) was digested with HNO3 (5 mL) and HClO4 (0.5 mL) by heating in a Multiwave 3000 microwave. After the cool-down to room temperature, the solution was diluted with deionized water to reach a final volume of 20 mL, and mineral levels were assessed by the Inductively coupled plasma atomic emission (ICP-AES). The bioconversion rates of Se and Zn were calculated by dividing the Se or Zn content in dry cell mass by the total Se or Zn content added in the broth. The strain with the highest levels of selenium and zinc bioaccumulation was selected as the experimental strain for in vivo study.

Bioconversion rate of Se (%) = (Se content in dry cell mass/total Se content added in broth) × 100%

Bioconversion rate of Zn (%) = (Zn content in dry cell mass/total Zn content added in broth) × 100%

### *2.4. Gene Analysis of Se*/*Zn Uptake and Resistance in L. plantarum SeZi*

The genomic DNA of pure cultured *L. plantarum* SeZi isolate was extracted by using MG™ Cell Genomic DNA Extraction SV kit (MGmed, Seoul, Korea), following the manufacturer's instructions. Whole genome sequencing was carried out by using a Nextera XT Library Preparation kit (Illumina, San Diego, CA, USA) and sequenced at a read length of 300 bp with paired-end library via an Illumina MiSeq sequencer (Illumina, San Diego, CA, USA). The Illumina sequencing raw data in the FASTQ format were assembled with SPAdes 3.9.0. Gene-finding and functional annotation pipeline of whole genome assemblies used in EzBioCloud genome database (http://www.ezbiocloud.net, ChunLab Co., Ltd., Seoul, Korea) [28]. The tRNA genes were investigated via tRNAscan-SE 1.3.1 [29]. The rRNA and other non-coding RNAs were explored by using a Rfam covariance model version 12.0 [30]. Protein-coding sequences (CDSs) were predicted via Prodigal 2.6.2 [31], and classified into different functional groups (EggNOG 4.5; http://eggnogdb.embl.de) [32]. In order to obtain more functional annotation, the UBLAST program [33] was utilized to search and compare the predicted CDSs in the protein databases, including Swissprot [34], Kyoto Encyclopedia of Genes and Genomes (KEGG) [35] and subsystems-based annotations (SEED) [36]. The comparative genomics analysis was conducted by using the genome sequences of closely related *Lactobacillus plantarum* strains from the EzBioCloud database and analyzed via ChunLab's comparative genomics tool (http://www.ezbiocloud.net/contents/cg).

### *2.5. E*ff*ect of SeZn-Enriched L. plantarum SeZi in an ICR Mouse Model*

### 2.5.1. Preparation of SeZn-Enriched *L. plantarum* SeZi for Mouse Study

The *L. plantarum* SeZi strain was anaerobically grown at 37 ◦C for 24 h in MRS medium with the addition of 0.01 mM Na2SeO3 and 3.5 mM ZnSO4·7H2O. For harvesting probiotic powder, the bacterial pellets were collected by centrifugation (15,600× *g*, 5 min) after SeZn enrichment, thoroughly washed with the PBS buffer, and frozen at −80 ◦C for lyophilization.

### 2.5.2. Animals and Diets

Seven-week old male ICR mice were purchased from Central Lab Animal (Seoul, Korea). The animal breeding environment was adjusted to a dark cycle of 12 h light/12 h dark at a temperature of 23 ± 1 ◦C and a humidity of 40–60%. The mice were acclimatized in the laboratory room for one week and then randomly divided into four groups (*n* = 8/group). The control group was provided with a normal chow diet AIN-93G purchased from Doo Yeol Biotech (Seoul, Korea). The treatment groups (SZ, Lp, SZ + Lp) were fed the same normal chow diet mixed with Na2SeO3 (1.2 μg/g Se4+) and ZnSO4·7H2O (5 <sup>μ</sup>g/g Zn2<sup>+</sup>), 10<sup>12</sup> CFU/mouse *Lactobacillus plantarum* SeZi, and 10<sup>12</sup> CFU/mouse SeZn-enriched *L. plantarum* SeZi, respectively. The daily administration was conducted for 4 weeks. The protocols and facilities utilized in this animal experiment were approved by the Institutional Animal Care and Use Committee of Seoul National University (SNU-180403-2-2).

### 2.5.3. Blood Analysis

To assess Se and Zn concentrations and oxidative stress-related parameters in mouse blood, blood samples were collected into a 1.5 mL heparinized tube from the mouse heart by cardiac puncture. Approximately 0.9 mL of the blood samples were centrifuged (2500 rpm, 10 min) to separate serum. The serum and whole blood samples were stored at −80 ◦C. The concentrations of selenium and zinc in the whole blood were measured using ICP-AES. GSH-Px activity, SOD activity, and malondialdehyde (MDA) level in serum were assessed via antioxidant enzyme detection kits purchased from Jiancheng Bioengineering Institute, Nanjing, China.

### 2.5.4. Bacterial Community Analysis by 16S rRNA Gene Amplicon Sequencing

Fecal DNA was extracted using a QIAamp DNA Stool Mini Kit (Qiagen, Manchester, UK). The V3-V4 hypervariable regions of the 16S rRNA genes in the stool DNA samples were targeted and amplified using interest-specific primers. A pooled library was constituted by attaching specific barcode sequences to the 16S rRNA amplicons. The denatured and diluted pooled library and PhiX control (Phix control v3, 30%, *v*/*v*) library were mixed and loaded onto a MiSeq v2 (500 cycle) reagent cartridge (Illumina, San Diego, CA, USA). The primers and methods were as described in our previous study [37]. After the metagenomic sequencing, paired-end FASTQ files were collected and imported into Quantitative Insights Into Microbial Ecology 2 (QIIME2) (ver. 2020.6, https://qiime2.org) for analysis. Operational taxonomic unit (OTU) taxonomy and related analysis were performed using QIIME2 as described in our previous study [37]. KEGG associated with selenocompounds metabolism pathways were assessed by conducting phylogenetic investigation of the community by reconstruction of unobserved states (PICRUSt) with the entire picrust2 pipeline command [38].

### *2.6. Statistic Analysis*

Differential abundance analyses were performed by non-parametric one-way analysis of variance (ANOVA) using the Kruskal-Wallis test, or non-parametric *t*-test with Mann-Whitney test. Other analyses were conducted by one-way ANOVA with Tukey's multiple comparisons test or paired *t*-test analysis. All statistical analyses were carried out via Graph-Pad Prism 8. Statistically significant difference was accepted at *p* < 0.05.

### **3. Results**

### *3.1. Screening and Selection of SeZn-Tolerant LAB Strains*

Amongst the three hundreds of isolated strains, only four LAB species grew in the presence of 60 mM selenite, including *L. plantarum, L. pentosus, L. fermentum,* and *L. rhamnosus*. All of these four strains are able to resist 100 mM ZnSO4·7H2O. Among these four SeZn-tolerant strains, *L. plantarum* SeZi yielded the greatest dry cell mass with the best selenium and zinc bioaccumulation capability as shown in Table 1. Based on the results of whole genome sequencing analysis in Tables 2 and 3, the genes coding DedA and CysA proteins related to Na2SeO3 uptake and detoxification, and zinc uptake regulation protein ZUR (zinc uptake regulation) and zinc resistance protein MerR (Mercury resistance) were found in *L. plantarum* SeZi genome. Thus, *L. plantarum* SeZi was selected for in vivo mouse study. The abundances of these microbial selenium/zinc metabolism-related genes in genomes from other LAB strains were investigated using publicly deposited genome databases (NCBI genome datasets, https://www.ncbi.nlm.nih.gov/genome/). Among the 1912 genome assemblies available, only a few *Lactobacillus* genomes contain genes encoding DedA, CysA, ZUR, or MerR proteins (Table 4).

**Table 1.** Generation of biomass and bioconvertion rates of selenium and zinc in the selected SeZn-tolerant probiotic strains in vitro.


Data are expressed as mean ± SD (*n* = 5). Treatments with different letters are significantly different at *p* < 0.05 (*n* = 5).



<sup>a</sup> Genes encoded on the minus strand are indicated with (−).



<sup>a</sup> Genes encoded on the minus strand are indicated with (−). ZUR: zinc uptake regulation.

**Table 4.** Other *Lactobacillus* strains with selenium and/or zinc resistance gene clusters. (From NCBI genome databases).


### *3.2. Increased Concentrations of Selenium and Zinc in Blood after L. plantarum SeZi Administration*

The selenium and zinc contents in mouse blood are presented in Figure 1. The blood selenium and zinc levels in the SZ + Lp group were significantly higher than that of the control group and the SZ group, respectively (*p* < 0.05). The differences between the control, SZ and Lp groups were not significant (*p* > 0.05).

**Figure 1.** The concentrations of blood selenium (**A**) and zinc (**B**) in mice after 4 weeks of administration. Data were analyzed by unpaired *t*-test analysis and expressed as mean ± SD. \*\* *p* < 0.01 (*n* = 8). Con, control; SZ, selenium and zinc supplemented; Lp, *Lactobacillus plantarum* SeZi; SZ + Lp, selenium- and zinc-enriched *L. plantarum* SeZi.

### *3.3. Increased Antioxidant Activities in Mice after L. plantarum SeZi Administration*

GSH-Px and SOD are imperative antioxidant defenses against oxidative stress [39,40], and MDA is the most commonly utilized biomarker of oxidative stress [41]. As shown in Figure 2, the GSH-Px activity was highest in the SZ + Lp group, followed by the SZ group when compared with other groups (*p* < 0.05). The activity of SOD was significantly increased in the SZ + Lp group compared to the Lp group (*p* < 0.05). Meanwhile, significant decreases in the MDA level were observed in the SZ group and SZ + Lp group compared to other two groups (*p* < 0.05).

**Figure 2.** Glutathione peroxidase (GSH-Px) activity (**A**), superoxide dismutase (SOD) activity (**B**), and lipid oxidation product malondialdehyde (MDA) level (**C**) in serum of mice at the final day. Treatments with different letters (a, b, c) are significantly different at *p* < 0.05 (*n* = 8). Con, control; SZ, selenium and zinc supplemented; Lp, *Lactobacillus plantarum* SeZi; SZ + Lp, selenium- and zinc-enriched *L. plantarum* SeZi.

### *3.4. Changes in the Gut Microbiota after L. plantarum SeZi Administration*

Gut microbiota alpha diversities were assessed by richness (Faith-pd) and Pielou's evenness analyses, which represent the number of species and the degree of species homogeneity, respectively. No significant difference was observed between the groups in the alpha diversity of richness (Figure 3A), while the evenness in the SZ group was significantly larger than Lp and SZ + Lp groups (Figure 3B). The results of beta diversity (Bray-Curtis dissimilarity) indicated that the clustering in microbial communities in the Lp and SZ + Lp group was distinct from that in the control and SZ group (Figure 3C).

The average relative abundances of the final day fecal samples at the phylum level (Figure 3D) and the genus level (Figure 3E) suggested the different microbial compositions amongst the groups after the oral administration of SZ and SZ + Lp. To further evaluate the effects of treatments on microbial compositional changes, the three significantly different genera between the groups were identified and are displayed in Figure 4A–C. The relative abundance of *Lactobacillus* in the SZ + Lp group was significantly higher than that of the control and SZ groups, and the *Lactobacillus* level in the Lp group was also significantly higher than the SZ group. *Adlercreutzia* was significantly abundant in the SZ group compared to the SZ and SZ + Lp groups. Interestingly, *Lactococcus* was highly enriched only in the SZ group with the relative abundance at 4.15%. In addition, the relative abundance of *Allobaculum* in the SZ group was much larger than the SZ + Lp group, although the difference was not statistically significant (Figure 4D, *p* > 0.05).

**Figure 3.** Comparison of diversity indices and microbial compositions amongst groups after 4 weeks of oral administration. Alpha-diversities of microbial communities are shown as (**A**) richness and (**B**) evenness. (**C**) Principal Coordinates Analysis (PCoA) plot represents beta-diversity based on Bray-Curtis dissimilarity. (**D**) Taxonomic profiles at the phylum level (**D**) and the genus level (**E**). Relative Abundance of taxa below 0.1% were excluded prior to analyses. Significance was accepted at \* *p* < 0.05, \*\* *p* < 0.01 (*n* = 8). Con, control; SZ, selenium and zinc supplemented; Lp, *Lactobacillus plantarum* SeZi; SZ + Lp, selenium- and zinc-enriched *L. plantarum* SeZi.

**Figure 4.** Relative abundances of *Lactobacillus* (**A**), *Adlercreutzia* (**B**), *Lactococcus* (**C**), and *Allobaculum* (**D**) in fecal samples after 4 weeks of oral administration. Data are expressed as mean ± SD. Significance was accepted at \* *p* < 0.05, \*\* *p* < 0.01 (*n* = 8). Con, control; SZ, selenium and zinc supplemented; Lp, *Lactobacillus plantarum* SeZi; SZ + Lp, selenium- and zinc-enriched *L. plantarum* SeZi.

### *3.5. Microbial Function Analysis Related to Selenocompounds Metabolism*

To investigate the functional changes in selenocompounds metabolism of the gut microbiome, KEGG analysis was performed by phylogenetic investigation of the community by reconstruction of unobserved states (PICRUSt).

As presented in Figure 5, *L. plantarum* SeZi significantly increased the relative abundances of the SCLY gene coding for selenocysteine lyase (EC: 4.4.1.16), CCBL gene coding for cysteine-S-conjugate beta-lyase (EC: 4.4.1.13), and MARS gene coding for methionyl-tRNA synthetase (EC: 6.1.1.10). These selenocompounds metabolism-related genes were responsible for the utilization of selenocysteine, seleno-cystathionine and seleno-methionine, respectively.

**Figure 5.** Selenocompounds metabolism pathway. Related enzymes include selenocysteine lyase (EC: 4.4.1.16), cystathionine gamma-synthase (EC: 2.5.1.48), cysteine-S-conjugate beta-lyase (EC: 4.4.1.13), homocysteine methyltransferase (EC: 2.1.1.13), and methionyl-tRNA synthetase (EC: 6.1.1.10). Arrows indicate the related genes that are involved in the corresponding pathway. Error bars represent means ± SD. Significance was accepted at \* *p* < 0.05, \*\* *p* < 0.01 (*n* = 8). Con, control; SZ, selenium and zinc supplemented; Lp, *Lactobacillus plantarum* SeZi; SZ + Lp, selenium- and zinc-enriched *L. plantarum* SeZi. K11717, cysteine desulfurase/selenocysteine lyase; K01760, cystathionine beta-lyase; K01874, methionyl-tRNA synthetase.

To detoxify selenite and selenate in the selenocompounds metabolism pathway, selenite is converted to selenate directly or via an intermediate, and selenate is further metabolized into hydrogen selenide. As shown in Figure 6, the relative abundances of enzymes related to oxidation of selenite to selenate (EC: 2.7.7.4, EC: 1.97.1.9) were significantly reduced in the Lp and SZ + Lp groups.

*Antioxidants* **2020**, *9*, 1028

**Figure 6.** Detoxification process of inorganic selenium. Related enzymes include 3 -phosphoadenosine 5 -phosphosulfate synthase (EC: 2.7.7.4) and selenate reductase subunit alpha (EC: 1.97.1.9). Arrows indicate the related genes that are involved in the corresponding pathway. Error bars represent means ± SD. Significance was accepted at \* *p* < 0.05, \*\* *p* < 0.01 (*n* = 8). Con, control; SZ, selenium and zinc supplemented; Lp, *Lactobacillus plantarum* SeZi; SZ + Lp, selenium- and zinc-enriched *L. plantarum* SeZi. K07310, Tat-targeted selenate reductase subunit YnfF; K07309, Tat-targeted selenate reductase subunit YnfE; K12527, putative selenate reductase; K00956, sulfate adenylyltransferase subunit 1; K00957, sulfate adenylyltransferase subunit 2; K00958, sulfate adenylyltransferase.

### **4. Discussion**

In this study, *L. plantarum* SeZi isolate was screened out from human fecal bacteria by the seleniumand zinc-tolerant abilities. The selenium- and zinc-enriched *L. plantarum* SeZi strain increased the levels of selenium and zinc and presented antioxidative properties in an ICR mouse model. A thorough search of the literature reporting the bioavailability and functionality of Se- and Zn-enriched microorganisms using a mouse model yielded only one related article [42]. To determine appropriate concentrations of selenium and zinc for mice, we referenced this article and 300 μg of selenium and 1.5 mg of zinc (/kg body weight/day) were used in our study. Yan et al. reported the antioxidant and antitumor activities were significantly increased by supplementation with Se/Zn-enriched mushrooms. However, the in vivo antioxidant activity of Se/Zn-enriched LAB has not been reported yet.

Selenium is an essential element that must be exogenously provided to reach the requirement of human and animal health [43]. The toxicity order of selenium species from high to low is selenate, selenite, nano-selenium, and lactomicro-selenium [44]. The accumulation of selenium in bacteria is processed by extracellular binding via active groups in the cell-membrane conjunction or intracellular binding via ion transportation on the membrane [45]. Many *Lactobacillus* strains are well-known to accumulate and biotransform toxic selenite into non-toxic seleno-amino acids (i.e., selenocysteine and selenomethionine) and selenoprotein [46,47] The utilization of Se-enriched *Lactobacillus* possesses unique advantages, including low toxicity, low cost production and additional probiotic effects.

In this study, we focused on the changes in the gut microbial composition as well as functional metabolism associated with Se/Zn uptake and utilization. Interestingly, Se/Zn supplementation greatly induced the enrichment of specific genus, *Lactococcus* (belonging to LAB), which confirmed the previous reports regarding in vitro tolerance of LAB to selenium [46,47]. However, this indigenously enriched LAB by inorganic Se/Zn supplement showed significantly different patterns in the microbial selenocompounds metabolism compared with Se/Zn-bioaccumulated *L. plantarum* SeZi strain.

To further investigate strain-specific functions, whole genome sequencing of *L. plantarum* SeZi strain was conducted and analyzed based on public databases. The genes coding *cysA* and *dedA* were observed in the whole genome sequencing of *L. plantarum* SeZi. Both of the *cysA* and *dedA* genes are associated with selenite uptake and detoxification. These genes were abundantly present in *L. plantarum* SeZi strain, but not observed in *Lactococcus* or other *Lactobacillus* spp. According to previous studies, selenite may enter the cells of *E. coli* through the sulfate permease CysA [48]. DedA can uptake selenite into cells as a direct transporter or a cofactor. Additionally, the *dedA* gene-contained mutant *E. coli* displayed selenite resistance [17]. Based on the results of targeted metagenome sequencing, the *L. plantarum* SeZi strain promoted the utilization of selenocysteine, seleno-cystathionine and seleno-methionine in selenocompounds metabolism pathway of gut microbiome. Besides, the transformation between the toxic inorganic selenium was reduced by *L. plantarum* SeZi in the detoxification process. These are consistent with the selenium-related function detected in the genome sequencing of *L. plantarum* SeZi.

Zinc plays an essential role in catalytic, structural, and regulatory functions in enzymes and protein domains [49]. Similar to the other trace elements (i.e., selenium), zinc in the organic forms have more bioavailability its inorganic forms [49]. Pharmacological zinc supplements often have low bioavailability and are easily overdosed [50]. Although internalizing zinc by *Lactobacillus* has not been studied in depth, the utilization of certain *Lactobacillus* species can be a promising alternative to deliver zinc in a highly organic form [51].

The selenium and zinc levels in the blood can be affected by dietary supplementation and related metabolism. In this study, the concentrations of selenium and zinc in the Se/Zn-enriched *Lactobacillus* group were significantly higher than the control or SZ group. This is consistent with a recently published study which indicates that feeding a diet supplemented with Se/Zn-enriched probiotics, 0.3 mg/L selenium, and 100 mg/L zinc significantly enhanced the blood selenium and zinc concentrations in Wistar rats [52]. It is also the only published study referred to Se/Zn-enriched probiotics in a murine model to date. However, the assessment of antioxidant capability and its potential mechanisms was not reported yet.

In this study, the enhancement of antioxidant ability in the SZ + Lp group was probably triggered by the increased selenium level. When selenium is incorporated into the selenoenzymes (i.e., GSH-Px), it enhances the antioxidant activities by suppressing the nuclear factor-kappa B (NF-κB) signal pathway [53]. Most selenoproteins take part in the defense against oxidative stress, protecting tissues and cells from oxidative damages [47]. Although the SZ group was administered with inorganic forms of Se/Zn, the SZ + Lp group was administered with bio-accumulated Se/Zn in bacteria. To understand the mechanisms of bioavailability and functionality related to inorganic/organic forms of trace nutrients, it is necessary to investigate the changes in selenium and zinc metabolites (e.g., selenoproteins, zincproteins, etc.) from blood and fecal samples.

Alteration of microbial communities was evaluated by 16S rRNA community analysis in this study. Up to now, the 16S metagenomic technique is still rarely used in the selenium or zinc-related animal studies. Obviously, the significant surge of *Lactobacillus* levels in the Lp and SZ + Lp groups was caused by the oral administration of *L. plantarum* SeZi. The relative abundance of *Adlercreutzia* was significantly higher in the SZ + Lp group compared to the control and SZ groups. According to the previous studies, a decreased level of *Adlercreutzia* was observed in multiple sclerosis patients and Alzheimer's disease patients [54,55]. Amongst *Adlercreutzia* species, *A. equolifaciens* is an equol-producing bacteria, promoting intestinal health [56]. Besides, a high relative abundance of *Lactococcus* (approximately 4.15%) was found in the SZ group, while this bacterium was almost undetectable in the other groups. It might be caused by utilization of selenium and zinc by *Lactococcus* spp. According to the previous studies, *Lactococcus lactis* is capable of selenium biotransformation and zinc uptake [57,58]. The relative abundance of *Allobaculum* spp., which was markedly reduced in the SZ + Lp group compared with other groups, was reported to be adversely associated with mRNA expression levels of tight junction protein genes (*zo-1* and *occludin*) and anti-inflammatory genes (*foxp3* and *Il-10*) in the colon of rats [59]. In some inflammatory bowel disease (IBD) patients, *Allobaculum* spp. was one of the uniquely observed species [60].

To the best of our knowledge, this study is the first attempt to evaluate the effects of Se/Zn-enriched LAB on in vivo gut microbiome changes. The microbiome analysis suggests a microbial aspect of selenocompound metabolism, however, lacks information on host's Se/Zn metabolic process. Future studies on evaluating functionality of *L. plantarum* SeZi in a disease-induced mouse model should include host's metabolites analysis as well.

### **5. Conclusions**

In conclusion, the selected strain *L. plantarum* SeZi is able to resist and biotransform inorganic selenium into organic selenium. In the in vivo study, the selenium and zinc-enriched *L. plantarum* SeZi increased blood selenium level, antioxidant capability and the utilization of seleno-amino acids. Therefore, the *L. plantarum* SeZi strain is a potential selenium and zinc-enriched probiotic for application as functional food ingredients in the future.

**Author Contributions:** Conceptualization, S.K., R.L., H.J.Y., and G.E.J.; Investigation, S.K., R.L., and H.J.; Data curation, S.K. and R.L.; Methodology, S.K. and R.L.; Formal analysis, S.K.; Software, S.K.; Visualization, S.K.; writing-original draft preparation, S.K. and R.L.; writing—review and editing, S.K., H.J.Y. and G.E.J.; supervising, H.J.Y. and G.E.J.; funding acquisition, H.J.Y. and G.E.J. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was carried out with the support of the Ministry of Small and Medium-sized Enterprises (SMEs) and Startups (MSS), Korea, under the "Regional Specialized Industry Development Program (R&D, Project number S2848321)" supervised by the Korea Institute for Advancement of Technology (KIAT). This work was also supported by the "K-BIO KIURI Center program (Project number 2020M3H1A1073304)".

**Conflicts of Interest:** G.E.J. holds BIFIDO Ltd. stocks. Other 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/).

**Małgorzata Makarewicz, Iwona Drozd˙ z, Tomasz Tarko and Aleksandra Duda-Chodak \* ˙**

Department of Fermentation Technology and Microbiology, Faculty of Food Technology, University of Agriculture in Krakow, 30-149 Kraków, Poland; malgorzata.makarewicz@urk.edu.pl (M.M.); iwona.drozdz@urk.edu.pl (I.D.); tomasz.tarko@urk.edu.pl (T.T.)

**\*** Correspondence: aleksandra.duda-chodak@urk.edu.pl; Tel.: +48-12-662-4792

**Abstract:** This review presents the comprehensive knowledge about the bidirectional relationship between polyphenols and the gut microbiome. The first part is related to polyphenols' impacts on various microorganisms, especially bacteria, and their influence on intestinal pathogens. The research data on the mechanisms of polyphenol action were collected together and organized. The impact of various polyphenols groups on intestinal bacteria both on the whole "microbiota" and on particular species, including probiotics, are presented. Moreover, the impact of polyphenols present in food (bound to the matrix) was compared with the purified polyphenols (such as in dietary supplements) as well as polyphenols in the form of derivatives (such as glycosides) with those in the form of aglycones. The second part of the paper discusses in detail the mechanisms (pathways) and the role of bacterial biotransformation of the most important groups of polyphenols, including the production of bioactive metabolites with a significant impact on the human organism (both positive and negative).

**Keywords:** intestinal microbiota; inhibition; metabolism; catabolism; biotransformation; bioactive compounds; health; metabolites; diversity

### **1. Introduction**

The intestinal microbiome plays an important, if not crucial, role in the metabolism of chemical compounds delivered with food, especially those that are undigested in the upper digestive tract. The enormous number of bacterial cells inhabiting the large intestine forms a complex ecosystem called the "intestinal microbiome". The word microbiome was introduced for the first time in 2001 to define the collective genomes of the microbiota [1]. Since then, much research and many projects were dedicated to assessing the impact of intestinal microbiota on a host's health, especially by determining its role in food metabolism, xenobiotics biotransformation and various disease development.

It is estimated that the microbiota of an adult is composed of ~10<sup>14</sup> bacteria cells [2] belonging to 1000–1150 species, with each individual harboring at least 160 species (usually about 500 species) [3]. Based on the large-scale 16S rRNA or metagenomic studies, scientists stated that ~80% of the bacteria identified by molecular tools in the human gut are uncultured and hence can be characterized only by metagenomic studies [4,5]. There are significant interindividual differences in the bacterial species found in the gastrointestinal tract. The composition, as well as the ratio of different species that form the intestinal microbiome, is very diverse within the human population, and each individual has his or her own unique profile of microbial species, which can be compared to a fingerprint. The differentiation of gut microbiota composition and profile is caused by the influence of multiple and diverse factors, such as age, origin, geographical location, environment, dietary habits (including probiotics), health, the application of antibiotics or even in the way an individual is born [6–8]. However, despite the great diversity of bacterial species, the majority of them belong to only four bacterial phyla: Firmicutes (64%), Bacteroidetes (23%), Proteobacteria

**Citation:** Makarewicz, M.; Drozd˙ z, I.; ˙ Tarko, T.; Duda-Chodak, A. The Interactions between Polyphenols and Microorganisms, Especially Gut Microbiota. *Antioxidants* **2021**, *10*, 188. https://doi.org/10.3390/ antiox10020188

Received: 22 December 2020 Accepted: 25 January 2021 Published: 28 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/).

(8%) and Actinobacteria (3%), whereas other taxons highly diverse [2]. Among the key functions of microbiota is occupying the intestinal surfaces and production of antimicrobial compounds that prevent the invasion of pathogens. Both commensal bacteria and gut pathogens (such as *Salmonella*, *Shigella*, *Helicobacter*, *Vibrio*, *Campylobacter*, *Yersinia*, *Clostridia*, *Aeromonas*, *Listeria*, *Streptococcus*, and *Staphylococcus*, as well as pathogenic strains of *Escherichia coli*, *Klebsiella pneumoniae*, *Enterococcus faecalis*) require similar ecological niches to colonize and proliferate in the intestine [9,10]. Therefore, various mechanisms designed to compete with each other have evolved. Commensal bacteria produce bacteriocins that specifically inhibit members of the same or similar bacterial species (e.g., *E. coli* versus pathogen enterohemorrhagic *E. coli*). Commensal bacteria produce short-chain fatty acids and cause pH reduction, thereby preventing the colonization by pathogens whose optimal pH for growth is neutral [11]. An altered bacterial community structure may facilitate the gut colonization by enteric pathogens but can also favor the overgrowth of potentially harmful subsets of indigenous bacteria, like virulent *E. coli* or *Clostridium difficile*.

The great diversity of bacterial species forming the gut microbiota implicates the large number of genes which they contain [2] and the enormous metabolic capacity of the intestinal microbiome, which is approximately 100-fold greater than that of the human liver [2,12]. The intestinal microbiota is equipped with a large set of different enzymes able to hydrolyze glycosides, glucuronides, sulfates, amides, esters and lactones through the action of enzymes such as α-rhamnosidase, β-glucuronidase, β-glucosidase, sulfatase and various esterases. Other reactions catalyzed by the gut microbial enzymes are aromatic ring cleavage, reductions (reductases, hydrogenases), decarboxylation (decarboxylase), demethylation (demethylase), isomerization (isomerase), and dehydroxylation (dehydroxylase) [13,14].

Intestinal bacteria contribute to the breakdown of polysaccharides and polyphenols as well as participate in the synthesis of vitamins (K, B12) and amino acids [8]. Many metabolites produced by gut microbiota are involved in various important physiological processes of the host, including energy metabolism and immunity. For example, the essential aromatic amino acid tryptophan can be metabolized by—among others—*Peptostreptococcus russellii*, *Clostridium sporogenes*, and *Lactobacillus* spp. to indole derivatives, which are ligands for aryl hydrocarbon receptor (AhR). This transcription factor plays an important role in the human immunological response, and via modulating T cell differentiation, Th17 development and IL-22 production may inhibit inflammation [15]. Branched-chain amino acids (BCAAs) (such as leucine, isoleucine, and valine) are essential amino acids that possess an aliphatic side chain with a branch, and that cannot be synthesized by humans. Therefore, they are provided by diet or synthesized by gut microbiota. The main species that contribute to the BCAAs production are *Prevotella copri* and *Bacteroides vulgatus* [15].

To date, thousands of microbial metabolites with known and unknown functions have been identified as components of the human metabolome. Well-known are short-chain fatty acids (SCFAs) with the acetate, propionate, and butyrate, being the metabolite of resistant starch and dietary fiber fermentation [15]. SCFAs are generally considered to have beneficial effects on host health, and they modulate metabolism, inflammation, hormone production, lipogenesis, and gut homeostasis. Gut microbiota can also metabolize dietary L-carnitine, choline, and lecithin into trimethylamine (TMA), which is then converted to trimethylamine-N-oxide (TMAO) in the liver of a host.

Owing to the multitude of direct and indirect interactions with the host organism, the intestinal microbiome is hence closely linked to the health of a host [16,17]. Dysbiosis, an imbalanced or disturbed microbiota composition, may play a significant role in etiology or the development of various gastrointestinal diseases such as inflammatory bowel disease (IBD), irritable bowel syndrome (IBS), colon cancer, and antibiotic-associated diarrhea [12,18]. The gut microbiota also plays a critical role in the transformation of dietary polyphenols into absorbable biologically active compounds. It is estimated that about 90–95% of the total polyphenol intake remains unabsorbed and colonic bacteria act enzymatically on their backbone, producing metabolites with a different physiological significance [19].

### **2. The Structure and Role of Polyphenols**

Polyphenols are secondary metabolites playing an important role in plant tissues. They provide the color to flowers and fruits (mainly anthocyanins), which attracts pollinators and seed dispersers; they are also responsible for flavor in fruit and vegetables, as well as protecting plant tissues against herbivores and other biotic and abiotic stressors, like UV radiation, cold, heat or salinity [20]. Flavonoids take part in energy transfers, the regulation of photosynthesis and morphogenesis, regulation of growth factors, and sex determination and—due to antimicrobial activity—protect against the spread of pathogens in plant tissues [21]. Polyphenols also influence human health. Because of antioxidant properties and free-radical scavenging activity, they are believed to protect against various diseases, e.g., cancer, stroke and myocardial infarction, cardiovascular diseases, and some immunological and neurological disorders; they are also thought to have a beneficial impact on humans with diabetes and obesity [22–31]. Several in vitro and in vivo animal studies have demonstrated the antioxidant and anti-inflammatory effects of polyphenols in the brain–liver–gut axis [32], and polyphenols have been shown to target different stages of the inflammatory cascade to reduce the severity of inflammation. Polyphenols can also modulate various signal pathways, for example, through interaction with AMP-activated protein kinase (AMPK), CCAAT/enhancer-binding protein α (C/EBPα), peroxisome proliferatoractivated receptor γ (PPARγ), and peroxisome proliferator-activated receptor-gamma coactivator 1-alpha (PGC-1α), sirtuin 1, and sterol regulatory element-binding protein-1c (SREBP1c) involved mainly in cellular energy metabolism and adipogenesis, as well as uncoupling proteins 1 and 2 (UCP1 and UCP2), and NF-κB that regulate antioxidant and anti-inflammatory responses [28].

Polyphenols are a large group of compounds that comprise phenolic acids, flavonoids, tannins, lignans, stilbens and coumarins (Figure 1 presents the chemical structure of flavonoids, Figure 2—non-flavonoids). In a human diet, they are provided mainly by plant food such as fruit, vegetables, tea, wine, coffee, and cocoa. However, even when phenolic compounds occur in the human diet in large quantities, they do not always show high biological activity after consumption.

Polyphenols' influence on human health depends both on their amount of food and on their bioavailability, bioaccessibility and the biological activity of metabolites produced in the human body. Some phenolic compounds have limited absorption in the digestive tract, while the others undergo an intensive metabolism to derivatives with a lower activity or they undergo a rapid elimination (degradation). The uniqueness of the gut microbiota composition causes that in one individual, a given polyphenol will undergo bacterial metabolism and will have an effect (beneficial or negative), whereas in another human being, the metabolism of the same polyphenol will follow a different path, and there will be no effect.

**Figure 1.** *Cont.*

**Figure 1.** *Cont.*

**Figure 1.** *Cont.*

**Figure 1.** *Cont.*

**Figure 1.** *Cont.*

164

**Figure 1.** The chemical structure of various classes of flavonoids. Based on [16,24,33–40].


$$\bigtimes\_{\ast}^{\ast}$$

**Figure 2.** *Cont.*

**Figure 2.** *Cont.*

**Figure 2.** *Cont.*

**Figure 2.** The chemical structure of various groups of non-flavonoid polyphenols. Based on [16,24,33–40].

### **3. The Impact of Polyphenols on Microorganisms and the Mechanism of Their Action**

Herbs and spices have a long history of being used as natural food preservatives and within folk medicine. It is due to substances with the antimicrobial activity they contain, such as flavonoids, anthocyanins, alkaloids, glycosides, saponins, coumarins, tannins, vitamins, phenolic acids and many more. Essential oils (a complex mixture of various bioactive compounds, inter alia, polyphenols), plant extracts and pure polyphenols are large groups of compounds with strong antibacterial properties. It has been proven many times that thyme, oregano, rosemary, sage, mint and other herbs and spices can inhibit Gram-positive and Gram-negative bacteria, including pathogens [20]. There are many scientific reports and publications demonstrating that pure polyphenols or bioactive compounds present in various types of plant preparations (e.g., aqueous, ethanolic or methanolic extracts, essential oils, enriched extracts) can exert both negative and positive impact on microorganisms. Some examples of such impact are presented in Table 1.









As can be seen in the table, not only pathogenic bacteria growth can be inhibited by polyphenols. It was proved that some beneficial microorganisms, inter alia lactic acid bacteria and probiotics strains, can also be inhibited. However, the bacteriostatic or bactericidal effect depends both on the polyphenol structure and bacteria species. Susceptibility to some polyphenols was also proved to be strain-dependent [43,47,50,64]. It was proved that some polyphenols exert a beneficial influence on bacteria. They can stimulate growth or at least change the composition of the microbiome in favor of beneficial bacteria such as *Bifidobacterium* and *Lactobacillus*, which both contribute to gut barrier protection; *Akkermansia muciniphila* and *Faecalibacterium prausnitzii* that possess anti-inflammatory effect by inhibiting the activation of NF-kB; and *Roseburia* sp.—the butyrate-producer [15]. *Akkermansia muciniphila* is an anaerobic, mucin-degrading bacterium residing in the healthy intestinal tract of a host that is believed to have several health benefits in humans [89]. Many studies show that various diseases, e.g., obesity, type II diabetes and inflammatory bowel diseases have an association with reducing *A. muciniphila* abundance [90]. Taking into account that polyphenols have been proven to increase the growth of *Akkermansia muciniphila* [72,73,75–77], the beneficial impact of polyphenols on human health may result from other than antioxidant activity. All these findings (Table 1) suggest that polyphenols reaching the large intestine may not only be catabolized to small phenolic acids but also elicit potentially beneficial effects of intestinal probiotic bacteria. Taken together, polyphenols appear to be able to alter gut microecology and, by affecting the total number of beneficial species in the gut, may confer positive gut health benefits.

### **4. Mechanism of Antibacterial Activity of Polyphenols**

As was reported above, the influence of pure polyphenols and plant extract, as well as the strength of that impact on bacteria, differs depending on the kind of both phenolic compounds and bacteria strain. The mechanism of antimicrobial activity of polyphenols against bacteria can differ and also depends both on the polyphenol type and bacteria species. Among the most important mechanisms of the antibacterial action of polyphenols are [58,91–98]:


### *4.1. Reactions with Proteins*

The antibacterial activity of flavonoids may result from their ability to form complexes with proteins through nonspecific forces such as hydrogen bonding and hydrophobic effects, as well as by covalent bond formation [99]. Due to protein binding by polyphenols, they are sequestered into soluble or insoluble complexes, which affects the function of both polyphenol and protein [100]. Proteins modified by polyphenols binding have some amino acids blocked or undergo conformation transitions, which can cause changes in protein structure, solubility, hydrophobicity, thermal stability, and the isoelectric point. In consequence, protein–phenolic complexation leads to changes in their physicochemical and biological properties, including the digestibility and utilization of food proteins as well as the activity of digestive enzymes [101]. It has been demonstrated that naturally occurring polyphenols, e.g., condensed tannins, can inhibit a number of digestive enzymes, including α-glycosidase, α-amylase, lipase, pepsin, trypsin, and chymotrypsin, changing the availability of nutrients and hence modulating the microbiota composition [102–106]. Furthermore, polyphenols can bind to important bacterial proteins such as adhesins, enzymes, cell envelope transport proteins and—by inactivating them—exert an antimicrobial impact. On the other hand, polyphenols complexation with proteins may influence the bioaccessibility and activity of phenolic compounds.

Quinones are known to complex irreversibly with nucleophilic amino acids in proteins, which leads to inactivation and loss of function in the proteins. They possibly interact with cell wall polypeptides, membrane-bound enzymes and surface-exposed adhesins of pathogenic bacteria [107].

Kaempferol-3-rutinoside (nicotiflorin) **(130)** was demonstrated to inhibit *Streptococcus mutans*. The affected protein was sortase A, a membrane enzyme that actively plays a crucial role in bacteria adhesion and the invasion of host cells [108]. A purified *S. mutans* sortase A was inhibited by curcumin **(216)** at a half-maximal inhibitory concentration; curcumin was also found to release the Pac protein to the supernatant and reduce *S. mutans* biofilm formation [109]. The sortase enzymes (cysteine transpeptidases) are used by Gram-positive bacteria to display proteins in a cell surface (e.g., glycoproteins), and they can attach to proteins in the cross-bridge peptide of the cell wall. Hence they are an important virulence factor. Morin **(110)**, myricetin **(114)**, and quercetin **(111)** exhibited strong inhibitory activity against sortase A and B from *S. aureus* [110].

Some flavones were active against *Escherichia coli* by forming complexes with extracellular and soluble proteins [111]. Chen et al. [112] proved that baicalein **(45)** decreased the expression of intracellular adhesin in *S. aureus*. EGCG can bind to porins, so probably this way, it affects the permeability of the outer membrane of Gram-negative bacteria via porin pores [113].

Nakayama et al. [114] demonstrated with two-dimensional electrophoresis that epigallocatechin gallate **(18)** strongly interacted with one of the outer membrane porin proteins of *E. coli*, especially with basic amino acids such as Arg, Lys and His. The docking simulation revealed that EGCG enters into the porin pore and binds to Arg residues present on the inner surface of the pore channel through hydrogen bonding, resulting in inhibition of the porin function.

It also has been shown that tea catechins [115] and various polyphenols [20] have the capacity to sensitize strains of methicillin-resistant *Staphylococcus aureus* to antibiotics. Taylor et al. [115] postulated that catechin gallates (compounds **13**–**18**) intercalate into phospholipid bilayers, and probably they affect both virulence and antibiotic resistance by perturbing the function of the key processes associated with the bacterial cytoplasmic membrane.

The inhibitory impact of polyphenols on the action of the bacterial efflux pump, which changes transport through the cell wall and cytoplasmic membrane, is also taken into account [91,116]. It has been demonstrated that quinones and chalcones are substrates of bacterial efflux pumps and could be used in combination with the efflux inhibitors in order to improve the accumulation of the drug in the cells to fight against MRSA infections [117]. Kaempferol **(108)** and galangin **(109)** were found to be effective efflux pump inhibitors in *S. aureus* [118].

Flavonoids can modulate the activity of bacterial enzymes, which are crucial for cell life, such are those catalyzing the synthesis of cell wall elements, cell membrane fatty acids or ATP. Fatty acid synthase II (FAS-II) is a key enzyme for the synthesis of fatty acids building the bacterial membranes. It catalyzes fatty acid chain elongation, from 16–24 carbons obtained de novo by FAS-I to long-chain fatty acids of 36–48 carbons as well as mycolic acids [92]. Flavonoids such as isoliquiritigenin **(142)**, butein **(144)**, fisetin **(107)** and 2,2 ,4 -trihydroxychalcone **(139)** inhibited FAS-II, thus preventing the growth of *Mycobacterium smegmatis* [119].

Epigallocatechin gallate (EGCG) **(18)** and the related tea catechins potently inhibited both the FabG and FabI reductase steps in the fatty acid elongation cycle [120]. The authors suggested that the presence of the galloyl moiety was essential for inhibitory activity, and EGCG was a competitive inhibitor of FabI and a mixed-type inhibitor of FabG, demonstrating that EGCG interfered with cofactor binding in both enzymes. Furthermore, EGCG inhibited acetate incorporation into fatty acids in vivo. Molecular docking studies conducted by Xiao et al. [121] revealed the importance of the 3-*O*-galloyl or 3-*O*-glycosides side chain at the flavonoid pyran ring in the mechanism of the inhibition of reductase flavoprotein, dihydroorotate dehydrogenase (PyrD), dihydrofolate reductase (DYR), NADH-dependent

enoyl-ACP reductase, and the DNA gyrase subunit in *E. coli*. Results obtained in the study demonstrated that EGCG has the strongest binding with NADH-dependent enoyl-ACP reductase (FabI) in comparison with other flavonoids, while quercitrin **(111)** was also the strongest inhibitor of DNA gyrase subunit B (GyrB) among tested 19 flavonoids. The results also indicated that flavonoids that have galloyl moieties, such as EGCG **(18)**, (−)-catechin gallate **(16)**, (−)-epicatechin gallate **(17)**, and (−)-gallocatechin gallate **(13)**, exhibited higher binding affinities to PyrD, FabI, and DYR than their cognates lacking the galloyl group, i.e., (−)-epigallocatechin **(14)**, (+)-catechin **(7)**, (−)-epicatechin **(12)**, and (−)-gallocatechin **(11)**, respectively.

Quercetin **(111)**, apigenin **(38)**, and sakuranetin **(84)** inhibited the activity of βhydroxyacyl-acyl carrier protein dehydratase from *Helicobacter pylori* (HpFabZ), which is necessary for bacterial fatty acid biosynthesis. These three flavonoids are all competitive inhibitors against HpFabZ by binding to the substrate tunnel and preventing the substrate from accessing the active site [122]. Similarly, β-ketoacyl acyl carrier protein synthase (KAS) III is a key catalyst in bacterial fatty acid biosynthesis. The docking studies between *Enterococcus faecalis* KAS III (efKAS III), and flavonoids proved that naringenin **(81)**, eriodictyol **(94)**, and taxifolin **(134)**, with high-scoring functions and good binding affinities, docked well with efKAS III, causing the *E. faecalis* growth inhibition. Hydrogen bonds between the 5- and 4 -hydroxy groups and the side-chain of Arg38 and the backbone carbonyl of Phe308 were the key interactions for efKAS III inhibition [123].

Both Gram-positive and Gram-negative bacteria produce hyaluronidases, which are an important virulence factor. They enable the bacteria to avoid the immune system and host defense mechanisms. Terpenes (e.g., glycyrrhizin) have been identified as hyaluronic acid lyases (Hyal B from *Streptococcus agalactiae*, Hyal S from *Streptomyces hyalurolyticus*, and Hay C form *Streptococcus equisimilis*). Compounds with many hydroxyl groups inhibited hyaluronate lyase stronger than those with only a few [124].

Flavonoid 5,6-dihydroxy-4 ,7,8-trimethoxyflavone, isolated from *Limnophila heterophylla* Benth, was found to effectively kill *Bacillus subtilis* by cell lysis. Moreover, they enhanced the activity of gluconeogenic fructose 1,6-bisphosphatase, but the decreased activity of phosphofructokinase and isocitrate dehydrogenase, the key enzymes of the Embden–Meyerhof–Parnas pathway and the tricarboxylic acid cycle, respectively, was demonstrated [125].

Isoflavones (4-(p-hydroxyphenethyl) pyrogallol and 7,8,4 -trihydroxyisoflavone **(80)** are potent inhibitors of urease, an enzyme produced by *Helicobacter pylori*, which catalyzes the hydrolysis of urea to produce ammonia and carbon dioxide and to protect the bacteria in the acidic environment of the stomach [126]. The structure–activity relationship of these polyphenols revealed that the two *o*-hydroxyl groups were essential for the inhibitory activity of polyphenol. When the C-ring of isoflavone was broken, the inhibitory activity markedly decreased.

It was observed [127] that treatment *Pseudomonas aeruginosa* with cranberry type-A proanthocyanidins **(23,24)** caused downregulation of a wide variety of proteins, including those related to ATP synthesis (likely cytochrome C PA2482), purine, carbohydrate, aminoacid and fatty acid metabolism (HmgA, GuaB, FdhE, FoaB, LdcA, PurU1) and involved in nucleic acid synthesis and repair (e.g., TopA, Rne, RplC, and Mfd). In addition, several citric acid cycle proteins, such as subunits of the acetyl-CoA carboxylase, aconitate hydratase and fumarase, were found to be significantly reduced. However, more than 30 proteins, mainly related to metal cation utilization, were upregulated.

### *4.2. Inhibition of Bacterial DNA Synthesis and Interaction with Nucleic Acids*

Flavonoids from *Elaeagnus glabra* were tested for their antibacterial activity against *Proteus vulgaris* and *Staphylococcus aureus*. A free 3 ,4 ,5 -trihydroxy B-ring and a free 3- OH group were necessary for antibacterial activity. DNA synthesis was predominantly inhabited by the active flavonoids in *P. vulgaris*, whereas RNA synthesis was inhibited in *S. aureus* [128]. The most active inhibitors of DNA synthesis were robinetin **(122)**, myricetin

**(114)**, and (−)-epigallocatechin **(14)**. It is probable that the B ring of the flavonoids could intercalate or form a hydrogen bond with the stacking of nucleic acid bases and further lead to the inhibition of nucleic acid synthesis in bacteria. The results of Lou et al. [129] demonstrated that *p*-coumaric acid **(166)** had dual mechanisms of bactericidal activity: disrupting bacterial cell membranes and binding to bacterial genomic DNA leading to inhibition of cellular functions, and ultimately to cell death.

Depolarization of membrane and inhibition of DNA, RNA, and proteins synthesis was observed in *S. aureus* and—in higher concentrations—cell lysis, when treated with flavonoids from *Dorstenia* sp., such as 6,8-diprenyleriodictyol **(106)**, isobavachalcone **(148)**, and 4-hydroxylonchocarpin **(149)** [130].

The synthesis of nucleic acid can be inhibited by polyphenols also through topoisomerase inhibition. Flavonoids are inhibitors of topoisomerases, and it plays an important role in their antimycobacterial activity. Docking studies have proved that quercetin **(111)** effectively binds to the subunit B of DNA gyrase through interaction with residues that are in the Toprim domain of the protein. Due to this activity, it inhibited the growth of *Mycobacterium smegmatis* and *Mycobacterium tuberculosis* [131].

Bandele et al. [132,133] have found that polyphenols may act against topoisomerase II in different a manner; (−)-epigallocatechin gallate **(18)** and (−)-epigallocatechin **(14)** were redox-dependent topoisomerase II poisons, kaempferol **(108)** and quercetin **(111)** were topoisomerase II "poisons", myricetin **(114)** utilized both mechanisms, while (−) epicatechin gallate **(18)**, and (−)-epicatechin **(12)** displayed no significant activity. Based on the observation, a set of rules has been formed to predict the mechanism of bioflavonoid action against topoisomerase II: while the C4 -OH in B ring is critical for the compound to act as a traditional poison, the addition of –OH groups at C3 and C5 increases the redox activity of the B ring and allows the compound to act as a redox-dependent poison. The second rule is that the aromatic and planar structure of the C ring in the flavonols that includes a C4-keto group allows the formation of a proposed pseudo ring with the C5-OH. Disruption of these elements abrogated enzyme binding and precluded the ability to function as a traditional topoisomerase II poison [132,133].

Although the above studies were conducted with human cells, and flavonoids are assumed to be poisons of human topoisomerase IIα and IIβ, there are some data about flavonoids as inhibitors of bacterial type II topoisomerases: DNA gyrase and topoisomerase IIA (also called topoisomerase IV) [92]. Gyrases are enzymes that modify the DNA topology, and they are present only in prokaryotes, making them an attractive target for antibacterial drugs. DNA gyrase consists of two catalytic subunits; GyrA is responsible for DNA breakage and reunion, while the subunit GyrB contains the ATP-binding site. Coumarins and cyclothialidines are natural products that inhibit the ATPase activity of DNA gyrase by blocking the binding of ATP to subunit GyrB [134]. Plaper et al. [135] demonstrated that quercetin **(111)** inhibits the supercoiling activity of the bacterial gyrase and induces DNA cleavage, and the mechanism is probably based on interaction with DNA. They showed that quercetin **(111)** binds to the 24 kDa fragment of gyrase B of *Escherichia coli* with a K(D) value of 15 μM and inhibits ATPase activity of gyrase B. Its binding site overlaps with the ATP binding pocket and could be competitively replaced by either ATP or novobiocin. The proposed mechanism is that quercetin **(111)** inhibits gyrases through either the interaction with DNA or with the ATP binding site of gyrase [135]. Other polyphenols that can inhibit bacterial DNA gyrase by binding to the ATP binding site of the gyrase B subunit are catechins, with epigallocatechin gallate **(18)** being the most active, followed by epicatechin gallate **(17)** and epigallocatechin **(14)** [136]. Furthermore, quercetin **(111)**, apigenin **(38)**, and 3,3 ,4 ,6,7-pentahydroxyflavone **(167)** demonstrated inhibitory activity against *Escherichia coli* DNA gyrase [137].

The quantitative structure–activity relationship (QSAR) and molecular docking of flavonoids were analyzed in the study of Fang et al. [138]. The QSAR models demonstrated that hydrophobicity, H-bond donor, steric and electronic properties are key factors for the antibacterial activity of flavonoids. Structure requirements including hydroxyl group at C-3, C-5, C-7 and C-3 , C2-C3 unsaturated double bond and the carbonyl group at C-4 are essential, while the presence of hydroxyl group at C-6, methoxyl group at C-8 and C-3 could decrease the antibacterial activity. Docking results indicated that half of the tested flavonoids inhibited GyrB by interacting with ATP pocket in the same orientation. Polymethoxyl flavones, flavonoid glycosides, and isoflavonoids changed their orientation, resulting in a decrease in inhibitory activity. Hydroxyl group at C-3, C-5, C-7 and C-4 , carbonyl group at C-4 are key active substituents of flavonoids for inhibiting GyrB by interacting with its key residues. Structure changes, including glycosylation, polymethoxylation or isoflavonoids, will change the action mode and result in a decrease in inhibitory activity [138].

Three flavonoids isolated from cottonseed flour which promoted *Escherichia coli* topoisomerase IV-dependent DNA cleavage were identified as rutin **(121)**, quercetin 3-*O*rhamnogalactoside, and isoquercetin **(126)**. Moreover, rutin **(121)** also inhibited topoisomerase IV-dependent decatenation activity and induced the SOS response of a permeable *E. coli* strain [139].

Arima et al. [140] observed that morin alone, at a concentration of 50 μg/mL inhibited the synthesis of DNA in the cells of *Salmonella enteritidis*, while its concentration equaled to a concentration of 12.5 μg/mL was enough if rutin **(121)** was added to the medium at a concentration of 12.5 μg/mL. Morin **(110)** alone also inhibited RNA and protein synthesis, but the rutin added did not influence the inhibition process.

Tannic acid **(215)** is strongly bound to DNA, which possibly had led to the covalent modification of DNA bases. Furthermore, tannic acid in the presence of Cu(II) caused strand cleavage in supercoiled plasmid DNA [141].

### *4.3. Interaction with the Bacterial Cell Wall or Inhibition of Cell Wall Formation*

Various strengths of antimicrobial activity against bacteria may be caused by differences in cell surface structures between Gram-negative and Gram-positive species [20,64]. The major function of the cell wall is to provide shape and cell integrity and to act as an osmotic barrier. Gram-negative bacteria were reported to be resistant toward many antibacterial substances due to the hydrophilic surface of their outer membrane and associated enzymes in the periplasmic space, which is capable of breaking down many molecules introduced from outside [142,143]. Moreover, the negatively charged lipopolysaccharide (LPS) of the outer membrane protects the bacterial cell against catechins [144]. Gram-positive bacteria seem to be more susceptible to the action of phenolic acids than Gram-negative bacteria [64]. One of the explanations is that the Gram-positive bacterium lacks an outer membrane, which would facilitate diffusion of the phenolic acids through the cell wall and intracellular acidification. Vattem et al. [145] postulated the hyperacidification at the plasma membrane interphase, being a consequence of dissociation of phenolic acids, as one of the possible mechanisms of the antimicrobial action of phenolic acids. This hyperacidification would alter cell membrane potential, making it more permeable, and cause irreversible alterations in the sodium-potassium ATPase pump, therefore leading to cell death.

Wu et al. [146] have demonstrated that quercetin **(111)** and apigenin **(38)** influence the synthesis of the bacterial cell walls by the inhibition of D-alanine:D-alanine ligase (an essential enzyme that catalyzes the ligation of d-Ala–d-Ala in the assembly of peptidoglycan precursors). Moreover, these two flavonoids could inhibit the FabZ enzyme from *H. pylori* [122]. Tasdemir et al. [38] found that quercetin **(111)** could inhibit three consecutive enzymes, β-ketoacyl-ACP reductase (FabG), β-hydroxyacyl-ACP dehydrase (FabZ) and enoyl-ACP reductase (FabI), in the FAS II pathway of *Plasmodium falciparum*, whereas apigenin **(38)** could only inhibit FabI.

Flavones form a complex with cell wall components and consequently inhibit further adhesions and microbial growth as well. The inhibition of bacterial enzymes (such as tyrosyl-tRNA synthetase) was observed for C-7-modified flavonoids containing the naringenin **(81)** core [147]. It was also demonstrated that they were also inhibitors of *S. aureus*, *E. coli*, and *Pseudomonas aeruginosa* growth. Baicalein **(45)** was an effective bactericide, and when combined with cefotaxime, the synergistic effects were observed by inhibiting extended-spectrum β-lactamase CTX-M-1 mRNA expression [148]. Inhibition of the bacterial efflux pump and increase in the susceptibility of existing antibiotics (by inducing depolarization of the cell membrane) is another possible mechanism of antibacterial activity. Artonin I **(63)**, from *Morus mesozygia*, was effective against *S. aureus* due to blocking the efflux mechanism and causing depolarization of the cell membrane [149]. Artonin I **(63)** reversed multidrug resistance and increased the susceptibility of existing antibiotics by lowering their minimum inhibitory concentrations.

Many researchers have proved that polyphenols can interact with the cell wall or outer membrane and with their components such as peptidoglycan, lipopolysaccharide. Zhao et al. [150] demonstrated that unlike dextran and lipopolysaccharide, peptidoglycan from *S. aureus* blocked both the antibacterial activity of EGCG **(18)** and the synergism between EGCG and oxacillin, suggesting EGCG may directly bind to the cell wall of *S. aureus* and interfere with its integrity. These results were confirmed by Yoda et al. [113]. As the bactericidal activity of EGCG **(18)** for *S. aureus* was blocked, dose-dependently by purified peptidoglycan, but not by lipopolysaccharide or dextran, it was suggested that EGCG binds directly to the peptidoglycan in the cell wall. These results are consistent with the opinion that the structure of the bacterial cell wall is responsible for the different susceptibilities of Gram-positive and Gram-negative cells to polyphenols.

Gram-negative bacteria (e.g., *Escherichia coli, Salmonella, Shigella*) have in their outer cell membrane a strong "endotoxin"—lipopolysaccharide (LPS). The fraction of procyanidins from cranberries composed of polymers with an average degree of polymerization of 21 can efficiently bind lipopolysaccharide and prevent the interaction of LPS with receptors on the surface of mammalian target cells [151]. On the other side, phenolic extracts of cloudberry and raspberry rich in ellagitannins disintegrated the outer membrane of examined *Salmonella* sp. [152] and released LPS from bacteria cells.

### *4.4. Alteration of Cytoplasmic Membrane Function*

The (inner) bacterial cell membrane is responsible for many essential functions: osmoregulation and respiration processes, transport, biosynthesis and the cross-linking of peptidoglycan and synthesis of lipids [153]. Any disturbance in its structure or functionality can result in metabolic dysfunction and cell death; hence the membrane disruption is postulated to be one of the mechanisms of the antibacterial activity of polyphenols. For example, catechins were shown to rupture the bacterial membrane by binding to the lipid bilayer and by inactivating or inhibiting the synthesis of intracellular and extracellular enzymes [154].

Apigenin **(38)** induced fungal membrane dysfunction and increased cell permeability [155], which caused the release of small intracellular constituents such as ions and sugars, but not proteins. Epicatechin-3-gallate **(17)** and caffeic acid **(167)** targeted both the cell wall and cytoplasmic membrane of *P. aeruginosa* [156]. The cellular membrane destruction and ensuing membrane permeability perturbation of *P. aeruginosa* had led to the ascending access of hydrophobic antibiotics, a release of potassium ions, and leakage of nucleotides. Phenolic acids, due to their partially lipophilic nature, pass through the cell membrane by passive diffusion and cause an increase in membrane permeability. They possibly reduce the intracellular pH and induce protein denaturation [157]. Methanol extract of *Coriolus versicolor* rich in polyphenols disabled *S. aureus* cell division (i.e., the formation of septa) and led to the accumulation of peptidoglycan and teichoic acid precursors in the cytoplasm [157]. In this case, the extract acted directly on the cytoplasmatic membrane, whereas in Gram-negative *Salmonella* Enteritidis, the cell envelope was damaged. On the other side, at high concentrations, catechins were found to generate an oxidative burst by the generation of reactive oxygen species (ROS) that cause alteration in the membrane permeability and membrane damage [158]. Purified flavonoids from *Graptophyllum glandulosum* possessed antimicrobial activities against multidrug-resistant

*Vibrio cholerae* and caused cell lysis and disruption of the cytoplasmic membrane upon membrane permeability [159].

Flavonoids (acacetin **(40)** and apigenin **(38)**) and flavonols (morin **(110)** and rhamnetin **(115)**) caused destabilization of the membrane structure by disordering and the disorientation of the membrane lipids and induced leakage from the vesicle [160]. The inverse correlation between the number of hydroxyl groups in the flavonoids and their capacity to leakage induction was noted. Studies of Chabot et al. [161] suggested that flavonoids lacking hydroxyl groups on their B rings (genistein **(65)**, hesperetin **(83)**, chrysin **(47)**, galangin **(109)**) were more potent inhibitors of microbial growth than those with the –OH groups. On the other hand, Adamczak et al. [67] have demonstrated that the presence of hydroxyl groups in the phenyl rings A and B usually did not influence the level of the antibacterial activity of flavones. A significant increase in the activity of the hydroxy derivatives of flavone was observed only for *S. aureus*. What is interesting, in contrary to other studies, the compounds tested in the study were generally more active against Gram-negative bacteria: *Escherichia coli* and *Pseudomonas aeruginosa* than Gram-positive ones: *Enterococcus faecalis* and *Staphylococcus aureus*.

Tsuchiya [162] reported that the catechin impact on membrane fluidity also depended on the stereospecificity. (−)-Epicatechin **(12)**, (+)-epicatechin **(9)**, (−)-catechin **(10)** and (+)-catechin **(7)** reduced membrane fluidity in increasing order of intensity; it means that epicatechins in a *cis* form were more effective for reducing membrane fluidity than catechins in a *trans* form. Stereospecificity in the membrane effects of catechin stereoisomers may be induced by the different hydrophobicity of geometrical isomers and the chirality of membrane lipid components. Lipophilic flavonoids may also disrupt microbial membranes [107]. It was suggested that the mode of action of terpenes and their related alcohols involves disruption of microbial membranes by their lipophilic components [163]. According to Tsuchiya [164], bioactive components with amphiphilic or hydrophobic structures interact with biological membranes resulting in the modification of membrane fluidity, microviscosity, order, elasticity, and permeability. The author postulated that interactions of flavonoids with lipid bilayers involve two mechanisms; the first is associated with the partition of the more nonpolar compounds in the hydrophobic interior of the membrane, while the second one includes the formation of hydrogen bonds between the polar head groups of lipids and the more hydrophilic flavonoids at the membrane interface. The membrane interactions and localization of flavonoids play a vital role in altering membrane-mediated cell signaling cascades [165].

The studies of Arora et al. [166] demonstrated that flavonoids and isoflavonoids preferentially enter into the hydrophobic core of membranes. In plant tissues, flavonoids occur mainly in the form of glycoside, and the presence of glycosidic residues in the flavonoid skeleton influences the hydrophobicity of flavonoids. Flavonoids with a greater hydrophobicity have been reported to influence the transmembrane potential to a greater extent than the less hydrophobic flavonols, probably because they can enter deeper into the lipid bilayer, thereby disrupting the compact packing of lipids [157]. Moreover, the spatial configuration is also important; a substantially higher affinity for artificial membranes was reported for flavonols (planar) than flavanones (tilted) [167]. Using the fluorescence anisotropy technique, it was reported that naringenin and naringin enhanced membrane fluidity, while membrane interaction with quercetin **(111)**, daidzein **(64)**, luteolin **(49)**, galangin **(109)**, kaempferol **(108)** and genistein **(65)** resulted in rigidified membranes [165]; however, the impact depended on the lipid composition of membranes.

Wu et al. [168] have shown the positive correlation between antibacterial capacity and membrane rigidification effect of the polyphenolic compounds. Authors have observed that flavonoids decreased the membrane fluidity with the potency being kaempferol **(108)** > chrysin **(47)** > baicalein **(45)** > quercetin **(111)** > luteolin **(49)**, whereas isoflavonoids increased the membrane fluidity with the potency being puerarin **(70)** > ononin **(75)** > daidzein **(64)** > genistin **(68)** [168]. Kaempferol **(108)**, located deeply in the hydrophobic core of the lipid bilayer, decreased the membrane fluidity most and exhibited the highest

antibacterial capacity against *E. coli*. The number and the position of hydroxyl groups influenced the membrane interaction with polyphenols; the OH group at C-3 in the C ring was important for decreasing membrane fluidity. He et al. [169] suggested that for flavonoids to be effective antimicrobial agents, interaction with the polar head–group of the model membrane followed by penetration into the hydrophobic regions must occur. The antimicrobial efficacies of the flavonoids were consistent with liposome interaction activities and decreased in the order: kaempferol **(108)** > hesperetin **(83)** > (+)-catechin **(7)** > biochanin A **(76)**.

Sophoraflavanone B (8-prenylnaringenin) **(101)** caused cell wall weakening, membrane damage and intracellular constituents leaking from the cell of methicillin-resistant *S. aureus* [170]. In the study, the direct binding of sophoraflavanone B **(101)** to peptidoglycan was demonstrated. It also has been proposed that sophoraflavanone G **(98)** and (−)-epigallocatechin gallate **(18)** inhibited cytoplasmic membrane function [171].

### *4.5. Inhibition of Energy Metabolism*

Many aspects of cellular metabolism revolve around ATP production and consumption, and ATP is regarded as the universal energy exchange factor that connects anabolism and catabolism but also enables processes such as motile contraction, phosphorylation, and active transport (uptake) of nutrients. Membrane-bound F1F0 ATP synthase from bacteria is an enzyme responsible for ATP production through oxidative phosphorylation or photophosphorylation. It has been demonstrated that morin **(110)**, baicalein **(45)**, silibinin **(138)**, and epicatechin caused complete inhibition of ATPase activity, while hesperidin **(86)**, chrysin **(47)**, kaempferol **(108)**, diosmin **(51)**, apigenin **(38)**, genistein **(65)**, or rutin **(121)** exert partial inhibition of about 40–60% [172].

Dadi et al. [173] demonstrated that resveratrol **(202)**, piceatannol **(203)**, quercetin **(111)**, quercitrin **(123)**, or quercetin-3-β-D glucoside **(126)** inhibited *E. coli* ATP synthase, but to different degrees. The most potent inhibitor was piceatannol **(203)** (~0 residual activity); inhibition by other compounds was partial and ranged from ~20% residual activity for quercetin **(111)** to ∼60% residual activity for quercitrin **(123)** or resveratrol **(202)**. Inhibition was identical in both F1F0 membrane preparations as well as in isolated, purified F1, but in all cases, inhibition was reversible. Interestingly, resveratrol **(202)** and piceatannol **(203)** inhibited both ATPase and ATP synthesis, whereas quercetin **(111)**, quercitrin **(123)** or quercetin-3-β-D glucoside **(126)** inhibited only ATPase activity and not ATP synthesis. The membrane-bound ATPase activity of *E. coli* was also inhibited by eugenol **(185)** or carvacrol **(183)** [174]. Similar results were obtained for thymoquinone that completely inhibited both purified F1 and membrane-bound F1F0 *E. coli* ATP synthase, and the process of inhibition was fully reversible [175].

It has been proposed that licochalcones A **(150)** and C **(151)** can inhibit energy metabolism [171]. Haraguchi et al. [176] have proved the antibacterial activity of licochalcones A **(150)** and C **(151)** against *S. aureus* and *Micrococcus luteus*, which resulted from inhibited oxidation of NADH in bacterial membranes. As licochalcones inhibited NADH-cytochrome c reductase, they exerted their antibacterial activity by inhibiting the bacterial respiratory electron transport chain.

EGCG **(18)** directly interacts with proteins and phospholipids in the plasma membrane and regulates signal transduction pathways, transcription factors, DNA methylation, as well as mitochondrial function and autophagy [177].

### *4.6. The Inhibition of Biofilm Formation and Interfering with Bacterial Quorum Sensing*

Biofilm is an assemblage of microbial cells that are irreversibly linked to a surface with bacteria embedded in an extracellular matrix of self-produced biopolymers. The ability to form biofilm is an important property of various bacteria such as pathogenic species and is associated with quorum sensing (QS) or cell-to-cell communication. Therefore, bacterial cell-to-cell communication has received attention to manifest the role of quorum signals in the attachment and growth of pathogenic bacteria in foods. QS participates

in the biofilm formation as well as controls the expression of various virulence factors, inter alia, the production of proteases that degrade connective tissue, the production of siderophores that facilitate iron uptake, the releasing of toxins that disrupt cellular processes, the formation of phenazines that favor the reactive oxygen species generation, and production of exopolysaccharides that are necessary for the phagocytosis-resistant capsules structure [178].

An anti-quorum sensing (anti-QS) agent curcumin **(216)** from *Curcuma longa* (turmeric) was shown to inhibit the biofilm formation of pathogens, such as *Escherichia coli*, *Pseudomonas aeruginosa* PAO1, *Proteus mirabilis* and *Serratia marcescens*, possibly by interfering with their QS systems [179], because the biofilm maturation was disturbed by a biomass reduction and by the interruption of swimming motility. Chlorogenic acid **(172)** was proved to significantly inhibit the formation of biofilm by *P. aeruginosa*, its ability to swarm, and virulence factors including protease and elastase activities and rhamnolipid and pyocyanin production. Moreover, the QS related genes were downregulated in *P. aeruginosa*, and the inhibitory rates were as follows: lasI 85.09%, lasR 48.63%, rhlI 27.98%, rhlR 34.7%, pqsA 73.08%, and pqsR 45.85%, respectively [180].

Quercetin **(111)** efficiently reduced the biofilm formation and other QS regulated phenotypes like violacein inhibition, exopolysaccharide production and alginate production in foodborne pathogens *K. pneumoniae*, *P. aeruginosa*, and *Y. enterocolitica* [181]. Furthermore, quercetin **(111)** significantly inhibited the swimming and swarming behavior of *P. aeruginosa* and *Y. enterocolitica*.

*L. paracasei* exposed to resveratrol **(202)** displayed changes in the physicochemical properties of their surface, especially with a global increase in negative charges, a more basic nature and an increase in their hydrophobicity. These changes may largely contribute to the enhanced adhesion, induced formation of bacterial aggregates and biofilm formation abilities of resveratrol-treated *L. paracasei* [182]. However, the majority of studied polyphenols have shown the opposite impact on the ability of biofilm formation. Apple flavonoid phloretin **(152)** was reported to control *E. coli* O157:H7 biofilm formation by a mechanism that implies repressing the curli genes (csgA and csgB), which are involved in fimbriae production [183]. Epigallocatechin-3-gallate **(18)** eliminates the biofilm matrix by directly interfering with the assembly of curli subunits into amyloid fibers and by triggering the σ<sup>E</sup> cell envelope stress response and thereby reducing the expression of a crucial activator of curli and cellulose biosynthesis (CsgD) [184]. Recently, it has been shown that EGCG **(18)** act against biofilms by strongly interfering with the assembly of amyloid fibers and the production of phosphoethanolamine-modified cellulose fibrils [185].

EGCG **(18)** also inhibited the formation of *Streptococcus mutans* biofilms [186,187]. The growth of *Streptococcus mutans* decreased, and the biofilm formation was inhibited by pinocembrin **(95)**, apigenin **(38)**, quercetin **(111)**, while caffeic acid phenethyl ester decreased, probably due to changes in bacterial architecture [188].

According to Xu et al. [189], the ECGC **(18)** activity against *S. mutans* is due to disrupting at the transcriptional level the adherence of bacteria to surfaces and hence inhibiting the biofilm formation. Authors hypothesized that EGCG at sublethal concentrations directly suppressed the expression of gift genes encoding glucosyltransferases, enzymes that synthesize polysaccharides necessary for biofilm formation [189]. Moreover, EGCG **(18)** was found to inhibit the enzymatic activity of the F1F0-ATPase and lactate dehydrogenase [190].

Morin **(110)** at its sub-MICs demonstrated a significant dose-dependent inhibitory efficacy against *Listeria monocytogenes* biofilm formation [191]. Moreover, morin-treated *Listeria* showed a significant reduction in hemolysin secretion and a concentration-dependent decrease in the flagella directed swimming and swarming velocity. The biofilm formation and biofilm-related genes in *L. monocytogenes* also were inhibited by thymol **(184)**, carvacrol **(183)** and eugenol **(185)** [192].

Bap (biofilm-associated protein) is expressed by *Staphylococcus* sp. in order to adopt functional amyloid-like structures as scaffolds of the biofilm matrix. Quercetin **(111)**, myricetin **(114)** and scutellarein **(44)** specifically inhibited Bap-mediated biofilm formation

of *S. aureus* and other staphylococcal species [97] by preventing the assembly of Bap-related amyloid-like structures.

The ability of bacteria to adhere was also inhibited by phenolic acids. Adhesion was less favorable when the bacteria were exposed to gallic acid **(159)** (*P. aeruginosa*, *S. aureus* and *L. monocytogenes*) and ferulic acid **(168)** (*P. aeruginosa* and *S. aureus*). Both phenolics were able to inhibit bacterial motility and prevented biofilm formation, as well as reducing the mass of biofilms formed by the Gram-negative bacteria [193]. Further studies proved that gallic **(159)** and ferulic **(168)** acids led to irreversible changes in membrane properties, such as its charge, intra and extracellular permeability, and physicochemical properties. Both acids caused changes in hydrophobicity and negative surface charge and induced the local rupture or pore formation in the cell membranes leading in consequence to essential intracellular constituent leakage [194].

### *4.7. Substrate Deprivation*

When polyphenol forms complex with protein, the biological function might change. Depending on the function of a complexed protein, the influence on bacteria cells will differ. As mentioned above, a decreased or inhibited activity of enzymes can result in a lack of energy for bacteria and, in consequence, might lead to cell death. Lack of energy also means a disturbed transport of nutrients across the cell wall and cytoplasmic membranes, diminished bacteria proliferation and limited mobility, as well as inhibited ability to biofilm formation or even blocked sporulation [195].

The deprivation of the substrates required for microbial growth, especially essential mineral micronutrients such as iron and zinc (via proanthocyanidin chelation with the metals), together with the destabilization of the cytoplasmic membrane, the permeabilization of the cell membrane, the inhibition of extracellular microbial enzymes, direct actions on microbial metabolism, were supposed to be the mechanism of the antibacterial activity of the A-type proanthocyanidin **(23,24)** [196].

Scalbert [197] suggested that tannin toxicity for bacteria is due to the direct impact on bacterial metabolism by inhibiting the oxidative phosphorylation as well as by deprivation of the substrates required for microbial growth, especially an iron deprivation. Generally, tannins are reported to be strong inhibitors of many various hydrolytic enzymes such as α-amylase, pectinase, cellulase, xylanase, lactate dehydrogenase, malate dehydrogenase, peroxidase, β-glucosidase, so they can inhibit the activity, growth or proliferation of microorganisms [198]

The inhibitory effect of tannic acid **(215)** on the growth of intestinal bacteria may be due to its strong iron-binding capacity. The growth of *E. coli* was restored by the addition of iron to the medium after the precipitate caused by tannic acid **(215)** was removed [199]. In the study, neither *Bifidobacterium infantis* nor *Lactobacillus acidophilus* required iron for growth, which probably contributes to their resistance to tannic acid. It is known that only a few bacteria, including lactobacilli, do not require iron. It is an essential trace element for most gut bacteria, and many have active Fe transport systems and other mechanisms to scavenge Fe [200]. For example, *Bacteroides* spp. are highly dependent on heme and iron, whereas many members of the Enterobacteriaceae have developed mechanisms, including siderophores, to acquire Fe in competition with other bacteria and the host.

It is well known that catechol and gallol structure (Figure 3) and hence many polyphenolic compounds are effective metal chelators. After deprotonation, which is required for metal binding, catecholate and gallate groups may be complexed with metal ions that prefer octahedral geometry, such as Fe2+ and Fe3+ (Figure 3) [201].

**Figure 3.** Expected octahedral coordination geometry of general iron-polyphenol complexes, (**a**) gallol, (**b**) catechols. Coordination requires deprotonation of the polyphenol ligands. Based on [201].

It was established for flavones and for the flavanone naringenin **(81)** that the binding metal sites are preferentially at the 5-hydroxyl and 4-oxo groups [202]. On the other hand, the study of Mladenka et al. [203] demonstrated that the most effective iron-binding site of flavonoids is 6,7-dihydroxy structure, present, for example, in baicalein **(45)**. The simultaneous presence of 3-hydroxy-4-keto conformation, 2,3-double bond and the catecholic B ring were associated with significant iron chelation; however, the catecholic B ring did not play an essential role in more acidic conditions. Quercetin **(111)** and myricetin **(114)** that contain all mentioned structural requirements, had activity similar to baicalein **(45)** at the neutral conditions but were clearly less active in lower pH. On the other hand, baicalein **(45)**, additionally possessing the 6,7-dihydroxyl groups, was very efficient even in the acidic condition. The 5-hydroxy-4-keto configuration has only moderate activity at all pH conditions. It was also proved that isolated keto, hydroxyl, methoxyl groups or ortho methoxy–hydroxy groups were not associated with iron chelation at all.

Polyphenols also have strong binding interactions with Cu2+, and stability constants for Cu2+ catecholate complexes are even larger than for Fe2+ [201].

As bacteria contain various metalloenzymes, flavonoids by binding metal ions can inhibit their activity and lead to various metabolic disorders (enzyme inhibition, impairment of ion channel functions). Furthermore, the metabolic functions of the human gut microbiota that involve metalloenzymes may also be altered [204]. An enzyme methionine aminopeptidase (MetAP) carries out the removal of the initiator methionine residue from newly synthesized proteins, and this removal is critical for the activation, distribution and stability of many proteins. It was proved that the adjacent hydroxyl groups on the phenyl ring (catechol moiety) were essential for effective inhibition of the Fe (II)-a form of *E. coli* MetAP and growth inhibition of bacterial cells [205].

Polyphenols can also cause iron deficiency in the digestive tract, which will affect sensitive bacterial populations and change the composition of the intestinal microbiota. Oral bacterium Fusobacterium nucleatum is associated with colon cancer, causes erythrocytes lysis, and therefore releases hemoglobin, which provides an iron source to bacteria and other periodontopathogens, promoting their proliferation in periodontal pockets [206]. The tea polyphenols were proved to inhibit dose-dependently the hemolytic activity of *F. nucleatum*.

The virulence factors such as gelatinase, collagen-binding antigen, cytolysins, and proteases enhance colonization, survival and persistence of *E. faecalis* in the root canal. Treatment of *E. faecalis* with a sublethal concentration of EGCG **(18)** (2.5 mg/mL) significantly inhibited the expression of responsible genes (collagen adhesin (ace), cytolysins activator (cylA), gelatinase (gelE) and serine protease (sprE)) by >75% compared to the untreated control [207]. The elastase, protease and pyocyanin production in *P. aeruginosa* were inhibited by curcumin **(216)** in a dose-dependent manner [179]. EGCG **(18)** caused the inhibition of glucose uptake by *E. coli*, which can suggest that EGCG inhibits the major function of porin proteins, namely the passive transport of small hydrophilic molecules such as glucose, leading to growth inhibition of *E. coli* [114].

#### *4.8. The Relationship between Polyphenols Structure and Antibacterial Activity*

The mechanism of inhibition by polyphenols may differ depending both on the structure of the polyphenolic compound and bacteria species. The amphipathic character of flavonoids plays a very important role as hydrophilic and hydrophobic moieties must be present together and well-spaced in these compounds [208].

Flavans with prenyl group at the A ring were potent antibacterial compounds against *Staphylococcus aureus*, and the number and position of prenyl groups on this ring influenced the activity [91].

The number of hydroxyl groups in the B ring in flavonols and flavones is associated with the antimicrobial activity against lactic acid bacteria (LAB). Myricetin **(114)** is a flavonol possessing three hydroxyl groups in the B ring as pyrogallol structure, whereas quercetin **(111)** and kaempferol **(108)** have one and two hydroxyl groups less in the B ring than in myricetin, respectively. Myricetin, as a pure compound, significantly inhibited the growth of all tested LAB that originated from the human gastrointestinal tract, as well as the Gram-positive *Enterococcus faecalis* and *Bifidobacterium lactis*, while quercetin **(111)** and kaempferol **(108)**, with a more lipophilic nature, had no inhibitory impact on the above bacteria [47]. Flavone luteolin **(49)** has a structure similar to quercetin **(111)**, but it lacks the OH group at position 3 in ring C. Luteolin **(49)** was bacteriostatic against some of the tested LAB as well as against *E. faecalis* and *B. lactis*, while other flavone apigenin **(38)**, which has one hydroxyl group less in the B ring had no such effects.

Baicalein (flavon) **(45)** and myricetin (flavonol) **(114)** show the most significant antibacterial effects among the tested flavonoids. Both have a pyrogallol structure, but baicalein in ring A (5, 6, 7–OH) and myricetin in ring B. Results proved that the pyrogallol structure was an important element for the potent antibacterial activity for flavonoids [60]. Echeverria et al. [208] made the comparison between a flavone (planar) and flavanone (not planar) with similar lipophilicity and oxygenated substitution patterns in the A and B rings (e.g., pinocembrin **(95)** and 3-*O*-methylgalangin **(113)**) and showed that flavones have higher antibacterial activity. On the other hand, possessing at least one hydroxy group in the ring A (especially at position C-7) seems to be crucial for antibacterial activity of flavones, and an additional OH group in another position such as C-5 and C-6 can further increase the activity [91].

All the flavonols and flavanones with antibacterial activities had two hydroxyl substituents on C-5 and C-7 of ring A in common, such as quercetin **(111)**, rutin **(121)**, naringenin **(81)**, and hesperetin **(83)** [60]. Moreover, the authors suggest that flavanones were more active than the corresponding flavones. For example, naringenin **(81)** showed antibacterial effects on all the tested bacteria, whereas apigenin **(38)** showed almost no effect. Such results indicate that the saturation of the C2-C3 double bond increased the antibacterial activity.

On the other side, Wu et al. [168] demonstrated that flavonoids were more effective *E. coli* inhibitors than isoflavonoids with relative activity being as follows: kaempferol **(108)** > quercetin **(111)** > chrysin **(47)** > luteolin **(49)** > baicalein **(45)** > tangeretin **(57)** and daidzein **(64)** > genistin **(68)** > ononin **(75)** > puerarin **(70)**. The only structural difference between quercetin **(111)** and luteolin **(49)** is that quercetin has a hydroxyl group at position 3 in the C ring, while luteolin has none. It means that the 3-OH group is important to the activity of flavonoids against Gram-negative bacteria *E. coli*. Further analysis of structure–activity relationships revealed that the methylation of OH groups could decrease the antimicrobial activity of flavonols. It also has been shown a significant positive correlation between the antibacterial capacity of flavonoids and the membrane rigidification effect. A quantitative structure–activity relationship (QSAR) study revealed that the activity of the flavonoid compounds could be related to molecular hydrophobicity and charges on the C atom at position 3 [168].

The hydrophobic substituents such as prenyl groups, alkylamino chains, alkyl chains, and nitrogen or oxygen-containing heterocyclic moieties usually enhance the antibacterial activity for all the flavonoids [98]. It was concluded that hydroxyl groups on special sites are favorable for antibacterial activity, such as 5,7-dihydroxyl substitution for flavone and flavanone and 2 or 4 hydroxylation for chalcones. The hydroxyl group at position three on the C ring of flavone also increased the activity. However, the methylation of the hydroxyl groups generally decreased the activity. The lipophilicity of ring A is therefore of great importance for the activity of chalcones. In addition, hydroxy groups at 4 , 4, and 6 of A and B rings increase the activity of chalcones [91].

The substitution of the flavonoid ring system with prenyl groups increases the lipophilicity of the molecule and results in a strong affinity to biological membranes. Prenylated flavonoids, i.e., featuring C5 isoprenoid substituents, have a relatively narrow distribution in the plant kingdom and are constitutively expressed in plants, as compared with prenylated isoflavonoids, which are produced in response to an attack or damage [209]. Xanthohumol **(145)** is the main component (80–90% of the total flavonoids) and is the most abundant prenylated chalcone in hops. It exerted high antimicrobial activity against *Bacteroides fragilis, Clostridium perfringens* and *Clostridium difficile* [210]. β-bitter acids (lupulones) were less effective, and the least effective against anaerobic pathogens were α-bitter acids (humulones). Xanthohumol **(145)**, naringenin **(81)**, chalconaringenin **(140)** and 4 hydroxy-4 -methoxychalcone inhibited the growth of *S. aureus* [211]. The presence of at least one hydroxyl group and especially at the C-4 position was crucial for the antibacterial activity against *S. aureus*. The lack of hydroxyl group or its replacement by a halogen atom (–Cl, –Br), nitro group (–NO2), ethoxy group (–O–CH2CH3), or aliphatic groups (–CH2CH3), (–CH3) led to inactivation of the compounds. Prenylated flavonoids, such as artocarpin **(62)** and isobavachalcone **(148)**, exhibited strong antibacterial activity towards *B. cereus*, *E. coli*, and *Pseudomonas putida* or only Gram-positive species, respectively [212]. It has been demonstrated that any isoflavonoid modification that results in the absence or cyclization of the prenyl group decreases the antibacterial activity of the compound.

Campos et al. [213] had demonstrated that hydroxycinnamic acids (p-coumaric **(166)**, caffeic **(167)** and ferulic **(168)** acids) induced greater potassium and phosphate leakage than hydroxybenzoic acids (protocatechuic **(157)**, gallic **(159)**, and vanillic **(161)** acids) across the membranes of *Oenococcus oeni* and *Lactobacillus hilgardii*.

Flavonoids can occur in two forms: free as "aglycons" or in the form of "glycosides", where an aglycon is combined with sugar moiety ("glycone"). Flavonoid glycosides occur in a diet generally in ring A or C as O-glycosides, and a corresponding substitution in ring A has a far greater impact on activity [180]. Aglycones of most flavonoids are more hydrophobic than their glycosides [91]. Both the number of glycosylation as well as the position and structure of saccharides are of great significance for the antioxidant, antibacterial, anticancer, anti-inflammatory and antidiabetic activity of a compound [180]. It has been postulated that glycosylation of flavonoids enhances antimicrobial activity, but their antioxidant, anti-inflammatory, anticancer and cardioprotective properties decreased [180]. However, it seems that the impact of glycosylation on antibacterial activity depends on the flavonoid class as well as the position at which sugar moiety is added. The results of Duda-Chodak [63] demonstrated that flavonoid aglycones, but not their glycosides, may inhibit the growth of some intestinal bacteria. In this study, rutin (quercetin 3-*O*-rutinoside) **(121)** had no inhibitory influence on the intestinal bacteria analyzed, and even slight stimulation of the growth of *Lactobacillus* spp. was observed. In contrast, its aglycone quercetin **(111)** exerted a dose-depended inhibitory effect on intestinal bacteria (except on *Bifidobacterium catenulatum*), and this was especially strong on *Ruminococcus gauvreauii*, *Bacteroides galacturonicus* and *Lactobacillus* spp. growth. The same was true for flavanones. Naringin **(85)** and hesperidin (flavanone 7-*O*-glycosides) **(86)** had no impact, but their aglycones (naringenin **(81)** and hesperetin **(83)**, respectively) inhibited the growth of almost all bacteria analyzed. A similar result, showing that 7-*O*-glycosylation of flavanones (naringenin and hesperetin) and flavones (baicalein **(45)**) decreased the antimicrobial activity against *E. coli*, *S. aureus*, *S. typhimurium*, *Enterobacter sakazakii* and *Vibrio parahemolyticus* were demonstrated by Xie et al. [60]. The opposite results were obtained by Adamczak et al. [67]; flavonol aglycones kaempferol **(108)** and quercetin **(111)** displayed a moderate activity only against *E. coli*, while quercetin 3-*O*-rutinoside **(121)** demonstrated inhibitory influence on all strains tested.

Docking results have revealed that the substitution of galloyl or glycosides at position 3 of heterocyclic pyrane ring in flavonoids enhanced the binding affinity to three targets, i.e., fumarate reductase flavoprotein subunit (FrdA), dihydroorotate dehydrogenase (PyrD) and NADH-dependent enoyl-ACP reductase (FabI). Such a phenomenon was observed for flavonoids and their glycosides; quercetin 3-rhamnoside **(123)** and myricetin 3-galactoside **(127)** were more potent inhibitors to PyrD, FabI, and DYR than quercetin **(111)** and myricetin **(114)**, respectively [121]. One of the most potent bacterial inhibitors among flavan-3-ol is EGCG **(18)**, possessing both pyrogallol and galloyl structures in a moiety. EGCG is a stronger inhibitor of pathogens than other flavan-3-ols having fewer or no galloyl groups and pyrogallols. Antibacterial activity of tea flavan-3-ols was in decreasing order EGCG **(18)** > ECG **(17)** > EC **(12)** ≥ theaflavins ≥ gallic acid **(159)** > EGC **(14)** against *S. aureus* and *P. aeruginosa* [214]. The importance of these free galloyl groups for antibacterial activity was also proved in the study of Puljula et al. [215]. Salicarinin A **(213)** and rugosin D **(211)** possess many free galloyl groups, inhibited the growth of *S. aureus* completely at a 0.5 mM concentration. Other ellagitannins, with lower numbers of galloyl or pyrogallol substituents, were less effective.

#### *4.9. The Impact of Food Matrix on Polyphenol Activity*

There are large discrepancies between the results, i.e., in some studies, it is shown that a given polyphenol class inhibits bacteria, and in others, it does not affect or even stimulates their growth. This may be due to the structure of the used polyphenol, including the molecule size, number and position of hydroxyl groups, their substitutions, the presence/absence and position of glycosylation, hydrophobicity and hydrophilicity of the moiety, and others. The observed discrepancies could also be attributed to the

changes in the structure of polyphenols when dissolved in various solvents (water, ethanol, methanol, organic solvents and their mixtures) or after their addition to the medium with bacteria. It is because polyphenols do not dissolve in every solvent, and they can precipitate (affecting the actual concentration of the tested compound) after changing the solvent [216–218]. Polyphenols also have different rates of diffusion depending on the medium and environmental conditions.

Moreover, the type of microorganism (Gram-positive, Gram-negative, anaerobic or aerobic or microaerophile, etc.) also has a significant impact when the activity of polyphenols is assessed. It should be borne in mind that used assays, analytical methods, as well as conditions and incubation time, strains of microorganisms, inoculum size, and even concentrations of tested polyphenol may differ between scientific laboratories. There are also big differences between results when the impact of polyphenol on intestinal bacteria is assessed using pure polyphenols solution, plant extract containing polyphenols mixture of whole food in which polyphenols are bound to a food matrix. When in vitro studies are performed, usually pure cultures of bacteria are tested, and interactions with other members of gut microbiota, the impact of human digestive enzymes, the host health, or interactions with other components of a meal are not taken into account. However, all mentioned factors are important for the final results.

It is obvious that the results obtained from in vivo and in vitro studies should not be compared directly. When in vivo studies are conducted, the scientists introduce an ingredient into the diet and analyze changes in the abundance or composition of the gut microbiota, usually focusing on the effect on the entire bacterial population rather than on individual species. In such experiments, many factors contribute to the final results: the chemical composition of the food matrix, the bioaccessibility of polyphenols, their bioavailability, the interactions between particular bacterial strains present in the gut, the health of consumers and many more. Depending on the polyphenols present in the plant, different effects can be achieved because each polyphenol reacts differently with the components of plant tissues. Moreover, each plant differs in its composition. Tarko and Duda-Chodak [219] proved the differences between the bioaccessibility of polyphenolic compounds originating directly from fruits (black chokeberry, elderberry, hawthorn, Cornelian cherry, apple and Japanese quince) and that of those present in the fruit extracts during their digestion conducted in a simulated human gut. They proved significant differences in polyphenols bioavailability that resulted from their interactions with food matrixes. It was caused by polyphenols bounding to the matrix, which is known to modify the polyphenols extractability and susceptibility to digestive enzymes and bacterial metabolism [220]. The interaction with the food matrix also modulates the impact of polyphenols on bacteria inhabiting the colon.

During in vivo studies, it should also be considered that some polyphenols present in a diet are absorbed before they reach the colon, and hence, do not influence the microbiota. For example, quercetin glycosides can undergo partial hydrolysis by pepsin during their passage through the stomach [221], and the released aglycone quercetin **(111)** may be then absorbed in the stomach and secreted in the bile. Glycosides of other flavonoids can be hydrolyzed to aglycones in the small intestine due to the activity of human digestive enzymes, such as lactase phlorizin hydrolase and cytosolic β-glucosidase. It refers to the glycosides that contain glucose, xylose or galactose; as mentioned, humans enzymes e have an affinity for those sugars. It means that only polyphenols resistant to the action of human enzymes are not absorbed in the small intestine and pass to the colon, where they may exert their inhibitory or stimulatory activity towards microbiota, or they may be cleaved by bacterial enzymes to produce derivatives and metabolites of various activity.

Another important issue is the diversity of the chemical composition of plant tissues. For example, chokeberries and apples contain much higher amounts of pectin than the elderberry fruit, which resulted in small amounts of polyphenols in the sediment obtained after elderberry digestion [219]. Further, fruits of the Cornelian cherry are rich in pectin and also in low-molecular-weight phenolic acids that can firmly bind with pectin and so pass to the colon intact [219]. However, the differences between the food matrix could also be related to the cell wall composition of the fruit, resulting in an observed different bioaccessibility of polyphenols present in apples, chokeberries and Japanese fruit [219]. The flesh of Japanese quince fruits contains much pectin, whereas, in the cell walls, cellulose dominates [222]. On the other hand, apples are rich both in pectin and cellulose, but they also contain lignin [223]. The presence of lignins was believed to reduce the proanthocyanidin adsorption in skin cell walls when compared to that of the flesh cell walls [224], causing that unbound proanthocyanidins were more sensitive to enzymatic digestion and acidic pH in the stomach.

Proanthocyanidins are of neutral charge, so they are easily absorbed by the cell wall polysaccharides, while anthocyanins—which are positively charged molecules—could rather selectively bind to a negatively charged pectin [225]. The ratio of bound to free proanthocyanidins depends mainly on their concentration and degree of polymerization. The susceptibility of anthocyanins, anthocyanidins, and proanthocyanidins to digestion can also depend both on the structure of the cell wall polysaccharide network in fruits and the structure of pectin. Voragen et al. [226] have demonstrated that 47% of the structural elements of pectin in apples are neutral side chains, while in bilberry or black currant, more than 60% are homogalacturonan]. Yet another structure was reported for Japanese quince pectin, which consisted of four different populations, mainly arabinans and highly methylated homogalacturonans [227]. The simultaneous presence of pectin, cellulose and hemicellulose in food favors the bounding of procyanidins and anthocyanins and protects them against digestive enzyme activity. In consequence, they are not released from the food matrix at this digestion stage. Moreover, during proanthocyanidins degradation, free (+)-catechin **(7)** could be released, which can bind effectively to cellulose [228].

Tarko and Duda-Chodak [219] also revealed that procyanidin B1 in hawthorn was almost insensitive to digestive enzymes, and probably the saponins, which presence in the hawthorn fruit is characteristic, had such a protective impact. Saponins are poorly absorbed in the intestine mainly due to their unfavorable physicochemical traits, such as large molecular mass (>500 Da), high hydrogen binding capacity (>12), and high molecular flexibility (>10).

Concluding, the presence/absence of the food matrix, as well as its chemical composition, can affect the bioaccessibility, bioavailability and biological activity of polyphenols and their bidirectional interactions with the intestinal microbiota.

### **5. Polyphenols Biotransformation by Intestinal Bacteria**

It is believed that only undigested and unabsorbed polyphenols can reach the large intestine and exert their impact on bacteria inhabiting there. As described above, many of the polyphenols can inhibit the growth of microbiota residing in the colon. However, some of the phenolic compounds act as prebiotics and stimulate the growth of particular species. Hence, polyphenols modulate the composition of human gut microbiota. On the other hand, only unabsorbed polyphenols can undergo biotransformation during the activity of bacterial enzymes. Products of bacterial metabolism can further be metabolized to various derivatives and absorbed into the human body [15].

Due to the great diversity of species forming the intestinal microbiota in different individuals, the profile of polyphenol metabolites that are generated and their final effect on the body are highly variable within the human population. The dietary polyphenols can be metabolized by various pathways leading to the formation of a number of different phenolic derivatives characterized by small and low molecular weight as well as a modified biological activity. For example, aglycones and oligomers are released by microbial glycosidases and esterases, which enhances their absorption [229]. On the other side, released aglycones can inhibit intestinal microbiota growth and activity, preventing the metabolism of other polyphenolic compounds from the diet. Some reactions of bacterial metabolism really improve the bioavailability and activity of polyphenolic compounds. In many situations, only the product of bacterial metabolism of a polyphenol can be absorbed and exert a beneficial impact in humans. However, other bacterial metabolites may be harmful to human cells or other members of the microbiota. Hence, apart from interindividual variation in a daily intake of polyphenols, interindividual differences in the composition of the human microbiota may lead to differences in bioavailability and bioefficiency of polyphenols and their metabolites and cause a different impact on host health [230–232].

The identified pathways of bacterial metabolism of the most important groups of polyphenolic compounds are presented below.

#### *5.1. Isoflavonoids*

One of the best examples of how significant the role is of the intestinal microbiota in polyphenols impact on human health are the nonsteroidal estrogens. A lack of particular species within the microbiota may cause that isoflavonoid cannot exert its expected effect even though it has been consumed. Isoflavonoids, including daidzein, genistein, and glycitein, are present in soybeans, but they are rather inactive. Only their metabolites, e.g., S-equol or O-desmethylangolensin (O-DMA), are able to exert their pro-healthy effects. Equol, because of its high binding affinity to the estrogen receptor (S-equol preferentially activates estrogen receptor ERβ), can alleviate the symptoms of menopause. Moreover, the antiandrogenic activity and inhibition of osteoclast formation, anticancer activities and anti-inflammatory effects have been observed [6]. It was demonstrated that equol has about 100 times higher estrogenic activity than the daidzein itself [232]. Although O-DMA did not exhibit agonistic or antagonistic activities toward the glucocorticoid receptor (TRa1, or TRb1) and has very weak agonistic activities against ERα and ERβ, it can influence the growth of cancer cells, osteoclast formation, scavenging superoxide radical or exert leptin secretion inhibitory activity [233]. Bacteria strains producing small to moderate amounts of dihydrodaidzein and/or O-DMA from daidzein and dihydrogenistein from genistein are recognized more often than equol producers [234]. O-DMA is found in 80–90% of the human population, whereas equol is found in only 30–50% of the population [235].

The possible metabolic pathways of daidzin and genistin degradation by bacteria are presented in Figures 4 and 5, respectively.

**Figure 4.** Possible pathways of microbial metabolism of daidzin and daidzein. Based on [236–243]. The dashed arrows indicate hypothesized reactions of microbiological degradation that were observed in vitro but were not reported in vivo.

**Figure 5.** Possible pathways of microbial metabolism of genistin and genistein. Based on [236–243]. The dashed arrows indicate hypothesized pathways of microbiological degradation that were observed in vitro but were not reported in vivo.

The isoflavones biotransformation generally starts with glycoside hydrolysis to release the aglycon. For example, daidzin (daidzein 7-*O*-glycoside) can be hydrolyzed to daidzein by *Eubacterium ramulus* [244]. Then, the hydrogenation of the double bond between C2 and C3 in ring C of daidzein (DZN) and genistein (GN) generates dihydroisoflavones such as dihydrodaidzein (DHD) and dihydrogenistein (DHG), respectively [236]. Dihydroisoflavones are further subjected to bacterial metabolism and can undergo: (a) the reductive pathway leading to equol formation, (b) the cleavage of the C ring, followed by the fission of the molecules into two moieties. Equol is generally produced from daidzein through the reductive metabolism, through dihydrodaidzein (DHD), tetrahydrodaidzein (cis-THD and/or trans-THD) and dehydroequol (DE) as intermediates; however, some equol-producing bacteria have also been shown to convert the genistein into dihydrogenistein and finally to 5-hydroxy-equol [245,246]. *Clostridium* sp. strain HGH136 cleaved the C-ring of daidzein to produce O-desmethylangolensin, probably via 2-dehydro-*O*demethylangolensin [241]. O-DMA may be further partially metabolized to resorcinol and 2-(4-hydroxyphenyl) propionic acid [233].

Daidzein was in part degraded by *E. ramulus* to *O*-desmethylangolensin, while genistein was completely degraded via 6 -hydroxy-*O*-desmethylangolensin to 2-(4-hydroxyphenyl) propionic acid [243]. It means that the OH group in position 6 of O-DMA was crucial for its further degradation. It is interesting that dihydrogenistein was neither observed as an intermediate in this transformation nor converted itself by growing cells of *E. ramulus*. Genistein-7-*O*-glucoside was partially transformed by way of genistein to the product 2-(4-hydroxyphenyl)-propionic acid.

*E. ramulus*, strain CG19-1 is capable of cleaving both 6 -hydroxy-*O*-desmethylangolensin and *O*-desmethylangolensin to phloroglucinol and resorcinol, respectively; and 2-(4 hydroxyphenyl) propionic acid was additionally formed from both O-DMA and 6 -OH-*O*-DMA [239].

A different metabolic pathway was revealed by Murota et al. [235]. They reported that the metabolites of genistein and glycitein that are primarily found in human urine were dihydrogenistein, 6 -OH-*O*-DMA, 2-(4-hydroxyphenyl)-propionic acid and phloroglucinol for genistein, while dihydroglycitein, 5 -methoxy-*O*-DMA and 6-methoxy-equol for glycitein. Moreover, strain CG19-1 cleaved both O-desmethylangolensin and 6 -hydroxy-*O*-desmethylangolensin to yield 2-(4-dihydroxyphenyl) propionic acid. The corresponding cleavage product, resorcinol, was only observed for O-desmethylangolensin.

According to Rossi et al. [242], the metabolites arising from glycitein include dihydroglycitein, which can be further O-demethylated to 6,7,4 -trihydroxyisoflavone (proved in vitro for *Eubacterium limosum*) and reduced to dihydro-6,7,4 -trihydroxyisoflavone, and further reduced to 6-hydroxyequol or cleaved to 5 -hydroxy-*O*-desmethylangolensin. The other pathway of dihydroglycitein degradation was through the C-ring cleavage producing 5 -*O*-methoxy-*O*-desmethylangolensin or reduction to 6-methoxy-equol (Figure 6).

*Slackia isoflavoniconvertens* is capable of contributing to the bioactivation of daidzein and genistein by the formation of equol and 5-hydroxy-equol, respectively [246].

It should be underlined that some bacteria can produce equol from either daidzein or its glycoside daidzin, but some cannot produce equol unless several other species of bacteria metabolize daidzin to aglycone daidzein and daidzein to DHD or other derivatives that are also present [235]. For example, *Clostridium* sp. strain HGH6 and *Lactobacillus* sp. Niu-O16 can reduce daidzein to dihydrodaidzein but did not convert dihydrodaidzein to equol [247]. On the other hand, *Eggerthella* sp. Julong 732 is capable of converting dihydrodaidzein, but not daidzein, to equol [238,247]. *Eggerthella* sp. Strain YY7918 converted substrates daidzein and dihydrodaidzein into S-equol but did not convert daidzin, glycitein, genistein, or formononetin into it [248]. Strain TM-40 (93% of homology with *Coprobacillus catenaformis*) isolated by Tamura et al. [249] produced dihydrodaidzein from both daidzein and daidzin. Decroos et al. [240] isolated from human feces a stable mixed microbial culture (*Enterococcus faecium* strain EPI1, *Lactobacillus mucosae* strain EPI2, *Finegoldia magna* strain EPI3 and an as yet undescribed species related to *Veillonella* sp.) that was able to covert daidzein into equol.

Among intestinal bacteria that were proved to metabolize the soya isoflavone daidzein and genistein to equol, DHD and/or O-DMA are *Slackia equolifaciens* (DZN to equol) [250], *Slackia isoflavoniconvertens*, *Adlercreutzia equolifaciens*, *Asaccharobacter celatus*, *Enterorhabdus mucosicola* (DZN to equol), *Peptoniphilus gorbachii* (DZN and GN to equol and O-DMA), *Gordonibacter urolithinfaciens* (DZN to O-DMA), some strains of *Eggerthella lenta* (DZN and GN to O-DMA), *Enterococcus lactis* (to O-DMA), some strains of *Bifidobacterium adolescentis* (DZN and GN to O-DMA), *B. animalis* (DZN to O-DMA) and *B. longum* (DZN and GN to O-DMA), some members of Coriobacteriaceae, e.g., *Collinsella massiliensis* (DZN and GN to O-DMA) and *C. aerofaciens* (DZN to O-DMA) [234], *Eggertella* strain Julong 732 (DHD via THD to equol) [238], *Lactococcus garvieae* strain 20–92 [251], *Eubacterium ramulus* Julong 601 (DZN to O-DMA, GN to 2-(4-hydroxyphenyl) propionic acid) [252], *Clostridium* sp. HGH 136 (DZN to DHD) [241] and HGH6 (DZN do DHD, GN to DHG), and *E. coli* HGH21 (DZN to DHD and GN to DHG) [253].

**Figure 6.** The pathways of bacterial metabolism of glycitin. Based on [237,242].

Puerarin is a daidzein 8-C-glucoside and was reported to be metabolized to daidzein by human intestinal flora such as *E. ramulus* CG 19-1 [239] or intestinal strain PUE, converting puerarin to daidzein by cleaving a C-glucosyl bond [254]. Formononetin and biochanin A are the principal isoflavones of red clover (and as a consequence, equol is present in cow milk) and can be consumed in the form of dietary supplements. Hur et al. [255] demonstrated that *Eubacterium limosum* is able to produce daidzein and genistein from formononetin and biochanin A, respectively. It means that due to bacterial metabolism, more potent phytoestrogens have been formed in the colon, as the estrogenic potencies of the mentioned compound for both estrogen receptors ERα and ERβ showed the affinities in the order of genistein > daidzein > biochanin A > formononetin. In the urine samples of volunteers consuming formononetin and biochanin A, other metabolites were also identified, such as dihydroformononetin and angolensin for formononetin and dihydrobiochanin A and 6 -hydroxyangolensin for biochanin A [256] (Figure 7).

**Figure 7.** The human metabolism of formononetin and biochanin A. Based on [16,255,256].

### *5.2. Other Phytoestrogens*

In addition to soy isoflavonoids, there are other ligands for estrogen receptors that are produced by intestinal microbiota, such as enterolactone, enterodiol, urolithins and 8-prenylnaringenin. Enterolactone and enterodiol are derivatives of plant lignans from sesame seed or flaxseed. It was proved at concentrations that can be achieved with high consumption of products rich in lignans, both, but enterolactone to a lesser extent can potently activate human estrogen receptors ERα and ERβ [257]. The bacterial transformation of lignans into phytoestrogens (Figure 8) was carried out mainly by *Peptostreptococcus* and *Eubacterium* species and included their demethylation and dihydroxylation, leading to enterolactone production [258]. Enterolactone can further be converted into enterodiol, and various studies proved that both the mentioned mammalian lignans are produced by human colonic microbiota from dietary precursors. Production of enterodiol is about 2000 times more efficient, meaning that the enterodiol-producing bacteria are dominant in the human gut.

**Figure 8.** Lignans metabolism by gut microbiota. Based on [12,237,258–261]. The dashed arrows indicate hypothesized or multistep process.

> The main bacteria converting lignans to enterolactone and enterodiol are *Peptostreptococcus* sp. SDG-1 and *Eubacterium* sp. SDG-2 [262], *Bacteroides distasonis*, *B. fragilis*, *B. ovatus*, *Eubacterium callanderi*, *Eubacterium limosum*, *Clostridium cocleatum*, *Clostridium scindens*, *Eggerthella lenta*, *Butyribacterium methylotrophicum*, *Butyribacterium pseudocatenulatum*, *Bifi-*

*dobacterium longum*, *B* > *breve*, *B. catenulatum*, *B. pseudocateunaltum*, *Enterococcus faecalis*, *Ruminococcus* sp. END-1 [261], *Clostridium saccharogumia* and *Lactonifactor longoviformis* [263]. Two organisms able to demethylate and dehydroxylate secoisolariciresinol were isolated from human feces. Based on 16S rRNA gene sequence analyses, they were named *Peptostreptococcus productus* SECO-Mt75m3 and *Eggerthella lenta* SECO-Mt75m2 [264]. It was demonstrated both in vivo and in vitro that the major metabolite of sesamin in humans is enterolactone [265]. The intestinal pathways of enterolactone and enterodiol production from lignan are presented in Figure 8.

Ellagitannins (ELT) are one of the main groups of hydrolyzable tannins that are characterized by high antioxidant activity. They are common in some fruits, such as pomegranates, black raspberries, raspberries and strawberries, as well as in walnuts and almonds. Chemically they are different esters of hexahydroxydiphenic acid (HHDP) and a polyol, usually glucose or quinic acid [266]. According to the number of HHDP groups linked to the sugar moiety, ellagitannins can be classified into monomeric, oligomeric, and polymeric ELT. The main ellagitannins identified in foods are punicalagin (Figure 9), sanguiin H-6 (dimer of casuarictin) **(212)**, lambertianin C (trimer of casuarictin) **(214)**, pedunculagin, castalagin **(209)**, casuarictin **(210)** and potentillin. Because of their size (634 Da for sanguiin H4 to up to 3740 Da for lambertianin D), these molecules are characterized by very low bioavailability and are not absorbed in the gastrointestinal tract until they are metabolized by gut bacteria. Intact ellagitannins and a product of their acidic or basic hydrolysis—ellagic acid (Figure 9), reach the distal part of the gastrointestinal tract where they are transformed by intestinal microbiota into dibenzopyran-6-one derivatives, known as urolithins, that are much better absorbed [267].

**Figure 9.** Degradation of ellagitannins to ellagic acid. Based on [268].

The generic name of urolithins includes different hydroxylated 6H-dibenzo[b,d] pyran-6-one derivatives. The bacterial transformation includes reduction of one of the two lactone groups followed by decarboxylation and sequential dehydroxylation involving a step-bystep reduction to tetrahydroxy (urolithin D), trihydroxy (urolithin C), dihydroxy (urolithin

A and isourolithin A), and monohydroxy dibenzopyranones (urolithin B). The pathway of bacterial metabolism is presented in Figure 10. Bacteria able to catalyze the biotransformation of ellagitannins to urolithins are *Gordonibacter urolithinfaciens* and *G. pamelaeae* that belong to the family Coriobacteriaceae [269,270] and *Ellagibacter isourolithinifaciens* from Eggerthellaceae [271].

**Figure 10.** The bacterial metabolism of ellagic acid to urolithins and derivatives. The red arrows represent the pathways reported in *Gordonibacter urolithinfaciens* and *G. pamelaeae.* Based on [272–274].

Although urolithins are characterized by lower antioxidant activity than ellagitannins, they circulate in the plasma as glucuronide and sulfate conjugate and display benefit influence on human health due to their estrogenic and/or anti-estrogenic activity, as well as anticancer activities. It means that bacterial metabolism is crucial for the pro-healthy properties of various berries [266,272].

It has been shown that the production of the potent hop phytoestrogen 8 prenylnaringenin (8-PN) depends on the activity of human intestinal microbiota [275]. This compound is generated from xanthohumol and isoxanthohumol that unaltered reach the small intestine (Figure 11). Among bacteria that catalyze the demethylation of isoxanthohumol into 8-PN are *Eubacterium limosum* and *E. ramulus* [275–277]. In addition to a strong impact on the ERα receptor, 8-prenylnaringenin inhibits angiogenesis and metastasis, prevents bone loss in rats and exhibits antiandrogenic activity [275,278].

**Figure 11.** Bacterial biotransformation of prenylflavonoids xanthohumol and isoxanthohumol. Based on [275,276,279].

### *5.3. Bacterial Transformation of Anthocyanidins*

Anthocyanidins (ACD) are plant pigments responsible for flower, fruit and vegetable color. Their structure and color depend on pH value and the presence of copigments. In plant tissues, ACD are generally present in the form of glycosides, called anthocyanins (ACN), that are susceptible to hydrolytic conversion into their corresponding anthocyanidins. Glucose, galactose, rhamnose and arabinose are the sugars most commonly encountered, usually as 3-*O*-glycosides or 3,5-*O*-diglycosides; however, rutinosides (6-*O*-Lrhamnosyl-D-glucosides), sophorosides (2-*O*-D-glucosyl-D-glucosides) and sambubiosides (2-*O*-D-xylosyl-D-glucosides) also occur, as do some 3,7-diglycosides and 3-triosides [280]. Moreover, some of the hydroxyl groups can be methylated, giving the big diversity of plant anthocyanidins. The anthocyanidins occur in the vacuole as an equilibrium of four molecular species that affects their color (Figure 12). However, after fruit and vegetable consumption, the form of the flavylium cation exists only in the stomach, while other forms are present in the lower parts of the gastrointestinal tract and in the tissues (if absorbed).

Bacterial metabolism of ACN involves the cleavage of glycosidic linkage and breakdown of the anthocyanidin heterocycle. Aura et al. [281] demonstrated that cyanidin-3 rutinoside was degraded through cyanidin-3-glucoside and cyanidin aglycone as intermediary metabolites. After hydrolysis of the protective 3-glycosidic linkage, the released aglycons are stable under acidic pH but unstable under neutral or slightly basic pH. It

means that under physiological conditions in the small intestine, the cleavage of the heterocyclic flavylium ring occurs [274]. An attack of the flavylium carbon at position 2 produces an unstable hemiketal that rapidly forms a ketone (Figure 13). Through keto-enol tautomerism of the neighboring enol functionality, the resulting α-diketone is very reactive and is easily decomposed by gut microbiota to phenolic acids (mainly protocatechuic acid, syringic acid, vanillic acid) and aldehydes (mainly phloroglucinol aldehyde) [274].

**Figure 12.** Anthocyanins equilibria [39].

**Figure 13.** Main steps of anthocyanin degradation. Based on [274,282].

It was demonstrated for raspberry anthocyanins; when incubated with fecal suspensions under anaerobic conditions, that they underwent a transformation by the colonic microflora. After C-ring fission in cyanidin, aglycone phenolic acids were released, originating from both the A and B rings. It was proved that some of the colonic catabolites

entered the circulation and were further metabolized before being excreted in urine (e.g., as hippuric acid) [283].

Phenolic acids may be utilized as a source of energy by the intestinal microflora. Keppler and Humpf [282] demonstrated that bacterial metabolism of the methoxyl derivatives as syringic acid and vanillic acid was accompanied by O-demethylation and resulted in the formation of gallic acid and protocatechuic acid (PCA), respectively. As phloroglucinol aldehyde (PHA) was degraded by the intestinal microflora very similar in comparison to the sterilized control samples, it was not possible to distinguish between the chemical or microbial transformation of the aldehyde. However, the phloroglucinol acid was detected as the oxidation product of PHA in very low amounts only in the non-sterilized inoculum filtrate [282], indicating the role of gut microbiota in the transformation.

Some in vitro studies revealed that the numbers of potentially beneficial bacteria (bifidobacteria and lactobacilli) increased after the consumption of purple sweet potato anthocyanins and grape seed extract [284]. As anthocyanins are hardly absorbed in the small intestine, they may be transformed into small molecular phenolic acids by colonic microbiota through ring cleavage, dihydroxylation and methylation reactions. Such metabolites generated from polyphenols may selectively stimulate the growth of beneficial bacterial, whereas the proliferation of harmful bacteria would be inhibited. Ávila et al. [285] analyzed various strains of *Lactobacillus plantarum* and *L. casei*, as well as probiotic strains *Lactobacillus acidophilus* LA-5 and *Bifidobacterium lactis* BB-12. They proved the enzymatic potential of selected strains for bioconversion of delphinidin and malvidin glycosides to their metabolites. Incubation of malvidin-3-glucoside with *B. lactis* BB-12, *L. plantarum* IFPL722, and *L. casei* LC-01 cell-free extracts led to different patterns of gallic, homogentisic and syringic acid formation.

It was also reported that gallic acid and free anthocyanins activated cell growth and the rate of malic acid degradation; vanillic acid showed a slight inhibiting effect, while protocatechuic acid had no effect. Finally, gallic acid and ACN were metabolized, especially by growing cells [88]. Incubation of malvidin-3-glucoside with fecal bacteria mainly resulted in the formation of syringic acid, while the mixture of anthocyanins resulted in the formation of gallic, syringic and *p*-coumaric acids [286].

The most abundant anthocyanins in fruit and vegetables are cyanidin, pelargonidin, petunidin, peonidin and delphinidin. The hypothesized pathways of their bacterial degradation are presented in Figures 14 and 15. Mayor ACN metabolites generated in the human colon by bacteria are protocatechuic acid, syringic acid, vanillic acid, gallic acid, phenylacetic acid, 3,4-dihydroxyphenylpropionic acid, 3,4-dihydroxyphenylacetic acid, 4-hydroxybenzoic acid, but also 4-hydroxyphenylethanol (tyrosol), catechol, benzoic acid [282,284,287–289].

Zhu et al. [284] reported that 2,4,6-trihydroxybenzoic acid, 4-hydroxybenzaldehyde, benzoic acid, phenylacetic acid, and phenylpropionic acid were found in the medium after bacterial metabolism od cyanidin-3-*O*-glucoside. The metabolism of cyanidin-3- *O*-glucoside and cyanidin-3-*O*-rutinoside mainly resulted in the formation of protocatechuic, vanillic, and *p*-coumaric acids, as well as 2,4,6-trihydroxybenzaldehyde, while the main metabolites of delphinidin-3-*O*-rutinoside were gallic acid, syringic acid and 2,4,6 trihydroxybenzaldehyde. Among minor metabolites identified after microbial metabolism of mentioned glycosides were: protocatechuic acid-glucoside, caffeic acid, tartaric acid, catechol, as well as pyrogallol, ferulic acid, 4-hydroxybenzoic acid [290]. This research indicated that the intake of ACNs might result in the appearance of specific metabolites that exert a protective effect in host physiology.

The main phenolic acid detected in fecal suspensions incubated with raspberry anthocyanins were: 3-phenylacetic acid, 3-(4 -hydroxyphenyl) lactic acid, tyrosol, 3-(4 hydroxyphenyl) propionic acid, 3-(3 -hydroxyphenyl) propionic acid, 4-hydroxybenzoic acid and 3,4-dihydroxybenzoic acid, but lower amounts of catechol, resorcinol, pyrogallol and 3-(3 ,4 -dihydroxyphenyl) propionic acid were also found [283]. Seven metabolites formed by human fecal bacteria were observed by LC/MS after incubation with cyanidin3-*O*-glucoside and cyanidin-3-*O*-galactoside, and protocatechuic acid (major metabolite), 2,4,5-trihydroxy-benzaldehyde, 5-hydroxy-2-(4 -hydroxyphenyl)-2H-chromen-3,7-dione, 3,5,7-trihydroxy-2-(3 ,4 -dihydroxyphenyl)-2H-chromene, and 5-hydroxy-2-phenyl-2Hchromen-3,7-dione were identified [288]. The deglycosylation, decomposition, hydrogenation, and dihydroxylation reactions were involved in their generation.

**Figure 14.** Biodegradation of anthocyanins and their main metabolites: cyanidin → 3,4-dihydroxybenzoic acid (protocatechuic acid), peonidin → 3-methoxy-4-hydroxybenzoic acid (vanillic acid) and pelargonidin → 4-hydroxybenzoic acid. Based on [282,284,287,288].

Malvidin-3-glucoside was completely degraded into syringic acid after incubation with a human fecal slurry for 24 h, whereas gallic acid, *p*-coumaric, and syringic acid were

formed after a mixture of various anthocyanins were incubated with healthy human fecal bacteria [291].

### *5.4. Metabolism of Procyanidins and Catechins by Intestinal Bacteria*

Condensed tannins (also called catechol-type tannins or non-hydrolyzable tannins) are an important class of polyphenols that do not contain sugar residues. They are also called proanthocyanidins as, under oxidative conditions, they depolymerize, yielding anthocyanidins. Therefore, different types of condensed tannins exist, such as the procyanidins, propelargonidins, prodelphinidins, profisetinidins, proteracacinidins, proguibourtinidins or prorobinetidins.

Condensed tannins are formed from flavan-3-ols or flavan-4-ols. These particular types of condensed tannins are procyanidins, which are not susceptible to cleavage by hydrolysis. Procyanidins are polymers of 2 to 50 (or more) catechin units (usually catechin and epicatechin molecules) joined by carbon–carbon bonds. The most ubiquitous are B-type procyanidins, abundant in apple, cocoa, pear, blueberries; these subunits are linked by single bond C4–C8 or C4–C6. In A-type procyanidins, present in cranberries, a double linkage exists; the C4–C8 or C4-C6 bond is accompanied by an additional C2–O–C7 or C2–O–C5 ether bond.

Some studies demonstrated that highly polymeric procyanidins (PPs) administration markedly decreased the Firmicutes/Bacteroidetes ratio and increased by eight times the proportion of *Akkermansia*, suggesting that PPs influence the gut microbiota and the intestinal metabolome to produce beneficial effects on metabolic homeostasis [76]. On the other hand, some species of intestinal bacteria are able to degrade oligomeric procyanidins. Spencer et al. [221] proved that procyanidin oligomers (trimer to hexamer) are hydrolyzed in simulated gastric juice to mixtures of epicatechin monomer and dimer, thus enhancing the potential for their absorption in the small intestine. Proanthocyanidins that undergo partial acid-catalyzed cleavage then decompose to monomeric flavan-3-ols, which are also metabolized by colonic bacteria. The bacterial degradation of flavan-3-ols and proanthocyanidins follow a similar pathway, and both lead to a generation of a unique compound 5-(3 ,4 -dihydroxyphenyl)-γ-valerolactone, which further undergoes dihydroxylation and oxidation to produce phenolic acids [291,293] (Figure 16).

**Figure 16.** Microbial metabolism of procyanidin B2, (−)-epicatechin and (+)-catechin. Based on [274,291,294–297].

The incubation of purified (+)-catechin, (−)-epicatechin, procyanidins A2 and B2, as well as partially purified apple and cranberry procyanidins with human gut microbiota, resulted in their degradation. The common metabolites were benzoic acid, 2 phenylacetic acid, 3-phenylpropionic acid, 2-(3 -hydroxyphenyl) acetic acid (OPAC), 2-(4 - hydroxyphenyl) acetic acid, 3-(3 -hydroxyphenyl) propionic acid, and hydroxyphenylvaleric acid. Interesting, that 5-(3 ,4 -dihydroxyphenyl)-γ-valerolactone and 5-(3 hydroxyphenyl)-γ-valerolactone were identified as the bacterial metabolites of epicatechin, catechin, procyanidin B2, and purified apple procyanidins, but not from the procyanidin A2 or cranberry procyanidin ferments, while 2-(3 ,4 -dihydroxyphenyl) acetic acid was only found in the fermented broth of procyanidin B2, A2, apple, and cranberry procyanidins [298].

The monomeric flavan-3-ols, which are usually shared by condensed tannins, can be degraded by fecal microbiota into low molecular weight aromatic compounds, including phenylpropionic acid, 2-(3 -hydroxyphenyl) acetic acid, 3-(3 -hydroxyphenyl) propionic acid, 5-(3 -hydroxyphenyl) valeric acid, phenylacetic acid, and 2-(4 -hydroxyphenyl) acetic acid [294]. Similar results have been obtained Appeldoorn et al. [296]. Purified procyanidin dimers, when incubated with human microbiota, have been transformed and among major identified metabolites were 2-(3 ,4 -dihydroxyphenyl) acetic acid (DOPAC) and 5-(3 ,4 -dihydroxyphenyl)-γ-valerolactone. Other metabolites detected were: OPAC, 2-(4 -hydroxyphenyl) acetic acid, 3-(3 -hydroxyphenyl) propionic acid, phenylvaleric acids, monohydroxylated phenylvalerolactone, and 1-(3 ,4 -dihydroxyphenyl)-3-(2,4,6 trihydroxyphenyl) propan-2-ol. In studies of Deprez et al. [297], polymeric procyanidins were metabolized by colonic bacteria into low-molecular-weight phenolic acid, and the main metabolites were 3-phenylpropionic acid, 3-(4 -hydroxyphenyl) propionic acid, 3-(3 hydroxyphenyl) propionic acid, 5-(3 -hydroxyphenyl) valeric acid, 2-(3 -hydroxyphenyl) acetic acid, and 2-(4 -hydroxyphenyl) acetic acid.

Among intestinal bacteria that are able to convert catechins are *Eggerthella lenta* and *Flavonifractor plautii* (formerly *Clostridium orbiscindens*) [295]. *Eggerthella lenta* rK3 reductively cleaved the heterocyclic C-ring of both (−)-epicatechin and (+)-catechin giving rise to 1-(3 ,4 -dihydroxyphenyl)-3-(2,4,6-trihydroxyphenyl) propan-2-ol (Figure 16). The conversion of catechin proceeded five times faster than that of epicatechin. *Flavonifractor plautii* aK2 and *Flavonifractor plautii* DSM 6740 further converted 1-(3 ,4 -dihydroxyphenyl)- 3-(2,4,6-trihydroxyphenyl) propan-2-ol to 5-(3 ,4 -dihydroxyphenyl)-γ-valerolactone and 4-hydroxy-5-(3 ,4 -dihydroxyphenyl) valeric acid.

According to Tzounis et al. [59], the initial conversion of (+)-catechin to (+)-epicatechin is required to the generation of 5-(3 ,4 -dihydroxyphenyl)-γ-valerolactone, 5-phenyl-γvalerolactone and phenylpropionic acid as metabolites. The prebiotic effects of both (+) catechin and (−)-epicatechin was observed, suggesting that the consumption of flavanolrich foods may support gut health through their ability to exert prebiotic actions.

Monagas et al. [299] demonstrated that some phenolic acids, including 3-*O*-methyl gallic, gallic, caffeic, 3-(4 -hydroxyphenyl) propionic, phenylpropionic, and 2-(4 -hydroxyphenyl) acetic acids derived from the microbial degradation of tea catechins, were able to inhibit the growth of several pathogenic and non-beneficial intestinal bacteria without significantly affecting the growth of beneficial bacteria (*Lactobacillus* spp. and *Bifidobacterium* spp.). It is possible that *Bifidobacterium* sp. are resistant to flavan-3-ols, being the are important ironchelating compounds, because these bacteria do not use heme-containing enzymes [237]. Growth of certain pathogenic bacteria such as *Clostridium perfringens*, *Clostridium difficile*, *Streptococcus pyogenes*, and *Str. pneumoniae* was significantly repressed by tea phenolics (catechin, epicatechin, gallic acid, caffeic acid), while commensal anaerobes like *Clostridium* spp., *Bifidobacterium* spp. and probiotics such as *Lactobacillus* sp. were less severely affected [70]. Similarly, the bacterial metabolites, such as 3-(4 -hydroxyphenyl) propionic acid, 3-phenylpropionic acid and 2-(4 -hydroxyphenyl) acetic acid, strongly inhibited the growth of *E. coli, S. aureus* and *Salmonella* sp. without influencing beneficial *L. casei* strain Shirota and *Bifidobacterium breve*.

Alakomi et al. [300] have shown that DOPAC, OPAC, 3-(3 ,4 -dihydroxyphenyl) propionic acid, 3-(4 -hydroxyphenyl) propionic acid, 3-phenylpropionic acid, and 3-(3 hydroxyphenyl) propionic acid efficiently destabilized the outer membrane of *Salmonella enterica* subsp. *enterica* serovar Typhimurium and *S. enterica* subsp. *enterica* serovar Infantis. Moreover, DOPAC, OPAC and 3-(3 ,4 -dihydroxyphenyl) propionic acid increased the susceptibility of *Salmonella* Typhimurium strains for novobiocin. It means that beneficial bacteria residing in the human gut can inhibit *Salmonella* growth by the transformation of food flavonoids to active antimicrobial metabolites.

### *5.5. The Bacterial Metabolism of Flavones and Flavonols*

It is interesting that some compounds are common and can be generated by colonic microbiota during the metabolism of various polyphenols, although to a different extent. For example, flavan-3-ols, as well as flavonols and hydroxycinnamic acids, lead to the generation of 3-(3 ,4 -dihydroxyphenyl)-propionic acid, 3-(3 -hydroxyphenyl) propionic acid, and 3-(4 -hydroxyphenyl) propionic acid. It means that some enzymes and metabolic pathways are quite common among bacteria. When quercetin is metabolized by bacteria, ring fission is done, leading to the generation of DOPAC, OPAC and protocatechuic acid (PCA) (Figure 17). Braune et al. [301] examined the degradation mechanism of the flavonol quercetin and the flavone luteolin and had demonstrated that *Eubacterium ramulus* converted quercetin through taxifolin and alphitonin, resulting in the formation of DOPAC and phloroglucinol. Flavonol luteolin was transformed by *E. ramulus* to 3-(3 ,4 dihydroxyphenyl) propionic acid via eriodictyol and derivatives. In both pathways, ring fission had taken place. Glycosides of quercetin, such as common in onions quercetin 4 -*O*-glucoside and quercetin 3-*O*-glucoside, are first hydrolyzed to aglycone and then are also catabolized, giving ring-fission products [235]. *Bifidobacterium animalis* subsp. *lactis* AD011, isolated from infant feces, has been shown to catalyze quercetin 3-*O*-glucoside and isorhamnetin 3-*O*-glucoside into quercetin and isorhamnetin, respectively [302].

The degradation of flavones and flavonols by *Clostridium orbiscindenss* was studied by Schoefer et al. [303]. They confirmed the quercetin degradation via taxifolin and alphitonin. Flavone apigenin and luteolin were converted to 3-(4 -hydroxyphenyl) propionic acid and 3-(3 ,4 -dihydroxyphenyl) propionic acid, respectively, and phloroglucinol was released in both cases (Figure 17). The intermediate metabolites were naringenin and phloretin for apigenin and eriodictyol and dihydrochalcone for luteolin [303]. However, the isolated *C. orbiscindens* strain was unable to hydrolyze the glycosidic bonds of luteolin 3- *O*-glucoside, luteolin 5-*O*-glucoside, naringenin 7-*O*-neohesperidoside (naringin), quercetin 3-*O*-glucoside, quercetin 3-*O*-rutinoside (rutin), and phloretin 2 -*O*-glucoside, suggesting that other bacteria are required for the initial steps in the metabolism of flavonoid glycosides in the human intestine. Similar pathways were reported for degradation of myricetin, kaempferol as well as quercetin, apigenin and luteolin glycosides, and among bacteria involved in their metabolism were *Enterococcus casseliflavus*, *Eubacterium ramulus*, *Eubacterium oxidoreducens*, *Butyrivibrio* spp., *Clostridium orbiscidens*, *Eggerthella* sp., *Flavonifractor plautii*, *Bacteroides uniformis*, *Bacteroides ovatus*, *Bifidobacterium* spp., *Bacteroides distasonis*, and *Blautia* sp. [235,237,304–307].

Experiments using radiolabeled quercetin 4 -*O*-glucoside (Q4 G) revealed that Q4 G passes through the gastrointestinal tract of rats and that almost all of Q4 G is converted into phenolic acids, with DOPAC and OPAC being the most abundant, and a small amount of PCA was also generated [235]. Moreover, 69% of Q4 G radioactivity was recovered in the form of phenolic acid derivatives, such as OPAC and hippuric acid, in the urine. It means that the first ring-fission product is DOPAC, which is subsequently subjected to dehydroxylation to form OPAC, followed by further catabolism into hippuric or benzoic acids (Figure 17). DOPAC also has been identified as a major catabolite of quercetin glycosides, such as rutin, as well as procyanidins (Figure 16). It is important because DOPAC is known to be a metabolite of the neurotransmitter dopamine, suggesting the existence of a metabolic pathway for DOPAC in humans. It has been demonstrated that DOPAC exerts anticancer, anti-inflammatory, cardioprotective and neuroprotective impact. However, DOPAC may inhibit mitochondrial respiration in brain mitochondria (when NO radical is present) and thus lead to mitochondrial dysfunction, which is assumed to be an important mechanism involved in Parkinson's disease [308].

**Figure 17.** Possible pathways of the transformation of flavones and flavonols due to metabolism by intestinal bacteria. Based on [239,301,304,309,310].

### *5.6. Microbial Catabolism of Phenolic Acids*

Phenolic acids can be delivered to the intestine with food, but they are also generated as the final metabolites during the degradation of various polyphenols. Phenolic acids play an important protective role in degenerative diseases as they exert antioxidant, antitumor, apoptotic, neuroprotective, hepatoprotective, anti-inflammatory and antimicrobial properties [35]. However, there has been some controversy about the bioactivity of polyphenols after metabolism. Once ingested, these molecules are metabolized and transformed into methylated, glucuronated and sulfated metabolites, and there is much evidence proving

both the enhanced and decreased biological activity of phenolic acid metabolites [35]. Not all phenolic acids are absorbed, and some of them reach the colon and can be metabolized by bacteria. It is supposed that the presence of an ester moiety lowers hydroxycinnamic acids (HCAs) absorption. Actually, HCAs in a free form are rapidly absorbed throughout the gastrointestinal tract, while HCAs esters or HCAs attached to cell walls require to be hydrolyzed by bacterial esterases before absorption [311]. As a large interindividual variation of phenolic acid metabolites (Figure 18) was observed, it may suggest that the catabolic pathways of both chlorogenic acid and other phenolic acids depend mainly on the colon microbiota composition.

**Figure 18.** Pathways of some phenolic acid metabolism conducted by various bacteria. Based on [37,304,312–315].

Incubation of coffee samples (a rich source of phenolic acids) with the human fecal microbiota led to the rapid metabolism of chlorogenic acid and the production of dihydrocaffeic acid and dihydroferulic acid, while caffeine remained unmetabolized [316]. Caffeic acid esters can be rapidly transformed to 3-(3 -hydroxyphenyl) propionic acid by human fecal microbiota [304] by de-esterification followed by a reduction of a double bond and dehydroxylation at the C4 position (Figure 18). Monteiro et al. [317] revealed that the main chlorogenic acid metabolites identified in urine after coffee consumption were: dihydrocaffeic, gallic, isoferulic, ferulic, vanillic, caffeic, 5-*O*-caffeoylquinic, sinapic, 4-hydroxybenzoic, and *p*-coumaric acids, with gallic and dihydrocaffeic acids being the major ones. Similar results were reported by Clifford et al. [37]. One of the most abundant sources of caffeic acid in nature is 5-*O*-caffeoylquinic acid (neochlorogenic acid), which was also proved to be hydrolyzed to caffeic and quinic acids by esterases from colonic microflora and is not degraded and absorbed in the upper gastrointestinal tract [311]. The lack of colonic microbiota (e.g., in germfree rats) resulted in the inhibition of hippuric acid formation, indicating that esterase enzymes of the colonic microbiota are involved in this pathway [318].

### *5.7. Bacterial Metabolism of Resveratrol and Curcumin*

Resveratrol is a natural polyphenol widely found in its *trans* isomer form in various fruits, especially grapes and berries, peanuts, and red wine. It was reported that purified resveratrol inhibited the growth of some pathogens, among other intestinal bacteria such as *Helicobacter pylori*, *Enterococcus faecalis*, *Pseudomonas aeruginosa*, *Vibrio* spp. [319]. However, some bacteria are able to metabolize *trans*-resveratrol. *Slackia equolifaciens* and *Adlercreutzia equolifaciens* [320] and *Eggerthella lenta* ATCC 4305 [319] converted resveratrol to dihydroresveratrol (Figure 19); while *Bacillus cereus* NCTR-466, *Achromobacter denitrificans* NCTR-774, and *E. coli* ATCC 47,004 metabolizes trans-resveratrol into resveratrol 3-*O*-glucoside (piceid) and resveratrol 4-*O*-glucoside (resveratroloside) [319]. Among other colonic metabolites of resveratrol, lunularin and 3,4 -dihydroxy-*trans*-stilbene [320] were identified. However, their bacterial producers are unknown (Figure 19). The 16S rRNA sequencing of fecal samples demonstrated the association of lunularin producers with a higher abundance of Bacteroidetes, actinobacteria, Verrucomicrobia, and Cyanobacteria and with a lower abundance of Firmicutes than either the dihydroresveratrol or mixed producers [321]. The bacterial metabolites of resveratrol can exert a beneficial impact on human health. Dihydroresveratrol reduced fatty acid-binding protein-4 expression, involved in fatty acid uptake in human macrophages treated with oxidized LDL and stimulates fatty acid oxidation in human fibroblasts, lunularin reduced the expression of proinflammatory mediators in endothelial cells [322], while 3,4 -dihydroxy-*trans*-stilbene increased glucose uptake and induced adenosine monophosphate kinase phosphorylation in C2C12 myotubes independently of insulin [323].

Jarosova et al. [324] examined the metabolism of six stilbenoids resveratrol, oxyresveratrol, piceatannol, thunalbene, batatasin III, and pinostilbene by colon microbiota from various donors. It was demonstrated that resveratrol, oxyresveratrol, piceatannol and thunalbene were subjected to metabolic transformation via double bond reduction, dihydroxylation, and demethylation (Figure 19), while batatasin III and pinostilbene were stable at simulated colon conditions. Authors reported strong interindividual differences in speed, intensity, and pathways of metabolism among the fecal samples obtained from the donors, suggesting that microbiota composition plays a crucial role in the influence of resveratrol on human health.

Curcumin is a lipophilic polyphenol characterized by quite poor bioavailability. It is supposed that curcumin passes through the stomach without any chemical modifications and reaches the large intestine, where it undergoes extensive phase I and II metabolism. The reductive pathways of metabolism by phase I enzymes lead to the formation of dihydrocurcumin, tetrahydrocurcumin, and hexahydrocurcumin (Figure 20) [325].

**Figure 19.** The effect of bacterial metabolism on resveratrol, oxyresveratrol, thunalbene and piceatannol. Based on [320,324].

However, consecutive reduction of the double bonds in the curcumin chain resulting in the formation of dihydrocurcumin, tetrahydrocurcumin, and hexahydrocurcumin can occur in the gut by a CurA reductase (NADPH-dependent curcumin/dihydrocurcumin reductase) that has been isolated from intestinal *E. coli* [326]. The 24-h fermentation of curcumin, demethoxycurcumin and bis-demethoxycurcumin by human fecal microbiota resulted in 24%, 61% and 87% degradation, respectively. Three main metabolites were identified: tetrahydrocurcumin, dihydroferulic acid and 1-(4-hydroxy-3-methoxyphenyl)-2 propanol [327]. A similar experiment was performed by Burapan et al. [328], but a mixture composed of curcumin, demethoxycurcumin, and bis-demethoxycurcumin was metabolized by the human intestinal bacterium *Blautia* sp. MRG-PMF1. New metabolites generated from curcumin and demethoxycurcumin by the methyl aryl ether cleavage reaction were identified. Demethylcurcumin and bisdemethylcurcumin were sequentially produced from curcumin, while demethyldemethoxycurcumin was produced from demethoxycurcumin [328]. Bis(demethyl)tetrahydrocurcumin and bis(demethyl)-hexahydrocurcumin were identified among colonic metabolites of curcumin, demethoxycurcumin and bisdemethoxycurcumin [329].

All these metabolites can undergo phase II metabolism by glucuronidases and sulfotransferases that are capable of conjugating glucuronic acid or sulfate molecule, respectively, to produce the corresponding glucuronide and sulfate O-conjugated metabolites. Furthermore, gut microbiota may deconjugate the phase II metabolites and convert them back to the corresponding phase I metabolites or to fission products such as ferulic acid and dihydroferulic acid in the colon [325].

It is interesting that CurA reductase, besides the ability to conversion of curcumin to tetrahydrocurcumin, is also able to metabolize resveratrol [326].

**Figure 20.** Metabolic pathway of curcumin. Reactions conducted by *E. coli* CurA and *Blautia* sp. are indicated. Based on [274,325,326,328–330].

### **6. Conclusions**

This review describes a bidirectional relationship between polyphenols delivered with food and the human gut microbiota. The manuscript presents a compilation of the knowledge from two perspectives. The first part describes the impact of various polyphenols classes on bacteria, with particular emphasis on human intestinal microbiota representatives. The mechanism of inhibitory impact of polyphenols, including protein binding, inhibition of nucleic acid synthesis, interaction with the cell wall and bacterial membranes, substrate deprivation, inhibition of energy metabolism and changes in cell attachment and biofilm formation, are discussed in details. In the second part, the role and pathways of

bacterial biotransformation of polyphenols are described, especially those reactions where bioactive metabolites with a significant impact on the human organism (both positive and negative) are produced. The role of interindividual variation in microbiota composition in the impact of food polyphenols on human health is explained. For example, the biotransformation of isoflavonoids and other phytoestrogens to bioactive O-DMA and S-equol, the generation of urolithins, the bacterial metabolites that can cross the blood–brain barrier, the degradation of complex condensed tannins and lignans as well as catabolic pathways of low-molecular-weight phenolic acids are elucidated.

The exact structures of all discussed phenolic compounds can be found in tables and figures, which will facilitate their comparison with the structures of other food ingredients and drawing one's own conclusions about their potential activity.

**Author Contributions:** All authors have read and agreed to the published version of the manuscript. Conceptualization, M.M. and A.D.-C.; Writing—Original Draft Preparation, M.M., I.D., T.T. and A.D.-C.; Writing—Review & Editing, M.M., I.D., T.T. and A.D.-C.; Visualization, M.M., I.D., T.T. and A.D.-C.; Supervision, A.D.-C.; Funding Acquisition, A.D.-C.

**Funding:** The APC was supported by the author's activation funds and through a research subsidy of the Department of Fermentation Technology and Microbiology, the University of Agriculture in Krakow.

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

### **References**


## *Review* **Translational Approaches with Antioxidant Phytochemicals against Alcohol-Mediated Oxidative Stress, Gut Dysbiosis, Intestinal Barrier Dysfunction, and Fatty Liver Disease**

**Jacob W. Ballway \* and Byoung-Joon Song \***

Section of Molecular Pharmacology and Toxicology, National Institute on Alcohol Abuse and Alcoholism, 9000 Rockville Pike, Bethesda, MD 20892, USA

**\*** Correspondence: jake.ballway@nih.gov (J.W.B.); bj.song@nih.gov (B.-J.S.)

**Abstract:** Emerging data demonstrate the important roles of altered gut microbiomes (dysbiosis) in many disease states in the peripheral tissues and the central nervous system. Gut dysbiosis with decreased ratios of Bacteroidetes/Firmicutes and other changes are reported to be caused by many disease states and various environmental factors, such as ethanol (e.g., alcohol drinking), Western-style high-fat diets, high fructose, etc. It is also caused by genetic factors, including genetic polymorphisms and epigenetic changes in different individuals. Gut dysbiosis, impaired intestinal barrier function, and elevated serum endotoxin levels can be observed in human patients and/or experimental rodent models exposed to these factors or with certain disease states. However, gut dysbiosis and leaky gut can be normalized through lifestyle alterations such as increased consumption of healthy diets with various fruits and vegetables containing many different kinds of antioxidant phytochemicals. In this review, we describe the mechanisms of gut dysbiosis, leaky gut, endotoxemia, and fatty liver disease with a specific focus on the alcohol-associated pathways. We also mention translational approaches by discussing the benefits of many antioxidant phytochemicals and/or their metabolites against alcohol-mediated oxidative stress, gut dysbiosis, intestinal barrier dysfunction, and fatty liver disease.

**Keywords:** gut microbiome; dysbiosis; leaky gut; endotoxemia; fatty liver disease; ethanol; oxidative stress; inflammation; phytochemicals; antioxidant

### **1. Introduction**

Ample research conducted over the past decade has revealed the expansive and critical role of the human microbiota in numerous physiological processes and pathological consequences. Before elaborating further on these microbial communities and their abundance, the terms microbiota and microbiome should first be distinguished, considering they are sometimes used interchangeably. A microbiota specifically refers to the "assemblage of microorganisms present in a defined environment" [1], while the microbiome includes "the microorganisms (i.e., bacteria, archaea, lower and higher eukaryotes, and viruses), their genomes (i.e., genes), and the surrounding environmental conditions" [1].

Various microbiomes within the human body have been well defined, including the gastrointestinal (GI) tract, lung, skin, urinary, and oral microbiomes, and collectively amount to trillions of bacterial cells that work in concert with (or sometimes in opposition to) human cells and experimental rodent models [2,3]. Current estimates suggest that 1.3 bacterial cells are present for every 1 human cell, contrary to previous suggestions of a 10:1 bacterial to human cell ratio [4]. However, these reduced estimates should not understate the breadth of microbial influence on bodily habitats. Despite differences in microbial composition among bodily habitats, the various microbiomes often perform similar functions, such as immune regulation in the GI tract, respiratory, oral, skin, and urinary microbiomes and promoting nutrient availability in the skin and oral microbiomes [2].

**Citation:** Ballway, J.W.; Song, B.-J. Translational Approaches with Antioxidant Phytochemicals against Alcohol-Mediated Oxidative Stress, Gut Dysbiosis, Intestinal Barrier Dysfunction, and Fatty Liver Disease. *Antioxidants* **2021**, *10*, 384. https:// doi.org/10.3390/antiox10030384

Academic Editor: Baojun Xu

Received: 15 January 2021 Accepted: 25 February 2021 Published: 4 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/).

Yet, while bacterial–human symbiosis can occur, so too can antagonistic interactions arise, resulting in diseases, such as periodontal disease (oral microbiome), urinary tract infection (UTI) (urinary tract microbiome), and pharyngitis and pneumonia (respiratory tract microbiome), to name several [2,3]. This review briefly addresses the nature of one of the most significant microbiomes: the gut microbiome. We also describe the mechanisms of the alcohol-mediated changes in gut microbiota, leaky gut, endotoxemia, and fatty liver disease. Based on the mechanistic insights, we finally propose the translational applications by describing the benefits of many antioxidant phytochemicals and/or their metabolites in preventing alcohol-mediated oxidative stress, gut leakiness, and fatty liver disease via changing gut microbiota.

### **2. The Gut Microbiome**

### *2.1. Microbiome Present in Many Tissues*

It is known that different parts of the GI tract may vary in bacterial and fungal composition and abundance, with the vast majority of gut microbes colonizing the colon [5–8]. The gut microbiota encompasses all the microbes found within the human GI tract, which stretches from the mouth, stomach, duodenum, jejunum, and ileum to the end of the large intestine (colon) at the rectum and anal canal. Among all the bacterial phylum present in the gut microbiota, the majority of bacteria fall into three phyla, Bacteroidetes, Firmicutes, or Actinobacteria, with a heavy slant toward Bacteroidetes and Firmicutes [9,10]. The major genera of these phyla found in gut microbial samples include *Bacteroides*, *Alistipes*, and *Prevotella* from the Bacteroidetes phylum, *Faecalibacterium*, and *Ruminococcus* from the Firmicutes phylum, and *Bifidobacterium* and *Collinsella* from the Actinobacteria phylum [9]. Three main enterotypes/clusters, derived from three of these specific genera, were developed to classify gut microbiotas based on their bacterial composition and their energy metabolism tendencies and capacities: Bacteroides (Enterotype 1), Prevotella (Enterotype 2) and Ruminococcus (Enterotype 3) [9,10]. However, as suggested in the study, usage of human fecal samples alone does not necessarily provide a comprehensive catalog of the gut microbial abundance and composition [9,11].

Indeed, a variety of tissues/organs lie between the mouth and anal canal, including the esophagus, stomach, and small and large intestines. Unsurprisingly, the physiological environments of the various organs/tissues of the GI tract vary considerably. Thus, although the gut microbiota is considered less diverse than other bodily microbiomes [12], microbial diversity does exist among the subsections of the GI tract. Indeed, sequencing of bacterial 16S ribosomal RNA from various regions of the GI tracts in healthy fasting adults revealed distinct microbial communities in the saliva, upper GI tract, lower GI tract, and feces [13]. For example, though *Bacteroides uniformis* levels in the salivary and upper GI regions were minimal, a significant increase was detected in the lower GI tract, which further increased in feces [13]. Additionally, diversity and heterogeneity of microbial communities are noted to decrease in regions farther down the digestive tract, owing to the increase in selective pressures and environmental changes present between the stomach and intestinal regions [13]. For instance, *Prevotella melaninogenica* abundance was high in the salivary region; yet, an insignificant decrease was noted toward the end of the upper GI tract, followed by a significant reduction in the lower GI tract, where it was nearly absent [13].

Taking a closer look at the lower GI tract, a marked environmental contrast exists between the acidic, oxygen- and antimicrobial-rich small intestine, and the more basic, oxygen- and antimicrobial-depleted large intestine (colon) [11]. Facultative anaerobes from the *Lactobacillaceae* (*Firmicutes* phylum) and *Enterobacteriaceae* (*Proteobacteria* phylum) families colonize the harsh small intestinal conditions, while anaerobes from families, such as *Bacteroidaceae (Bacteriodetes* phylum) and *Lachnospiraceae (Firmicutes* phylum), inhabit the large intestine in great numbers, owing to the tolerant conditions of the colon [11,14]. Additional factors determining microbial composition of the small intestine and colon

include differences in number, amount, and composition of mucus layers and nutrient availability (reviewed extensively in [11]).

Naturally, these differences in the bacterial composition suggest distinct functions for microbes at these particular regions. Indeed, in the intestines, the gut microbiota serves an important role in the metabolism of various endogenous and exogenous compounds, the catabolism of complex carbohydrates, amino acids, and fatty acids into smaller molecules, the synthesis of vitamins or short chain fatty acids (SCFAs), and the degradation of bile acids for the benefit of the host [2,11,12,15]. By the same token, it is also possible that a few microorganisms can metabolize some substrates and produce potentially harmful compounds, such as ethanol [16–18] and trimethylamine (TMA), or disturb the balance of certain secondary bile acids [19–21]. However, many metabolites of this microbial-mediated metabolism, especially SCFAs, are essential for host energy and signaling pathways as well as epigenetic regulation [2,11,12,15,19]. Additionally, the gut microbiota plays a role in immune regulation and host defense [15]. Specifically, the gut microbiota and its metabolites were shown to regulate T-lymphocytes and macrophages, interact with the enteric nervous system (e.g., by producing and/or increasing production of serotonin and melatonin), and mitigates pathogenic bacterial colonization [15,22–25]. Importantly, these functions may be altered or even compromised when changes to the composition of the gut microbiota take place. Indeed, the gut microbiota composition can vary not only between individuals and throughout the different stages of an individual's life, especially during infancy (thoroughly reviewed in [10]), but also in response to numerous environmental factors or other pathophysiological states.

### *2.2. Gut Microbiota Altered by Aging, Disease States, and Environmental Factors*

Changes in the gut microbiota due to normal aging or aging-related pathological states have been well described, and a significant portion of age-related changes occur during early human life [10,26]. Beginning with the transfer of maternal microbes during fetal development, numerous factors will subsequently alter and reshape the human gut microbiota, including the type of birth delivery, the method of milk administration, the weaning of the infant from milk onto solid food, and the interactions with family and the geographical environment during early development [26] (see [10] for a comprehensive list of microbial changes during development). Eventually, the fluctuations experienced by early gut microbiota as a result of these factors will give way to the adult gut microbiota, which face an array of environmental pressures [27]. Even so, the adult microbiota continues to fluctuate with continued aging, especially considering the results of one study, where notable differences in microbial composition, such as a significant increase in *Bacteroidetes* phylum members in elderly individuals compared to young adults, were observed [28]. Additionally, not only does the composition of the gut microbiota change in aging, but also the effectiveness of the microbiota in performing important functions. For instance, although the diversity of microbes actively synthesizing proteins was increased in elderly individuals, the levels of proteins involved in tryptophan and indole metabolism (TnaA and TrpB) were decreased in elderly people compared to infants. In fact, a predicted ~90% drop-off in tryptophan and indole production was observed for individuals 34 years old or above, which may alter immunological and neurological functioning [29].

In addition to normal aging and related physiological changes, pathological states can also play a role in altering the gut microbiota composition. Traumatic brain injury (TBI) represents one such example, and a recent study examining the fecal microbiome of individuals who have suffered chronic TBI (examined years after the acute incident) found significant decreases in *Prevotella* and *Bacteroides* species in addition to increases in the *Ruminococcaceae* family of bacteria, when compared to healthy individuals [30]. Additionally, the researchers suspect that the increases in *Ruminococcaceae* members, coupled with decreases in *Prevotella* species, specifically *Prevotella copri*, may explain the changes in amino acid metabolism and inflammation experienced by these individuals [30]. However, another study examining repetitive, mild TBI found no major alterations in the gut microbiota composition. Interestingly, an increase in the percent relative abundance of members from the *Desulfovibrionaceae* family was observed, which, as mentioned by the researchers, has been connected to cognitive impairments in previous studies, though no direct effect was evaluated or confirmed in this study [30,31].

Moreover, gut dysbiosis can also be affected by environmental factors such as various diets, including Western-style high-fat diets (HFDs) [32] containing fructose [33,34] or predominantly fructose alone [35]. Western-style HFDs containing n-6 polyunsaturated fatty acids can drive the incidence of obesity [36], which may progress to worse complications (e.g., insulin resistance, hypertension, etc.) characteristic of metabolic syndrome [37,38] or even type 2 diabetes. Western-style HFDs impact the resident gut microbiota, contributing to gut leakiness, lipopolysaccharide (LPS) translocation, and inflammatory damage in host intestinal tissue [39,40]. Specifically, a Western-style HFD appears to reduce the presence of carbohydrate-metabolizing proteins, in favor of proteins implicated in amino acid metabolism, with decreased amounts of the Ruminococcaceae family members, since they are involved in carbohydrate degradation [41]. Importantly, the increased abundance of members of the Proteobacteria phylum following HFD exposure to mice [42] is also physiologically important, considering that members of this phylum produce endotoxin LPS, which can enter host circulation due to HFD-mediated damage to host intestinal barrier and stimulate inflammatory damage in the GI tract and other tissues [39]. Interestingly, both Western-style high-fat and fructose diets can increase the abundance of members of the Proteobacteria phylum with a corresponding decrease in Bacteroidetes members, which, as previously stated, likely contributes to the increased leaky gut and LPS translocation [39,43]. Furthermore, chronic and binge alcohol (ethanol) consumption or exposure also affects the amount and composition of gut microbiomes [44,45], resulting in increased oxidative stress, intestinal permeability, endotoxemia, and damage in many tissues. In addition, Chen et al. [46] recently showed that high mobility group box-1 (HMGB1) contained in extracellular vesicles (exosomes) derived from gut dysbiosis can also promote non-alcoholic fatty liver disease (NAFLD) in adapter protein ASC-null mice after exposure to an HFD.

All these conditions clearly indicate the important role of changes in the amounts and composition of the gut microbiota, resulting in impairment or damage in various tissues or organs through the gut–liver–brain axis. In this review, we specifically focus on the role of alcohol-mediated oxidative stress in promoting intestinal barrier dysfunction, and fatty liver disease. In addition, we explain the mechanisms of alcohol-mediated gut dysbiosis, resulting in increased leaky gut and fatty liver (steatosis) and/or inflammation (steatohepatitis) via increased oxidative and nitrosative/nitrative stress. Finally, we briefly describe the beneficial effects of various antioxidant phytochemicals and their mechanisms of action against gut dysbiosis, intestinal barrier dysfunction, and fatty liver disease.

### **3. Oxidative Alcohol Metabolism and Progression to Alcoholic Liver Disease**

Following consumption, most alcohol (ethanol) molecules can be oxidatively metabolized to acetaldehyde and then irreversibly converted to acetate in numerous organs/tissues, including the liver, stomach, and possibly brain at very low levels. However, the primary site of oxidative alcohol metabolism occurs in the hepatocytes of the liver [47,48], although the stomach is also known to be involved in alcohol metabolism [49]. In addition, alcohol can be metabolized by the non-oxidative metabolic pathway such as fatty acid ethyl esters by cholesterol esterase, etc. [50,51]. Furthermore, unmetabolized ethanol can be excreted from the body through the breath, skin, sweat, and urine [52].

Owing to its low Km and high expression in the liver, cytoplasmic alcohol dehydrogenase (ADH) oxidatively metabolizes the majority of alcohol entering the liver into acetaldehyde, a highly reactive intermediate, while simultaneously reducing the cofactor NAD+ to NADH [47,48]. The reactive (and potentially harmful) intermediate acetaldehyde is then oxidized in the mitochondria by the low Km aldehyde dehydrogenase 2 (ALDH2) with the reduction of a cofactor NAD<sup>+</sup> to NADH, to generate acetate, which is converted

into acetyl-CoA for metabolic use or subsequently exported to other organs, including the brain [47,48]. The ADH- and ALDH-mediated oxidative ethanol metabolism pathway provides a reliable and efficient means of metabolizing the majority of consumed alcohol. Aside from the physiological consequences of alcohol overconsumption, social problems arise when alcohol consumption increases in frequency and amount. In fact, more than ~75% of alcohol-associated with sociomedical consequences are ascribed to the consumption of large amounts of alcohol in short periods of time [53,54].

Naturally, increased amount and frequency of alcohol consumption begins to deplete the cellular levels of the NAD+ cofactor by the ADH- and ALDH2-dependent reactions. A decrease in NAD+ levels, coupled with an increase of NADH levels, will cause redox changes and alter numerous cellular functions, resulting in elevated fat synthesis, decreased fat oxidation, and increased cell death processes, among others, with characteristic hallmarks of early alcoholic liver disease (ALD), specifically, steatosis and liver inflammation [47,55]. Furthermore, two additional enzymes are involved in oxidizing ethanol in the liver: peroxisome-resident catalase and endoplasmic reticulum (ER)- and mitochondrialocalized ethanol-inducible cytochrome P450-2E1 (CYP2E1), which represents a major component of the microsomal ethanol oxidizing system (MEOS) [48,56]. However, the role of catalase in hepatic ethanol metabolism is minor in comparison possibly due to the limited availability of hydrogen peroxide [56]. Thus, the CYP2E1 enzyme, constitutively expressed under normal physiological states and then induced by ethanol, at least via protein stabilization [57,58], becomes functionally important in ethanol metabolism, as well as in alcohol-mediated oxidative liver damage, particularly because CYP2E1-mediated alcohol metabolism results in the production of both highly reactive acetaldehyde and reactive oxygen species (ROS), such as the superoxide anion (O2 −) and hydrogen peroxide (H2O2) [47,59,60]. In fact, *Cyp2e1*-null mice are resistant to alcohol-mediated liver injury [61], while transgenic mice with overexpressed CYP2E1 [62] and *Cyp2e1* knock-in mice were more sensitive to liver injury by alcohol [63] or non-alcoholic substances, including a diet with 20% fat-derived calories [64]. Overwhelming increases in ROS levels strain the cellular antioxidant defense mechanisms, resulting in elevated oxidative stress, various post-translational protein modifications (PTMs), and apoptotic cellular damage through increased lipid peroxidation, ER stress, mitochondrial dysfunction, and DNA damage with genomic instability [47,65–69]. If alcohol consumption persists, hepatic damage will continue to increase and prime the liver for progression into more severe stages of ALD such as liver fibrosis, cirrhosis, and hepatocarcinoma [70,71].

ALD pathogenesis has been well reviewed elsewhere [72–75]; however, here we will briefly address the main phases of ALD and the pathological hallmarks of each stage that will be relevant when discussing the contributing role of the gut microbiota in ALD pathogenesis. The first major stage of ALD is the development of fatty liver, also termed steatosis, which involves the accumulation of lipid droplets within hepatocytes. The mechanisms underlying the development of alcohol-induced fat accumulation are numerous. Important mechanisms include a redox change with the decreased NAD+/NADH ratio during the oxidative ethanol metabolism and increased fat synthesis in the cytoplasm, through activated transcription factors, such as sterol regulatory element binding protein (SREBP-1c). In addition, ethanol intake can cause fat accumulation via decreased fat degradation, resulting from the suppressed mitochondrial enzymes for the fat oxidation pathway and acetaldehyde-mediated decreased transcription of peroxisome proliferator activator receptor α (PPARα) needed for fatty acid export and degradation, and increased import/transport of free fatty acids from adipose tissues after lipolysis. Persistent fat accumulation and oxidative stress can severely damage hepatocytes, resulting in apoptosis, which promotes the activation of liver-resident macrophages (Kupffer cells) and attracts infiltrating neutrophils, leading to inflammation and steatohepatitis. During steatohepatitis, Kupffer cells and other liver cells respond to both damage-associated molecular pattern (DAMP) molecules from apoptotic hepatocytes, in addition to other molecules, such as gut-derived pathogen-associated molecular pattern (PAMP) molecules such as

LPS, leading to the further secretion of proinflammatory cytokines and the persistence of oxidative stress, thus exacerbating liver damage [47,76]. Eventually, hepatic stellate cells can be activated by transforming growth factor-β (TGF-β) secreted by Kupffer cells attempting to resolve inflammation, which may propel the liver toward fibrosis [47]. Liver fibrosis arises as structural proteins, such as collagen and α-smooth muscle actin (α-SMA) derived from transformed hepatic stellate cells, assemble into the extracellular matrix to form a network of rigid, fibrotic scar tissue to surround damaged portions of the liver [77]. Acetaldehyde, produced during the oxidative ethanol metabolism and potentially elevated due to inactivation of ALDH2 under oxidative stress [65,78–80], is recognized to modulate important aspects of fibrosis, such as inhibiting PPARγ or increasing the transcriptional activity of C/EBPβ to stimulate collagen α1(I) expression [67]. Persistent fibrosis leads to liver failure during cirrhosis, as unresolved fibrotic tissue continues to damage liver architecture and hinders liver recovery and function [47,74]. Development of hepatocellular carcinoma can arise not only from the formation of various adducts between DNA or proteins and acetaldehyde, malondialdehyde (MDA), or 4-hydroxynonenal (4-HNE) (likely resulting at least partially from CYP2E1-mediated oxidative stress) [81], but also from the release of PAMP and DAMP molecules, that can activate immune cells and indirectly contribute to the abundance and increased activity of tumor-initiating stem-cell-like cells (TICs) [74].

With this basic overview of ALD in mind, we can now analyze the direct effect of alcohol on the gut function and the gut microbiota. Following this analysis, we will systematically address the effects of alcohol-induced changes to the gut microbiota during ALD. The contribution of specific bacterial groups and/or species to liver and gut damage following alcohol exposure will be discussed. Additionally, this review will highlight some of the numerous therapeutic options that may mitigate alcohol-induced oxidative stress, gut dysbiosis, leaky gut, and fatty liver by various dietary supplements, such as antioxidant phytochemicals, probiotics, small molecule metabolites, and traditional/ancient medications.

### **4. The Mechanisms of Alcohol-Mediated Gut Dysbiosis, Intestinal Barrier Dysfunction, and Consequences**

Over the past decade, several well-published articles have reviewed the role of alcoholinduced gut dysbiosis on ALD pathogenesis [20,82–85]. In fact, it has been reported that people with ALD, including liver cirrhosis, have elevated levels of serum endotoxin compared to control subjects [44], indicating increased intestinal permeability or gut leakiness (leaky gut). This seminal observation with AUD people was replicated by many other laboratories [86–88]. Furthermore, the elevated gut dysbiosis and leaky gut following long-term and/or binge ethanol exposure was also observed in experimental models with mice [89] and rats [86,90], indicating a common phenomenon conserved among different species. The following section will highlight alcohol-induced gut dysbiosis, the specific changes in their amounts and composition during ALD, and potential translational approaches designed to remedy these alterations.

### *4.1. The Effect of Alcohol on the Amounts and Composition of Gut Microbiota*

Many microbes are affected by the presence of alcohol in the various parts of the GI tract, and the changes in the abundances and composition of the gut microbiota have been extensively studied. Conveniently, a recent study [91] has compiled data from many publications describing changes in the gut microbiota in humans, such as individuals with AUD [92], those who have a history of chronic overconsumption of alcohol [93], and those in different stages of ALD [94]. Examining the data from a subset of these publications (in addition to several very recent publications) reveals important trends to consider for ALD and gut dysbiosis prevention.

Alcohol intake is known to increase the degree of small intestine bacterial overgrowth (SIBO), as originally reported [44,45,95]. In addition, several important phyla, including the major *Proteobacteria*, *Bacteroidetes*, *Firmicutes*, and *Actinobacteria*, are all impacted by the presence of alcohol in the GI tract. Numerous studies examining human colonic biopsies [96] and human [93] and mouse [97] feces indicate a higher abundance of members of the *Proteobacteria* phylum in response to alcohol [91]. As suggested elsewhere, this change can conceivably result from the ability of microbes in this phylum to persist in the high ROS environment generated by increased inducible nitric oxide synthase (iNOS) following alcohol exposure [98], since they are predominantly facultative anaerobes, which can withstand these conditions [92]. Specifically, at the family level, *Enterobacteriaceae* abundance was increased [91,92,96], in addition to elevated levels of certain genera from this family, such as *Escherichia* [91,92,96]. Owing to their Gram-negative status and endotoxin (e.g., LPS) producing capabilities, *Proteobacteria*, such as those from the genera *Escherichia*, are unsurprisingly seen as potential instigators of gut barrier dysfunction during alcohol consumption [93,97]. In particular, the ability of species in the *Escherichia* genus to produce harmful PAMP molecules, such as LPS [93], and to metabolize alcohol (in some strains) [99], producing the toxic metabolite acetaldehyde, supports the hypothesized harmful role of *Proteobacteria* in alcohol-induced gut dysbiosis. Additionally, although the *Proteobacteria* genera *Sutterella* displayed a decreased relative abundance in the colonic biopsies of heavy drinkers [96], another study examining the feces of individuals with a history of chronic alcohol overconsumption found a higher relative abundance of this genera and views the *Sutterella* increase in light of previous studies reporting its role in promoting inflammation [93]. This same study provides evidence for an inverse relationship between *Proteobacteria* and the presence of the anti-inflammatory SCFA butyric acid, which was decreased in the feces of individuals with a history of chronic AUD, although as noted by the authors, confirming this correlation is hindered by the nature of the study [93].

Alcohol exposure has been shown to decrease the abundance of members of the phylum *Bacteroidetes* in the feces of mice exposed to chronic alcohol [97] and from colonic biopsies [87] and feces [92] of AUD individuals. However, several other reports using chronic alcohol mouse models have described an increase in *Bacteroidetes* presence following the sequencing of colonic and cecal contents [20,100,101]. More conflicting results emerge when examining the genus *Bacteroides* in this phylum. Although one recent study did not detect any difference in the abundance of *Bacteroides* members in colonic biopsies from heavy drinkers [96], other studies examining stool from AUD patients found an increase in *Bacteroides* members [102], while the sequencing of feces from alcoholic individuals [92] found decreased abundance of this genus. As mentioned by the authors, unlike facultative anaerobic *Proteobacteria, Bacteroides* members are obligate anaerobes and may, therefore, struggle to survive in the presence of significant ROS during prolonged alcohol exposure, which may explain the decrease in this study [92]. Although *Bacteroides* are not active ethanol metabolizers [92], they are known to play a role in the metabolism of bile acid, which could interfere with farnesoid X receptor (FXR) signaling (due to its regulation by bile acids) [20]; however, elevated Bacteroides levels could lead to increased production of the gamma-aminobutyric acid (GABA) neurotransmitter [102], suggesting a potential interplay between members of the *Bacteroides* genus and the brain during ALD. Interestingly, one study showed that members of another genus, *Prevotella*, have the capacity to metabolize ethanol and generate acetaldehyde in vitro, suggesting a possible role for members of this genus in contributing to acetaldehyde production in vivo [92]. However, like the *Bacteroides* data, conflicting reports have emerged regarding the relative abundance of *Prevotella* during alcohol exposure, where *Prevotella* members displayed increased abundance in the stool of AUD patients [102]; yet, numerous studies have described a decrease in the relative abundance of members from this genus [91].

In a mouse model, the presence of members of the *Actinobacteria* phylum was noted to increase in the feces of alcohol-exposed mice, and this elevation, in conjunction with the amplified presence of *Proteobacteria,* has been suggested to play a role in the intestinal manifestations of ALD [97]. Indeed, several interesting genera from this Gram-positive phylum are altered in the gut following alcohol exposure, including *Corynebacterium, Bifidobacterium*, and *Collinsella* [91,97]. *Corynebacterium* was found to be increased in the

feces of mice chronically exposed to alcohol and, although the relevance of this elevation in ALD pathogenesis remains unknown, the authors posit that this amplification could be noteworthy, since other studies have found *Corynebacterium* infection in ALD [97,103,104]. The other genera, *Bifidobacterium* and *Collinsella*, display conflicting alterations among studies. While *Bifidobacterium* was decreased in human feces of individuals who habitually drink alcohol [92], this genus was found in increased abundance in the feces of active alcoholic patients with cirrhosis and severe alcoholic hepatitis, compared to individuals with cirrhosis and lacking severe alcoholic hepatitis [94]. Similarly, while *Collinsella* was found in increased abundance in the stool of AUD individuals [102], analysis of the relative abundance of *Collinsella* showed no significant change in the abundance of these genera in the feces of individuals who habitually consume alcohol [92]. Interestingly, both *Bifidobacterium* and *Collinsella* were characterized as potential acetaldehyde accumulators in in vitro aerobic conditions [105], and, in particular, the absence of *Bifidobacterium* was hypothesized to contribute to the observed decrement in alcohol metabolism in the feces of alcoholic individuals [92]. Some have postulated that the ability for *Collinsella* to metabolize ethanol may permit their observed overgrowth in the stool of AUD patients [102], or could potentially allow them to persist longer than other microbes.

Although the phylum *Firmicutes* was demonstrated to decrease in abundance in mouse fecal samples following chronic alcohol exposure [97], this diverse phylum has several genera that were both increased, including *Streptococcus, Coprobacillus, Holdemania,* and *Clostridium*, and decreased, including *Ruminococcus, Faecalibacterium, Subdoligranulum, Roseburia*, and *Lactobacillus*, among others, following alcohol exposure [91–93,96,101,102]. *Roseburia* and *Lactobacillus*, two therapeutically relevant bacterial genera, showed decreased abundance in colonic biopsies of heavy drinkers [96] and colonic contents of rats chronically exposed to alcohol [101], respectively. Similarly, *Faecalibacterium* and anti-inflammatory members of the genus, such as *F. prausnitzii*, are decreased in the feces of heavy drinkers [96] and individuals who have a history of AUD [93]. Expectedly, these decreased levels of *F. prausnitzii* during alcohol exposure likely impact the levels of beneficial SCFAs in the intestines [91], such as butyrate, and, unsurprisingly, a positive correlation was observed between *Faecalibacterium* and butyric acid levels in the feces of AUD individuals [93]. Overall, the Firmicutes phylum is quite diverse. While obligate anaerobes from the Ruminococcus genus are observed to decrease [92], likely the result of the increased oxidative stress in the gut following alcohol consumption, as previously reported [98], facultative anaerobes from the Streptococcus genus have been shown to elevate in both the stool of patients with AUD [102] and in the feces of alcoholic individuals [92]. Indeed, as others have postulated, the observed decrease in obligate anaerobes may give members of the Streptococcus genus (and other facultative anaerobes) an opportunity to proliferate, which may be of concern, considering that infections from Streptococcus members have been noted during cirrhosis [91].

Lastly, from the phylum *Verrucomicrobia*, the *Akkermansia* genus was noted to decrease in both the stool of people with AUD [102] and in colonic biopsies from heavy alcohol consumers [96]. The *Akkermansia* genus and, specifically, *Akkermansia muciniphila*, have numerous beneficial roles in the intestines, including protecting the gut barrier and aiding in the production of epithelial cell-protective mucus [91,96,102], and supplementation was shown to prevent manifestations of ALD in binge and chronic mouse models of alcohol exposure [106]. Importantly, in patients with AUD, low *Akkermansia* levels negatively correlated with the inflammatory marker MCP-1, indicating a potentially significant role for these microbes in inflammation regulation during alcohol consumption [102].

Importantly, when analyzing the alcohol-induced changes in the composition of the gut microbiota, one needs to take special consideration for the specific tissue/sample being examined. For instance, examining microbial changes in the feces of mice following alcohol exposure [97] or humans who have consumed alcohol at some time [92,93] does not necessarily provide a comprehensive assessment of the gut-wide microbial status. As suggested by others, analysis of microbial changes in samples other than feces is needed to pinpoint changes in the diverse environments of the gut [91]. For example, one study examining the microbial changes in jejunal and colonic contents of rats chronically exposed to alcohol found significant changes in colonic microbiota, but hardly any impact on composition in the jejunal microbiota [101]. Thus, future research may target specific areas of the intestines (duodenal, jejunal, ileal, colonic, rectal, etc.), which should provide a more complete profile of the gut microbial changes following alcohol exposure.

### *4.2. Mechanisms of Alcohol-Induced Damage to the Intestines, Resulting in Leaky Gut and Endotoxemia*

Before ethanol molecules reach bodily destinations, such as the liver and brain, they must first pass through the tissues of the GI tract. Alcohol can be absorbed in the mouth and esophagus (albeit in limited quantities), and absorption rates will begin to increase further down the GI tract, especially at the stomach, duodenum, and jejunum and, to a lesser extent, at the ileum [107,108]. Alcohol absorption at the proximal regions of the small intestine (duodenum and jejunum) [109] will result in the passage of these molecules into the capillaries and blood vessels, including the portal vein, where they will be delivered to the liver [110]. Although alcohol passes from the intestinal lumen into the bloodstream through simple diffusion [107,111], ethanol molecules first must pass through the layers of the intestinal barrier and can trigger changes at these various regions.

Considering the structure of the small and large intestinal barrier, both consist of a monolayer of intestinal epithelial cells with a layer of mucus on the luminal side and the immune cell-rich lamina propria on the non-luminal side. Aside from the chief intestinal epithelial cell, the enterocyte, the monolayer can contain numerous other cell types. These include intestinal epithelial stem cells (IESCs), which can differentiate to replace cells in the monolayer, Paneth cells, which produce antimicrobial peptides and regulate IESCs, and goblet cells, which produce and secrete mucin glycoproteins for incorporation into the luminal mucus layer [112,113]. Several differences exist between the small and large intestinal barriers. Specifically, unlike the small intestine, the colon contains two mucus layers, a loose and a thick layer [11], a proportionally greater number of goblet cells, and an absence of Paneth cells [112].

In the intestinal lumen, ethanol will encounter the mucus layer, which contains glycosylated mucin proteins and lipids, among other components, and serves as both a protective barrier and a regulator of the microbial environment, whereby species, such as *Akkermansia muciniphila*, will use barrier-generated mucus resources [11,112]. Interestingly, mRNA levels of mucin proteins 1-4 do not increase in the small or large intestines in response to binge alcohol [114]. However, examining the glycosylation status of mucin glycoproteins, studies using chronic models of alcohol exposure found altered patterns of mucin glycosylation in the intestinal mucosa, especially increased galactosylation [115], which may influence bacterial binding to the mucus layer or may prevent mucus adherence to the intestinal barrier, as demonstrated in a gastric mucosa study [116]. Importantly, one study has suggested that ethanol may also decrease the hydrophobicity of the mucus layer, specifically, through the ethanol-mediated dissolution of mucus-layer free fatty acids, which function as an absorptive barrier, thus likely promoting ethanol-induced gut permeability dysfunction [117].

Eventually, from the mucus layer, ethanol will pass into the monolayer through simple diffusion, where it will either continue to diffuse into the circulation for delivery to various bodily sites or be metabolized in the barrier [108]. Indeed, the presence of ADH and ALDH isozymes and their activities have been detected in the small and large intestines, although their contribution to alcohol metabolism is not as significant as the liver [118]. Although differences in ADH and ALDH expression vary across the intestinal landscape, importantly, ADH expression and activity appears greater in both the small and large intestines compared to those of ALDH, suggesting a greater buildup of reactive acetaldehyde over acetate in the monolayer following alcohol metabolism [118]. In addition, the presence of CYP2E1 and its elevated levels in the GI tract after alcohol exposure [119–121] are likely to produce elevated levels of ROS and acetaldehyde, especially due to low levels of ALDH2

expression in the gut [122,123]. As reviewed thoroughly elsewhere [108], the metabolism of ingested alcohol can also occur through non-oxidative pathways. Ingested alcohol may also interact with and be metabolized by gut microbes. Indeed, some species are capable of metabolizing ingested alcohol, such as certain strains of *Escherichia coli* [99], while other gut microbes display limited alcohol metabolism ability, such as some members of the lactobacillus genus [124]. Additionally, a recent study examining the feces of Japanese alcoholic individuals categorized several bacterial groups, such as the *Ruminococcus* and *Collinsella* genera, as potential acetaldehyde accumulators for their ability to metabolize ethanol in aerobic in vitro conditions [105,125].

With these factors in mind, along with the previous discussion of the alcohol-mediated alterations to the gut microbial composition, we can now examine the mechanisms and impact of gut dysbiosis and alcohol exposure on the gut barrier, specifically, on the increased intestinal permeability observed following chronic and/or binge alcohol exposure. Normally, the intestinal monolayer forms a tightly linked barrier with various proteins forming the intestinal tight junction (TJ), adherent junction (AJ), and desmosomes that keep microbes in the intestinal lumen out of the bloodstream [126], while also permitting the passage of luminal nutrients into the bloodstream, thus ensuring a useful, non-toxic blood supply for recipient organs, such as the liver via the portal vein [127]. However, alcohol exposure can alter the permeability of this intestinal barrier, resulting in an influx of harmful luminal molecules, such as bacterial endotoxin LPS and peptidoglycan, which can induce inflammation and oxidative damage both in the gut and in other organs [127]. Based on this information, we will concisely address the role of ethanol, acetaldehyde, ROS, and other factors in triggering this intestinal permeability dysfunction and will subsequently address the impact of this leaky barrier on the various stages of ALD pathogenesis, as illustrated in Figure 1. This review will place a particular emphasis on the role of oxidative and nitrosative/nitrative stress in gut dysfunction and its impact on ALD pathogenesis, in addition to the mechanistic role of enterocyte apoptosis and PTMs of paracellular junctional complex proteins in initiating gut barrier dysfunction (leaky gut), leading to inflammatory liver injury (Figure 1).

Early studies determined that exposure of ethanol to the human colonic Caco-2 cell line triggered apoptosis in these epithelial cells [128]. Although a mouse model of chronic ethanol exposure alone displayed no significant changes in apoptotic protein markers or intestinal permeability in the jejunum [129], binge alcohol exposure elevated endotoxin levels in rodent models, pointing to the development of a leaky gut in these animals [130,131]. The effects of binge alcohol exposure on time- and dose-dependent gut permeability change, elevated endotoxemia, and fatty liver injury were confirmed by other laboratories [121,132,133]. Mouse and rat models of binge alcohol exposure showed increased levels of apoptotic protein markers, such as BAX and cleaved caspase-3, and histological evidence for the apoptosis of intestinal epithelial cells, consequently indicating increased gut permeability [121]. In rodent models, treatment with antioxidants or a specific inhibitor of CYP2E1 significantly prevented binge alcohol-mediated leaky gut and fatty liver disease, while *Cyp2e1*-knockout mice were also quite resistant to these changes [120,121,132]. These results clearly indicate the important roles of CYP2E1 and consequent oxidative stress in promoting intestinal barrier dysfunction and ALD. Furthermore, a recent study using a mouse model of chronic plus binge alcohol exposure found increased apoptotic markers in the proximal small intestine, likely mediated by ER stress, and this corresponded to the observed increase in bacterial product translocation at this region [134]. Importantly, however, the effect of ethanol on the intestinal barrier is not limited to the induction of apoptosis in the epithelial cells within the gut monolayer.

**Figure 1.** Schematic overview of gut–liver communication and damage prompted by intestinal disorders or consumption of exogenous agents. Numerous exogenous agents (e.g., alcohol, high-fat diet (HFD), fructose, etc.) or underlying intestinal disorders (e.g., Crohn's disease, ulcerative colitis, etc.) can elicit changes to the abundance and/or composition of the gut microbiota. Elevated gut CYP2E1 and NADPH oxidases (NOXs) can increase oxidative stress. The resulting gut dysbiosis alters gut metabolism and damages the intestinal barrier through various mechanisms, including the oxidative stressmediated post-translational modifications (PTMs), leading to decreases in paracellular junction complex proteins. Sustained damage to the barrier causes gut leakiness and, subsequently, a gut-localized immune response and increased levels of harmful gut-derived compounds (e.g., lipopolysaccharide (LPS), peptidoglycan, exosomes, etc.) into the circulation. LPS (and other gut-derived metabolites) and alcohol will reach the liver and drive alcoholic liver disease (ALD) pathogenesis and progression. Kupffer cell activation, mediated by LPS and oxidative stress driven by metabolism of the ethanol by hepatic CYP2E1 and from activated NOXs, increases inflammatory cytokine levels (e.g., TNF-α, etc.) and instigates hepatocyte apoptosis. Eventually, sustained oxidative stress, LPS infiltration, and hepatocyte damage will lead to the activation of hepatic stellate cells, driving liver fibrosis and continued liver damage.

Several important proteins that normally connect cells of the monolayer and enhance the impermeability of the intestinal barrier are affected by ethanol. To limit the travel of molecules between adjacent monolayer cells (the paracellular pathway), TJ complexes composed of proteins, such as claudins, occludin, and zonula occludens-1 (ZO-1), along with AJ complexes, composed of proteins, such as E-cadherin and βcatenin [135,136], form between monolayer cells to prevent the mass influx of particles into the bloodstream. Exposure of ethanol to Caco-2 cells resulted in a time-dependent increase in epithelial cell permeability, which correlated with the time- and dosedependent decrease in ZO-1 and increase in claudin-1 protein levels [137]. Additionally, ZO-1 and claudin-1 were noted to be irregularly distributed upon localization. A similar observation was noted in Caco-2 cells treated with ethanol (10, 20, or 40 mM) for 3 h, whereby ZO-1 and occludin showed decreased localization to the membranes for intercellular interactions, which correlated with increased barrier permeability, despite no observed change in ZO-1 (and other tight junction proteins) mRNA levels [138]. Thus, the presence and proper localization of junctional complex proteins involved in paracellular transport are crucial for maintaining intestinal integrity, especially considering that occludin knockout mice exposed to ethanol displayed both increased permeability at the colon and increased triglyceride accumulation in the liver compared to wild-type (WT) mice exposed to ethanol [139]. In addition, ethanol and consequently CYP2E1-mediated oxidative and nitrosative/nitrative stress appear to induce changes

to the PTM landscape of intestinal proteins on a global level, with regard to increased acetylation, nitration, and ubiquitination [120] as well as phosphorylation [140–143].

### *4.3. Mechanisms of Gut Leakiness via Oxidative Stress and PTMs of Paracellular Proteins*

Interestingly, ethanol appears to induce changes to the PTM landscape of intestinal proteins on a global level, with regard to acetylation, oxidation, nitration, and ubiquitination [120,121]. One recent study found that the removal of intestinal NAD+-dependent class III deacetylase SIRT1 blunts alcohol-induced liver damage in an acute on chronic alcohol mouse model, likely through prevention of ferroptosis, which may indicate a potential role for global acetylation in intestinal barrier function in a model of ALD pathogenesis [144]. However, in particular, various PTMs of proteins involved in the blockage of paracellular transport, such as those comprising TJ, AJ, and cytoskeletal proteins, have received considerable interest for their possible role in stimulating gut leakiness following alcohol exposure. Indeed, many paracellular transport proteins, especially TJ proteins, are known to be post-translationally modified, which play a role in their functional capabilities [145,146].

In the context of alcohol exposure, acetylation of α-tubulin appears to interfere with ZO-1 recruitment to the membrane, which likely contributes to increased permeability observed in the in vitro Caco-2 cell model following exposure to ethanol or acetaldehyde [138]. Additionally, increased nitration and ubiquitin-conjugation of α-tubulin in the intestines of mice exposed to binge alcohol correlated with decreased protein levels of α-tubulin and endotoxemia in these mice [121]. Modification of specific TJ and AJ proteins has also been well documented. In this same study, increased nitration and ubiquitin-conjugation of β-catenin (AJ), plakoglobin (adherens junction/desmosomes), claudin-1 (TJ), and claudin-4 (TJ) was observed in the intestines of binge alcohol-exposed mice, which, once again, correlated with endotoxemia and, thus, gut leakiness and fatty liver disease [121]. The significantly decreased gut TJ and AJ proteins in binge alcohol-exposed rats compared to controls were further confirmed by quantitative mass-spectral analysis. Additionally, the role of paracellular protein phosphorylation in both TJ and AJ formation has been well characterized and is altered by acetaldehyde [140,147]. Specifically, the presence of acetaldehyde leads to the persistent phosphorylation of the TJ protein occludin and AJ proteins E-cadherin and β-catenin, which are hypothesized to prevent the binding of these proteins to actin filaments, evidenced by the decreased amounts of these proteins in actin-rich Triton-insoluble fraction, which, as indicated by the authors, is an accurate predicator of TJ and AJ complex integrity [148]. Further studies have pinpointed the specific enzymes involved in regulating paracellular protein phosphorylation status following alcohol exposure. The protein phosphatase 2A (PP2A)-mediated dephosphorylation of threonine residues on occludin [149] and the altered activity of protein tyrosine phosphatase 1B-mediated dephosphorylation of AJ proteins [147] are two such examples of alcohol-induced changes to the paracellular protein landscape. Further supporting the role of phosphorylation in paracellular protein regulation, the results of a study examining brain endothelial cells exposed to ethanol suggest that the increased permeability of bovine brain microvascular endothelial cells (BBMEC) observed following ethanol exposure is linked to the phosphorylation of claudin-5 and occludin, likely mediated by myosin light chain kinase (MLCK) [150].

Acetaldehyde has also received considerable attention for its ability to alter the intestinal barrier function following alcohol consumption, and this subject has been thoroughly reviewed elsewhere [151]. Like ethanol, acetaldehyde was also shown to increase membrane permeability of an in vitro 3-D Caco-2 spheroid model exposed to acetaldehyde concentrations as low as 0.025 mM; yet, this exposure did not affect the mRNA levels of TJ proteins when this same model was exposed to 0.2 mM acetaldehyde [138]. Additionally, acetaldehyde also affects the localization of ZO-1 and causes an increase in global protein acetylation, including α-tubulin [138]. Indeed, considerable research has been devoted to determining the role of acetaldehyde in altering the localization of ZO-1 and other paracellular barrier proteins (for a comprehensive list, see [151]). As previously

mentioned, various PTMs appear to play a significant role in determining the integrity of the barrier, whereby the presence of acetaldehyde leads to the persistent phosphorylation of several paracellular proteins, thus contributing to gut leakiness [140]. Furthermore, in vivo studies using WT and ALDH2(+/−) mice demonstrated increased intestinal permeability at the distal and proximal colon, jejunum, and ileum in ALDH2(+/−) mice following ethanol exposure, compared to the corresponding WT, which only displayed increased permeability at the distal colon [152]. This suggests an important role for acetaldehyde in mediating gut leakiness and for ALDH2 in moderating elevated acetaldehyde levels and thus affecting the re-distribution of TJ and AJ proteins in the mouse ileum and colon [152]. Interestingly, acetaldehyde was also shown to negatively impact an in vitro model of mucin-producing goblet cells, particularly through another catalyst of intestinal barrier dysfunction: oxidative stress [153].

Indeed, many adverse cellular manifestations, such as lipid peroxidation, and DNA damage arise as a result of oxidative stress, which ultimately leads to inflammation and apoptosis [47,74]. Oxidative stress was observed in the intestines of rats exposed to chronic alcohol conditions, evidenced by the increase in iNOS protein levels, nitrate and nitrite levels, and jejunal, ileal, and colonic protein nitration and oxidation, which correlated with increased intestinal permeability [98,154,155]. Specifically, the iNOS-dependent increase in miR-212 levels in Caco-2 cells exposed to ethanol appears to contribute to perturbations in the cell permeability [156]. In addition, an iNOS-mediated decrease in ZO-1 expression was observed in the ethanol-exposed Caco-2 cells [156]. The role of iNOS in instigating in-vitro increases in Caco-2 cell permeability was confirmed in-vivo using iNOS-KO mice, which displayed significantly less intestinal barrier dysfunction compared to the corresponding WT mice following exposure to alcohol in both mouse strains [156]. Additionally, elevated and activated iNOS appears to contribute to the activation of the Snail transcription factor, which was demonstrated to play a role in increased gut permeability [157]. However, iNOS is not the only enzyme responsible for the alcohol-induced development of oxidative stress. In a binge alcohol rat model, both iNOS and CYP2E1 protein levels were elevated in the intestines 1 or 2 h following the last binge dose [132]. CYP2E1 is able to oxidize ethanol into acetaldehyde, but in doing so, generates ROS molecules, such as the 1-hydroxyethyl radical [60,158]. Indeed, the role of intestinal CYP2E1 on gut leakiness observed following alcohol exposure has been well reviewed and involves important circadian proteins, such as CLOCK and PER2 [155]. Importantly, the observed increase in plasma endotoxin levels, oxidative stress (determined via increased 3-nitrotyrosine (3-NT) levels), and intestinal permeability was dependent on intestinal CYP2E1, since *Cyp2e1*-null mice were quite resistant to binge alcohol-mediated gut leakiness and subsequent fatty liver [120,132,159]. Additionally, the levels of nitrated and ubiquitin conjugated TJ and AJ proteins were markedly diminished in the same *Cyp2e1*-null mice exposed to binge alcohol, suggesting a role of CYP2E1 and/or CYP2E1-generated oxidative and nitrosative/nitrative stress in modulating the intestinal PTM landscape [120,132,159].

Furthermore, it is also possible that ROS can be provided through activated NADPHoxidase isozyme(s) (NOXs) present in the colon and in immune cells in the lamina propria of the GI tract [32,160,161]. However, treatment of Caco-2 cells with acetaldehyde, ethanol, and the NADPH oxidase inhibitor diphenyleneiodonium did not ameliorate evidence of increased barrier permeability and dysfunction as *N*-acetyl cysteine was demonstrated to do [162]. The oxidative stress-induced mitogen-activated protein kinases (MAPKs) [e.g., p38 protein kinase (p38k), c-Jun *N*-terminal protein kinase (JNK), and extracellular signal regulated protein kinase (ERK)] are also implicated in gut barrier damage, since they are observed to be phosphorylated (activated) in Caco-2 cells and played a role in TJ disruption and gut barrier dysfunction, particularly through increasing MLCK mRNA [163], whose activity has been implicated in regulating intestinal permeability [164]. The oxidative stress-mediated decrease in hepatocyte nuclear factor-4α (HNF-4α) was also implicated in tight junction disruption following alcohol exposure [165]. Additionally, studies have implicated the fatty acid ethyl esters ethyl oleate and ethyl palmitate in increased ROS

production in Caco-2 cells and these esters also caused redistribution of the paracellular proteins ZO-1 and occludin and caused increased ROS-mediated permeability of these cultured cells [166].

### *4.4. Crosstalk among Gut Dysbiosis, Intestinal Barrier Dysfunction, and ALD*

Bidirectional gut–liver communication occurs as gut-derived molecules pass through the intestinal barrier and enter the portal vein to reach the liver, while liver-derived molecules pass through the biliary tract to interact with the intestines [20,167]. Typically, in a resting, non-disease state, SCFAs, secondary bile acids and other diet-derived metabolites will pass into the liver to perform various functions, such as driving fatty acid oxidation in the case of SCFAs [168], or, in the case of secondary bile acids, which have been modified in the gut, these metabolites can be recycled for future use [20]. Concurrently, the liver will secrete primary bile acids, among other liver-derived molecules, to the intestines to increase lipid absorption [169] and potentially alter the gut microbial landscape through their antimicrobial properties [170]. Heavy alcohol consumption will alter communication between the gut and liver, not only by increasing the presence of circulating alcohol [108] and subsequently produced acetaldehyde [171], which can damage the liver at high concentrations, but also by enabling harmful gut microbiota-derived molecules to leak into the circulation and inflict damage on both the intestinal barrier and the liver [167].

The non-luminal side of the intestinal barrier is home to the lamina propria containing a wide variety of immune cells, which can respond to changes in gut barrier dysfunction. The response of immune cells (specifically leukocytes) to alcohol consumption does not appear to be uniform across the lower GI tract [172], possibly due to the differences in microbial composition and abundance found throughout the GI tract, gut environmental factors, and/or differing ethanol concentrations at these regions. Predictably, following ethanol consumption, these cells and others will interact with and be activated by luminalderived molecules (LPS/PAMPs) that traversed the damaged, leaky intestinal barrier [127]. LPS is a significant luminal-derived molecule that will enter circulation and directly alter local (gut) regions, triggering inflammation and injury to downstream organs, such as the liver, inflicting further damage [127,172] and enterocyte barrier dysfunction [173]. Although passage of LPS through the intestinal barrier has long been speculated to occur via paracellular transport, a recent study examining LPS transport across the intestinal barrier during lipid absorption (non-alcohol conditions) observed CD63- and lipid raftmediated transcellular transport of LPS across the barrier and absorption into the portal vein [174]. Regardless of the transport method across the barrier, the manifestation of endotoxemia is a common occurrence in individuals with ALD [175,176] and serum LPS levels appear to correlate with the amount of alcohol being consumed [177]. However, before LPS and other gut-derived harmful molecules (ethanol, acetaldehyde, cytolysin, candidalysin [178], DAMPs, exosomes, etc.) can travel through the portal vein to the liver, the gut-localized immune response will be affected [174]. For example, an early study examining mice chronically exposed to alcohol demonstrated that the mRNA levels of certain cytokines (IL-1β and TNFα) are upregulated in the ileum by ethanol alone, while others (IL-6 and IL-11) are upregulated in the presence of both ethanol and LPS [179]. A more recent study using a binge on chronic mouse model of alcohol exposure found increased production of IL-17 from Paneth cells, which contributed to inflammasome activation in the small intestine, evidenced by increases in activated caspase-1 and IL-18 [134]. Furthermore, alcohol exposure has been shown to alter populations of immune cells in the gut, such as T-lymphocyte populations in the small intestine of Rhesus macaques chronically exposed to alcohol [180]. Importantly, considering that NADPH oxidase levels are increased during exposure of mice to the Western-style HFD [32,161] and play a role in inflammation of the intestines during dextran sodium sulfate (DSS)-mediated colitis [160], it is possible that ROS generated by NADPH-oxidase isozyme(s) present in immune cells in the GI tract may also contribute to gut-localized inflammation in response to alcohol.

LPS and other luminal-derived molecules will travel through the portal vein and reach the liver, where they will interact with various cell types and influence the different stages of ALD pathogenesis [181–183]. Briefly, depending on the specific DAMP and/or PAMP, gut-derived molecules will interact with specific toll-like receptor (TLR) complexes present in resident liver macrophages, Kupffer cells (and other liver cells) to initiate a specific inflammatory cascade [181–183]. In particular, the pathway by which LPS induces inflammatory cascades in the liver has been investigated thoroughly [184]. Briefly, after the binding of LPS to LPS-binding protein (LBP) and CD14, LPS will bind MD2 to interact with TLR4 on the membrane, thus activating the complex [181,182,184]. The resulting TLR4 activation can elicit intracellular signaling cascades which, depending on the specific proteins involved [182,184], can increase the expression of proinflammatory cytokines, such as TNFα, which will both increase apoptotic liver damage and hepatic inflammation [185,186]. Indeed, LPS has been shown to drive steatosis and inflammation through mechanisms such as decreased autophagic response in the liver [187] and increased pro-inflammatory cytokine production [188], respectively. Additionally, the LPS-mediated induction of the liver damage has also been implicated in increased hepatic stellate cell (HSC) response to TGF-β and fibrosis onset, although LPS may also exhibit anti-fibrotic properties through by targeting HSC proliferation, as recently reviewed [189]. Importantly, a key mediator of gut-induced damage of the liver is oxidative stress and, specifically, ROS, generated not only by LPS-activated Kupffer cells but via the oxidative metabolism of ethanol by CYP2E1 as high concentrations of ethanol enter the liver from the circulation [67].

A recent study demonstrated that increased ROS levels, likely generated by CYP2E1, occur alongside ethanol-induced decreases in autophagy in alcohol-exposed mice [187]. Furthermore, studies have confirmed the role of CYP2E1 in LPS-induced liver damage via the production of ROS and activation of oxidative stress-sensitive downstream kinases, such as the MAP kinases, and mitochondrial damage [190,191]. MAP kinases are not only activated by CYP2E1-generated ROS [68,79], but also by the NADPH oxidase-mediated production of ROS, which activate ERK in a CYP2E1-independent manner, leading to TNF-α production [192]. Another study found that arachidonic acid supplementation with alcohol activated ERK via ROS and, subsequently, increased TNF-α levels, supporting the notion that other factors, such as dietary n-6 fatty acids, can act alongside LPS in inducing oxidative stress-mediated liver damage following alcohol exposure [193] or during other liver pathologies, such as palmitic acid-mediated NAFLD [194]. In a binge alcohol model, NADPH oxidase-mediated ROS production is also important for inflammatory signaling by increasing interleukin-1 receptor-associated kinase (IRAK) levels in Kupffer cells 21 h post binge alcohol exposure, a change that was dependent on NF-κB activity and which correlated with increased TNF-α levels in these Kupffer cells [195]. Interestingly, at the PTM level of regulation, LPS and acetate (and/or acetaldehyde) were demonstrated to decrease hepatic SIRT1 levels, with, expectedly, increased hyperacetylation of a subunit of nuclear transcription factor kappa B (NF-κB), leading to increased inflammatory response in in vitro rat Kupffer cells [196]. Though not explicitly evaluated, the authors hypothesized that ROS may regulate this mechanism.

Kupffer cell-localized NADPH oxidase (NOX) was hypothesized early on to generate ROS during infiltration of LPS and/or neutrophils following alcohol exposure [197] and the role of specific members of the NOX family of NADPH oxidases, such as NOX4, has been described. Indeed, recent in vitro and in vivo models of alcohol-induced liver damage confirmed the role of NOX4 in increasing mitochondrial ROS and mitochondrialmediated apoptosis, in addition to a partial role in steatosis development following alcohol exposure [198]. Translational approaches to combat macrophage-mediated ROS production during ALD have also revealed the mechanistic regulation of macrophage-localized NADPH oxidase, as globular adiponectin was shown to inhibit ROS and NOX2 expression through activation of liver kinase B1 (LKB1) and AMP-dependent protein kinase (AMPK) [199]. Besides immune cells, HSCs can also be targeted by LPS and, mechanistically, oxidative stress also appears to drive LPS-mediated increases in MCP-1 and IL-6 in HSCs [200]. Although gut-mediated hepatic oxidative stress represents an important contributing factor to ALD pathogenesis and progression, not all aspects of ALD pathogenesis appear to rely on oxidative stress, since other factors, such as insulin resistance, may also play a role [201].

### **5. The Antioxidant Properties, Metabolisms, and Health Benefits of Various Phytochemicals against Gut Dysbiosis, Intestinal Barrier Dysfunction, and Fatty Liver Disease**

As described in the previous sections, gut dysbiosis with decreased ratios of Bacteroidetes/Firmicutes and changes in the levels of various endogenous metabolites, including ethanol and SCFAs, are associated with many disease states. Thus, there have been many efforts to normalize or restore the gut microbiome by using different diets, such as microbiota-targeted vegan (vegetarian) diets and/or consuming microbiota-accessible carbohydrates (oligocarbohydrates) or dietary supplements with various phytochemicals [19]. In addition, many phytochemicals represent diverse classes of antioxidants contained in a variety of fruits and vegetables as well as medicinal plants [202]. Most of these phytochemicals are known to have very low bioavailabilities due to their water insolubilities [202,203]. These phytochemicals include different chemical classes, such as various polyphenols, lignans such as phytoestrogens, carotenoids, phytosterols/phytostanols, alkaloids and glucosinolates, sulfur-containing compounds, etc., as recently reviewed [202]. In one class of phytochemical, there are many chemical subgroups. For instance, polyphenols represent antioxidant chemicals of many different subgroups. Common subgroups include flavonoids (e.g., apigenin, quercetin) polyphenolic amides (e.g., capsaicin) and polyphenolic acids [204], which include important polyphenols such as resveratrol, curcumin, capsaicin, quercetin, rutin, genistein, daidzein, ellagic acid, and proanthocyanidins (tannoids), such as (-)-epicatechin (EC), (-)-epigallocatechin (EGC), (-)-epicatechin-3-gallate (ECG), and (-)-epigallate catechin-3-gallate (EGCG), among others, as recently reviewed [205]. Despite the different chemical structures of these phytochemicals, they have exhibited their beneficial effects in some human studies, many experimental rodent models, and in vitro cell culture models.

One of the characteristic properties of these phytochemicals is their very low water solubility, despite the presence of many hydroxyl groups in polyphenols [206,207]. It is known that some phytochemicals can be metabolized or modified by intestinal bacteria (specifically hydrolyzed, reduced, deglycosylated, degraded, conjugated, etc.) and some of these alterations may enhance absorption of these chemicals in the intestine [208–210]. For instance, resveratrol, contained in grapes, berries, and red wine, can be metabolized to dihydroxyresveratrol, 3,4 dihydroxybibenzyl (lunularin), and 3,4 -dihydroxy-trans-stilbene by gut bacteria or converted to the more bioavailable metabolite piceid [205,211]. Likewise, 3,4-dihydroxyphenylacetic acid, a gut-derived metabolite of the flavonoid quercetin, found that various fruits, vegetables, and beverages (mostly as quercetin glycosides) [212] can both act as a free radical scavenger and reduce markers of advanced glycation end products [194]. A multitude of polyphenols (e.g., flavonoids, thearubigins, chlorogenic acids) are known to be metabolized by gut microbiota [213] as well as conjugated, glucuronated, deglycosylated, etc., in the intestinal epithelial cells and liver, although each compound is supposed to produce its own unique structural derivatives. As reviewed elsewhere, these chemicals will undergo gut bacteria-mediated metabolic transformations or conjugation, which may increase absorption, as in the case of anthocyanins [214] and phytoestrogens [215]. In addition, bacterial transformations may yield new metabolites with beneficial effects, as in the case of equol derived from daidzein [215], where equol is known to possess anti-inflammatory and anticancer properties [214]; however, not all microbialderived metabolites exhibit beneficial effects or increased absorption [214,215]. In addition, many phytochemicals, despite their low bioavailabilities, are known to exert their biological activities in the gut independent of tissue absorption. In fact, these phytochemicals are likely to be present in the highest abundance in the gut after oral consumption, mainly due

to very low absorption rates through the gut membrane [208–210]. These phytochemicals exhibit their beneficial effects by altering the rates of transcription of certain genes and the absorption or levels of many essential nutrients or compounds, such as cholesterol, triglycerides, and bile acids [208–210]. Furthermore, they are known to exert their functional activities by their unique antioxidant activities by suppressing the activities of pro-oxidant enzymes and transcription factors, including NF-kB, which transcriptionally regulates many proinflammatory downstream targets such as TNFα and iNOS. These antioxidant compounds also show anti-inflammatory effects with decreased levels of pro-inflammatory cytokines and chemokines against gut-related disorders such as inflammatory bowel disease [216], in addition to combatting other metabolic syndrome-related symptoms (e.g., obesity-related fatty liver and type 2 diabetes associated cardiovascular disorders) [210] and possibly preventing neurodegenerative abnormalities, including Alzheimer's disease, through the gut(–liver)–brain axis [217].

As illustrated in Figure 2, oxidative and nitrosative/nitrative stress, inflammation and leaky gut, endotoxemia and inflammatory tissue injury arise following exposure to many environmental compounds such as ethanol (alcohol drinking), Western-style high-fat diets, and high fructose or genetic risk factors and/or certain disease states. However, certain phytochemicals and/or their metabolites are capable of ameliorating these pathological manifestations caused by some of these insults, particularly through changing (usually normalizing) the amounts and components of the gut microbiota, resulting in improved composition (i.e., gut eubiosis) [202,218–220] from gut dysbiosis. Additionally, these beneficial agents can alter the metabolism and production of many endogenous compounds such as ethanol, acetaldehyde, SCFAs, bile acids, and trimethylamine (TMA) by altering the compositions of the gut microbiome [202,221] or perhaps even ethanol and acetaldehyde, which have been shown to be produced by some gut bacterial strains [202,218–220,222]. These antioxidant phytochemicals and/or their gut metabolites can also suppress the increased oxidative and nitrosative/nitrative stress by inhibiting the enzymes, such as CYP2E1, NOXs, iNOS, and NF-κB [223–225], as listed in the Figure. Furthermore, these beneficial antioxidant compounds from dietary supplements can affect key enzymes/proteins in the cellular signaling pathways, such as AMPK [226] and hepatic Sirt-1 [227], or transcription factors, including NF-κB and PGC-1α [228,229], to show their biological effects, as demonstrated by polyphenols in green and black teas [230]. In addition, these antioxidant phytochemicals may improve the disease states by modifying the genetic and epigenetic regulations in the gut and other tissues possibly through the NAD+-dependent non-histone deacetylase Sirt-1 [227] and its isoforms. Finally, gut dysbiosis-mediated altered levels of SCFAs, especially propionate and butyrate, which are inhibitors of histone deacetylase [19], may result in different epigenetic profiles, which further regulate the gene transcription rates compared to those during eubiosis and/or in phytochemical-treated cases. All these changes are likely to contribute to the beneficial effects of phytochemicals and improvement of various disease states in experimental models, as well as in a few human studies.

**Figure 2.** Proposed mechanisms of the beneficial effects of antioxidant phytochemicals on gut dysbiosis and oxidative stress-mediated intestinal barrier dysfunction and inflammatory liver injury. As described in the text, many phytochemicals contained in various fruits, vegetables and dietary supplements can be metabolized by gut microbiota for improved absorption, leading to greater bioavailability. By improving the oxidative stress and gut dysbiosis, these antioxidant phytochemicals and/or their metabolites prevent leaky gut, endotoxemia, inflammation, and alcoholic and/or non-alcoholic fatty liver disease.

### **6. Translational Approaches and Therapeutics against Alcohol-Mediated Oxidative Stress, Gut Dysbiosis, Intestinal Barrier Dysfunction and Fatty Liver Disease**

The presence of pathogenic microbes and their metabolites in the gut may potentially promote ALD and its progression, since pretreatment with non-absorbable antibiotics such as neomycin and polymyxin B [231] or rifaximin [232] significantly prevented the severity of alcohol-mediated liver disease in different animal models. Yet, although neomycin and polymyxin B helped ameliorate fructose-mediated steatosis development [233], rifaximin only showed limited success in combatting certain manifestations of non-alcoholic steatosis and steatohepatitis (NAFLD/NASH) [234] and thioacetamide-induced liver injury [235]. Furthermore, gut microbiome analyses of control and age-matched cirrhosis patients revealed that altered gut microbiota is also positively associated with cirrhosis and progression [236], suggesting an important role of gut dysbiosis in various liver diseases. The opposite case may also be true, where the absence of all bacteria is also liable to drive liver damage, since one study showed that germ-free mice exposed to binge alcohol conditions displayed greater hepatic fat accumulation and inflammation compared to WT mice, in addition to increased CYP2E1 mRNA in the proximal small intestine, which may trend toward elevated CYP2E1 protein levels and, likely, a stimulation of ROS production and intestinal and liver damage [237]. This germ-free study and reports characterizing gut dysbiosis following alcohol exposure demonstrate that while the gut dysbiosis has the potential to escalate ALD, the complete absence of gut microbes does not necessarily attenuate manifestations of ALD. Thus, maintenance of the normal balance of microbes should help prevent the initiation of the 'second hit' in the two-hit hypothesis of ALD progression, whereby early, mild manifestations of ALD (fatty liver) can progress to more severe insults (inflammation or fibrosis) through a second hit, such as certain gut-derived molecules, such as LPS and reactive oxygen species, as proposed in NAFLD/NASH [238–241]. Indeed, numerous therapeutic studies have sought approaches to moderate oxidative and nitrosative/nitrative stress, gut dysbiosis, and intestinal barrier dysfunction following alcohol exposure. These approaches include supplementation with commensal microbes, gut-protective factors, and a wide range of dietary options from both synthetic and natural origins.

Numerous studies have reported the beneficial effects of commensal microbial (probiotics) supplementation on ALD manifestations and gut dysbiosis [242]. Recent exciting results with clinical studies revealed that fecal microbiota transplantation (FMT) from healthy donor people seems to be relatively safe and effective in attenuating the severity with improvement of health outcome in mouse models of alcoholic liver injury [243] and patients with severe alcoholic hepatitis or cirrhosis [244–246], as reviewed [85]. In addition to whole FMT, supplementation of a specific microbial strain(s) such as lactobacillus [247] can also improve systemic endotoxemia and liver conditions following alcohol consumption. Additional evidence for the beneficial role of microbial strain supplementation includes the decrease in high blood ALT and AST levels following ethanol exposure to rats supplemented with yogurt or cream cheese made with *Lactococcus chungangensis* CAU 28 [248], the decline in hepatic iNOS and global nitration levels following *Lactobacillus fermentum* administration, depending on time point [249], and the *Roseburia intestinalis*-mediated reduction in gut permeability through increased occludin mRNA and protein levels [250]. Additionally, other members of the Lactobacillus genus such as *Lactobacillus plantarum*, have been shown to reduce inflammatory markers, triglyceride levels and gut leakiness associated with ALD by way of epidermal growth factor receptor (EGFR) activity [251]. Inflammation, steatosis, and gut leakiness were also mitigated in a chronic alcohol mouse model supplemented with *Akkermansia muciniphila* gut population [106], which, as previously stated, is found to be decreased in individuals with steatohepatitis [106] and alcohol use disorder [20,102].

Alongside probiotics, certain dietary supplements have been shown to attenuate ALD pathogenesis and limit gut dysbiosis and leakiness. For example, administration of the comestible cricket *Gryllus bimaculatus* prior to acute alcohol exposure was demonstrated to mitigate ethanol-induced increases in intestinal oxidative stress (8-OHdG levels), hepatic apoptotic markers, and hepatic triglyceride accumulation [252]. Pomegranate extracts and indole-3-carbinol (I3C), derived from Brassica vegetables, both mitigated inflammation of the liver and hepatocyte and enterocyte apoptosis in binge alcohol [120] and chronic plus binge alcohol exposure models, respectively [159]; although, unlike pomegranate [120], I3C supplementation did not attenuate fatty liver [159]. Additionally, ellagic acid (EA) and urolithin A (UA), two polyphenols derived from pomegranate extracts, mitigated ethanol-induced increases in gut permeability in T84 colon cells [120]. Furthermore, EA, UA, and I3C were all capable of reducing ethanol-induced increases in hepatic CYP2E1 levels, in vivo (I3C) [120,159] and in vitro (EA, UA) [120], which should reduce hepatic oxidative stress and attenuate liver injury. Interestingly, the PTM landscape is also affected by these supplements, whereby both I3C and pomegranate supplementation decrease ethanolinduced hyperacetylation in the liver and intestines, respectively [120,159]. Pomegranate extract, in particular, reduced ethanol-induced global nitration and ubiquitination of intestinal proteins and, specifically, ethanol-induced claudin-1 nitration and ubiquitination, thus preventing its degradation, and all of these PTM changes correlated with the prevention of intestinal barrier damage, and endotoxemia [120]. Additionally, rats chronically administered alcohol and fed oats exhibited decreases in ethanol-induced protein nitration and oxidation in all small intestine subregions and the colon, which was suggested to be due to a decrease in oxidative stress as a result of oat-mediated decreases in iNOS, nitrite, and nitrate levels [253]. Additionally, β-glucans from various foods, such as oats and barley [224], were effective against gut dysbiosis in non-alcohol-related diseases, thus supporting human

health [225]. When co-administered with ethanol at various percentages, fish oil (high in n-3 polyunsaturated fatty acids) was shown to attenuate liver manifestations of ALD (steatosis and inflammation) [90], plasma endotoxin levels [90], alcohol-induced intestinal permeability dysfunction [254] and even impacted gut microbial composition, through a recovery of fecal Bifidobacterium members [254] and an increase in Bacteroidetes members (especially with supplementation of 25% fish oil), thus decreasing the ratio of Firmicutes to Bacteroidetes following alcohol exposure [90]. However, studies have reported that oxidation of fatty acids in fish oil (prior to administration) actually worsens liver outcomes following alcohol exposure and even increases the abundance of members of the Gramnegative, LPS-producing Proteobacteria phylum [255]. Thus, proper maintenance of fish oil in non-oxidized states appears to be very important to ensure its beneficial effects against alcohol-mediated fatty liver injury [256], possibly through preventing leaky gut [254].

Supplementation of zinc [130] may also attenuate liver and gut dysfunction following alcohol exposure. Zinc levels were observed to decrease in the ileum of mice chronically exposed to alcohol, which correlated with plasma endotoxemia, ileal oxidative stress and permeability, and decreased tight junction protein levels [257]. Mechanistically, by hampering the activity of enzymatic regulators of hepatic apoptosis (e.g., caspase-3), zinc supplementation can prevent hepatocyte death following alcohol exposure [258] and may benefit from, but does not require, the zinc-binding ability of metallothionein to exert its therapeutic effect [259].

Some alternative and complimentary remedies, such as traditional herbal medicines in China [260–262], Korea [263], and India [264,265], have recently proved effective in both combatting inflammation and in altering the gut microbial composition in models of ALD or models of intestinal disease. For example, co-administration of ethanol with the fungi *Wolfporia cocos* (or, more specifically, the water-insoluble polysaccharides from their fruiting bodies) decreased the hepatic triglyceride levels and MCP-1 levels, indicating decreased steatosis and inflammation of the liver [261]. Interestingly, these polysaccharides increased Firmicutes abundance and decreased the abundance of the Gram-negative, LPS-producing Proteobacteria phylum following chronic alcohol exposure [261]. Additionally, leaf extract from the plant *Dendropanax morbifera* leaf extract also altered microbial composition in rats acutely exposed to alcohol, whereby, for example, members of the Bacteroides operational taxonomic unit (OTU) increased upon co-administration of leaf extract with ethanol, compared to the ethanol group [263]. The *Dendropanax morbifera* leaf extract co-administered with ethanol contained several phytochemicals (e.g., rutin, caffeic acid, etc.) [263] and it is highly likely that the list of beneficial antioxidant phytochemicals capable of protecting against alcohol-mediated oxidative stress, gut dysbiosis, epithelial barrier dysfunction, and ALD (as well as NAFLD) will be increased in the future.

Microbiota-derived molecules may also help combat alcohol-induced tissue damage and ALD pathogenesis. Indeed, while microbial products derived from the colon and feces of alcohol-exposed mice were demonstrated to increase intestinal permeability in vitro, in addition to instigating in vitro T-cell activation [266], certain beneficial microbial products may reverse these adverse outcomes. SCFAs, in particular, butyric acid supplemented as tributyrin, may improve intestinal barrier functioning through the preservation of ZO-1 and occludin protein levels and proper localization at the barrier [267]. Additionally, butyrate can restrain cytokine (TNFα and MCP-1) production following acute alcohol exposure, although it does not appear to ameliorate alcohol-induced steatosis [267]. Indirect benefits of SCFA supplementation might also be achieved through administration of *Pediococcus pentosaceus*, which increased the levels of the SCFAs possibly due to the partial recovery of bacterial genera, such as Clostridium, whose levels were decreased in the mouse model of chronic plus binge alcohol exposure and whose recovery could increase SCFA production, as suggested by the positive correlation found between Clostridium and butyric acid levels in this study [268]. Administration of other SCFAs or other beneficial microbial products may yield similar results to butyrate supplementation, such as administration of indole-3-propionic acid (IPA), an indole metabolite generated by the metabolism of

tryptophan, which elevated gut tight junction proteins, thus preventing intestinal barrier dysfunction and liver damage in rats exposed to a Western-style HFD, possibly by the gut–liver axis [269]. However, another study exposing mice to HFD did not find a reduction in liver triglyceride levels with co-administration of IPA (in addition to other parameters of bodily damage or inflammation), which could be attributed to minor differences in the HFD or murine model used [270]. Nevertheless, these differing results highlight the need for continued studies on the role of microbial products in potentially ameliorating liver and tissue damage induced by a wide range of stimuli.

Certain synthetic drugs, such as sennoside A [271] and metformin [272,273], can exhibit their beneficial effects on leaky gut and endotoxemia, in addition to preventing NAFLD and/or learning and memory impairment. However, it is also known that some antibiotics and chemotherapeutic agents such as 5-fluouracil and vancomycin are known to cause dysbiosis of gut microbiota [274]. In the latter cases, a probiotic with digestive enzymes was formulated to protect against cancer-drug-related dysbiosis, suggesting a reminder of careful interpretation of the results in using some drugs to regulate gut dysbiosis. As mentioned in the previous sections, usage of some antibiotics such as neomycin and polymyxin B or rifaximin could be considered for treating alcohol-mediated gut endotoxemia and steatotic and inflammatory liver damage through repurposing their applications, since these beneficial changes were observed in rodent models [231,232]. Additionally, the composition of the gut microbiota was changed in several orders (e.g., Erysipelotrichales) following ethanol and ethanol with rifaximin administration; however, the physiological role of these compositional changes and others should be further studied [232]. These antibiotics may represent possible options for treating gut dysbiosis, but future studies will have to confirm this proposition. However, aside from already-approved antibiotics, the usage of other synthetic compounds may not be practical, since large-scale randomized clinical studies are likely to require considerable time, cost, and effort in evaluating the safety and efficacy tests needed for FDA approval.

The benefits of naturally occurring compounds found in many edible plants, fruits, vegetables, and dietary supplements should also be considered as alternative approaches. For instance, various phytochemicals (berberine, curcumin, resveratrol, and the numerous components of silymarin) from different foods and plants can prevent NAFLD and, in some cases, can normalize gut dysbiosis and intestinal permeability changes, especially in rodent models [262]. The antioxidant *N*-acetylcysteine (NAC) was shown to mitigate LPS-induced intestinal permeability in the IPEC-J2 intestinal porcine enterocyte cell line [275], and experiments using Caco-2 cells showing that NAC supplementation with alcohol can prevent an EtOH-induced increase in CLOCK and PER2 expression helped formulate the hypothesis that ROS generated by CYP2E1 in response to ethanol could increase the expression of these circadian rhythm proteins, leading to intestinal dysfunction and permeability, as recently reviewed [276]. Some of these phytochemicals, such as resveratrol, curcumin, silymarin, genistein, quercetin, rutin, anthocyanidins, ellagic acid, etc., may have very limited bioavailability, although some may be better absorbed than others [277]. In theory, intakes of extremely large amounts of these natural compounds are needed to observe their beneficial effects due to their very low bioavailability, though experiments testing this hypothesis using curcumin or ellagic acid suggest this may is not the case [205,208–210,226,278]. These phytochemical compounds may have paradoxically high functional activity despite the setback of their low bioavailability [205]. For instance, a subset of French people, who usually consume a diet with relatively high saturated fatty acids and cholesterol, exhibit low risks of major diseases, like coronary heart disease, and deaths known as the French paradox [279,280] possibly due to their habits of daily drinking of wine, which contains resveratrol and other antioxidants [281]. Part of the reason could be that these polyphenol compounds may have direct antioxidant effects that counteract potentially oxidative enzymes such as NADPH oxidases [282] and CYP2E1 [223,283]. Alternatively, these chemicals may possibly exhibit their benefits by preserving the

activities and/or levels of certain antioxidizing enzymes and proteins, as suggested by the importance of the presence of SOD2 for the resveratrol-mediated prevention of cytotoxicity in mouse hippocampal neurons pretreated with this polyphenol [284]. Additionally, other antioxidant-related proteins, such as ALDH2 and glutathione peroxidase (Gpx), which were demonstrated to be inactivated in the presence of acetaminophen [285], alcohol or non-alcoholic substances [68,79], may also be restored by the administration of antioxidant phytochemicals. Furthermore, these phytochemicals, including various flavonoid compounds, demonstrate their benefits by preventing gut dysbiosis and normalizing the amount and composition of beneficial bacteria such as *Lactobacillus, Akkermansia,* and *Bifidobacteria*, while decreasing the population of potentially harmful microbes, as suggested elsewhere [219,262,286]. In addition to their known beneficial interaction with the gut microbiota [202], certain phytochemicals may also affect bacterial strains known to produce ethanol and acetaldehyde, both of which are known to cause leaky gut and fatty liver disease [120,121], although this hypothesis has yet to be tested. In general, the usage of these phytochemicals and/or other prebiotics or probiotics as dietary supplements could be important in treating patients with alcoholic steatohepatitis with leaky gut and endotoxemia, especially considering the increasing incidence of ALD and the proposed high mortality rate (30~40%) at 1 month for individuals with severe alcoholic hepatitis [287–289]. Unfortunately, there is no clinically proven drug approved for effectively treating alcoholic hepatitis patients, although a few clinical studies are being conducted [290,291]. Early clinical trials using FMT from healthy donors into patients with alcoholic hepatitis, cirrhosis, AUD, etc., offer preliminary data suggesting that FMT may be a safe and promising therapeutic option [85,244–246], although additional large-scale and longterm multi-center studies are needed to confirm the safety/toxicity profile and efficacy of the newly emerging FMT therapy against various liver diseases, including ALD. Requests have been submitted for several phytochemicals listed in this review, such as trans-resveratrol, urolithin A, and curcumin, to be recognized as generally regarded as safe (GRAS by the FDA definition), and, if given FDA approval, these could be recommended as dietary supplements for ALD patients in addition to their clinical treatment protocols. However, these newly emerging areas need to be further studied to completely understand the benefits of various antioxidants contained in fruits, vegetables and dietary supplements.

### **7. Conclusions**

In this review, we have briefly described the role of gut microbiota alteration during normal development as well as in pathological conditions. We also mentioned the role of gut dysbiosis caused by various environmental factors as well as epigenetic and/or genetic risks. We also describe the patterns of gut microbiome changes in various pathophysiological conditions and provide evidence for the benefits of ameliorating gut dysbiosis by many different agents in human populations and experimental rodents as well as cell culture models. Moreover, we describe the causal role of increased oxidative stress in promoting intestinal barrier dysfunction, subsequently leading to elevated endotoxemia and fatty liver disease. We also described potentially safe methods of translational approaches against alcohol-mediated oxidative stress, gut dysbiosis, leaky gut, and ALD. These beneficial methods and agents include healthy lifestyle changes with proper intake of healthy diets, containing beneficial chemicals contained in many fruits and vegetables, or n-3 polyunsaturated fatty acids. Indeed, these beneficial agents have demonstrated their effectiveness against gut dysbiosis, intestinal barrier dysfunction, endotoxemia, and ALD and NAFLD in many experimental models. In particular, supplementation with certain phytochemicals represents an exciting prospective therapeutic option for mitigating alcohol-mediated tissue damage owing to the breadth of known phytochemicals and the observed benefits of these chemicals and/or their metabolites on the gut microbiome and redox regulation, despite many compounds having low bioavailabilities. Importantly, future certification

of many phytochemicals in various fruits, vegetables, and plants as GRAS could permit usage of these chemicals as dietary supplements in treating ALD. However, randomized large-scale clinical studies need to be conducted in the future to accurately demonstrate the efficacy of some of these antioxidant phytochemicals.

**Author Contributions:** Conceptualization by J.W.B. and B.-J.S.; writing—original by J.W.B. and B.-J.S.; review and revisions by J.W.B. and B.-J.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Intramural Research Fund of the National Institute of Alcohol Abuse and Alcoholism, National Institutes of Health.

**Acknowledgments:** Both authors are grateful to Wiramon Rungratanawanich for her excellent help in designing the Figures and significant editorial assistance.

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

### **References**


## *Article* **Quercetin Ameliorates Insulin Resistance and Restores Gut Microbiome in Mice on High-Fat Diets**

**Yuqing Tan 1,2, Christina C. Tam 3, Matt Rolston 4, Priscila Alves 2, Ling Chen 2,5, Shi Meng 6,7,\*, Hui Hong 1,\*, Sam K. C. Chang <sup>8</sup> and Wallace Yokoyama <sup>2</sup>**


**Abstract:** Quercetin is a flavonoid that has been shown to have health-promoting capacities due to its potent antioxidant activity. However, the effect of chronic intake of quercetin on the gut microbiome and diabetes-related biomarkers remains unclear. Male C57BL/6J mice were fed HF or HF supplemented with 0.05% quercetin (HFQ) for 6 weeks. Diabetes-related biomarkers in blood were determined in mice fed high-fat (HF) diets supplemented with quercetin. Mice fed the HFQ diet gained less body, liver, and adipose weight, while liver lipid and blood glucose levels were also lowered. Diabetes-related plasma biomarkers insulin, leptin, resistin, and glucagon were significantly reduced by quercetin supplementation. In feces, quercetin supplementation significantly increased the relative abundance of *Akkermansia* and decreased the Firmicutes/Bacteroidetes ratio. The expression of genes *Srebf1*, *Ppara*, *Cyp51*, *Scd1*, and *Fasn* was downregulated by quercetin supplementation. These results indicated that diabetes biomarkers are associated with early metabolic changes accompanying obesity, and quercetin may ameliorate insulin resistance.

**Keywords:** quercetin; high-fat diet; insulin resistance; gut microbiome

### **1. Introduction**

Obesity is recognized as a major global public health crisis. In 2016, over 1.9 billion adults globally were overweight and 650 million were obese globally [1]. Obesity, cardiovascular disease, and type II diabetes are considered inflammatory diseases. Antiinflammatory phytochemicals such as phenolics and polyphenolics are extremely potent against metabolic diseases. They are concentrated in leaves, peels, and seeds where they protect the plant against environmental pathogens. Their bioactivity against plant pathogens is broad and may be the basis for their demonstrated beneficial health properties in humans.

Quercetin is one of the most abundant and common flavonoids in plant foods [2]. It is well known as a potent antioxidant and scavenger of reactive oxygen species (ROS) and reactive nitrogen species (RNS). Quercetin has been shown to have beneficial effects in

**Citation:** Tan, Y.; Tam, C.C.; Rolston, M.; Alves, P.; Chen, L.; Meng, S.; Hong, H.; Chang, S.K.C.; Yokoyama, W. Quercetin Ameliorates Insulin Resistance and Restores Gut Microbiome in Mice on High-Fat Diets. *Antioxidants* **2021**, *10*, 1251. https://doi.org/10.3390/antiox10081251

Academic Editor: Baojun Xu

Received: 14 July 2021 Accepted: 3 August 2021 Published: 5 August 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/).

human studies [3]. However, the effect of chronic intake of quercetin on the gut microbiome and diabetes-related biomarkers remains unclear. However, quercetin content in typical meals for humans does not reach the high levels used in some animal studies. Therefore, understanding the effect of a chronic intake of a low concentration (0.05%) of quercetin on the gut microbiome in mice on a high-fat diet is vital. Quercetin is found as a glycoside in foods, and studies suggest that the glycoside must be hydrolyzed to the aglycone for efficient absorption [4]. The pharmacological effects of quercetin may also be partly due to its physicochemical properties. The solubility of quercetin is low, about 2.6 mg/L at 25 ◦C. Despite its low solubility, carbon-14 (14C) studies in humans reported that oral absorption was 36–53%. However, the combined urinary and fecal excretion was less than 10% [5]. The investigators found that as much as 23–81% was excreted as CO2, suggesting that microbial action may contribute to its degradation. A recent study suggested that the metabolism of the natural product asperuloside altered intestinal metabolites levels and composition via modulation of gut microbiota [6].

Quercetin has been shown to be highly effective in preventing obesity-related metabolic syndrome characteristics in animals. For example, quercetin (0.05%) reduced body weight, visceral fat, blood glucose, insulin, and TNF-α in C57BL/6J mice fed a high-fat diet for 20 weeks but not 8 weeks. However, this study did not investigate the gut microbiome [7,8]. In terms of a low-fat diet, no significant effects on body weight, visceral fat, or blood glucose and lipids were observed in C57BL/6J mice fed 0.05% or 1% quercetin in an AIN93G diet (7% fat by weight) for 20 weeks [9]. However, type 2 diabetic mice (db/db mice) fed 0.04% or 0.08% quercetin on a AIN93G diet for 6 weeks had lower fasting glucose but not insulin, as well as lower adiponectin at 0.08% but not 0.04% quercetin [10]. These results suggest that quercetin reduces biomarkers of metabolic dysfunction when mice have excessive body and visceral adipose weight gain via consumption of a high ft diet or in an obesogenic animal model.

Quercetin also changes the distribution of gut bacteria. In overweight humans and animals, the intake of probiotics and dietary fibers changed the patterns of gut microbiota in comparison to those of normal-weight animals or humans [11]. Gut bacteria transform quercetin to metabolites such as homoprocatechuic acid, protocatechuic acid, 4-hydroxybenzoic acid, and propionic acid [12]. These metabolites can be detected in blood and urine to assess bioactivity in human trials [13]. Quercetin (1% of diet) was reported to reshape the fecal microbiota composition of rats fed a high-fat, sucrose diet, but anti-obesity and anti-inflammation effects were not reported [14]. However, 1% quercetin supplementation is hard to achieve in typical meals for humans (usual consumption is 10–100 mg quercetin a day). While many studies have investigated the impact of quercetin intake on physiological effects including changes in gene expression and, levels of lipidand carbohydrate-metabolizing enzymes, few have reported alterations in biomarkers associated with insulin resistance (IR) and appetite. This is the first study to investigate chronic intake and its effect on gut microbiota profiles in diet-induced insulin-resistant mice. We hypothesize that the prevention of insulin resistance and obesity, as well as improvements in the gut microbiome, due to quercetin intake in diet-induced obese mice is accompanied by measurable changes in blood biomarkers for diabetes.

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

#### *2.1. Animal and Diets*

Male C57BL/6J mice (22.3 ± 1.5 g), 5 weeks old, were purchased from Jackson Laboratories (Sacramento, CA, USA). Mice were housed individually in a temperature-controlled room (20–22 ◦C, 60% relative humidity, 12 h alternating light/dark cycle). Mice were acclimated and given access to drinking water and chow diet (LabDiet #5001, PMI International, Redwood, CA, USA) *ad libitum* for 1 week before feeding of the experimental diet. Mice were sorted by weight, and each weight range was randomly divided into three groups with eight mice each and fed, ad libitum, semi-synthetic diets based on an AIN-93G formulation consisting of a high-fat (HF, 53% fat calories) diet supplemented

with 0.05% quercetin (HFQ), an HF control diet, and a low-fat (LF) reference diet (Table 1). Quercetin was dissolved in 5 mL of ethanol dispersed with the dry ingredients, followed by evaporation of ethanol to obtain 1 kg of food. Body weights were recorded once a week, and food intake was monitored twice a week. The study was reviewed and approved by the Institutional Animal Care and Use Committee, Western Regional Research Center, USDA, Albany, CA, USA (Protocol No. 18-4).


**Table 1.** Diet composition (grams).

### *2.2. Plasma and Tissue Collection*

After 6 weeks of feeding, mice were fasted for 16 h and then anesthetized with isoflurane (Phoenix Pharmaceutical, St. Joseph, MO, USA). Blood was collected by cardiac puncture into EDTA-rinsed syringes. Plasma was separated by centrifugation at 2000× *g* for 15 min at 4 ◦C, and the samples were stored at −80 ◦C for further analysis. Liver and epididymal adipose were collected, weighed, and frozen in liquid nitrogen for further analysis. Five livers were used for hepatic lipid content analysis, and the remaining livers (3–4) were used for PCR analysis.

### *2.3. Blood Glucose, Plasma, and Hepatic Lipid Analysis*

Tail-vein blood glucose level was determined using a OneTouch Ultrameter (Life Scan Inc., Milpitas, CA, USA). Plasma lipoprotein cholesterol was determined according to our previous method [15,16]. Plasma triglyceride (TG) was determined using an enzyme colorimetric assay kit (Sekisui Diagnostics PEI Inc., Charlottetown, PE, Canada) according to the manufacturer's instructions, and the absorbance was measured at 505 nm (Nanodrop 2000 C spectrophotometer, Thermo Scientific, Pleasanton, CA, USA). Liver lipids were determined as described previously [15]. Liver tissue from five mice from each group was used for hepatic lipid analysis and the residual liver tissue (*n* = 4) were saved for PCR analysis.

### *2.4. Glucose Tolerance Test (GTT)*

The glucose tolerance test was administered after 6 weeks of diet treatment. After 5 h of fasting, mice were orally administered a 20% glucose solution (10 mL/kg body weight). Tail-vein blood glucose levels were determined at 0, 15, 30, 60, and 120 min using a OneTouch Ultrameter (Lifescan Inc., Milpitas, CA, USA). GTT curves were obtained by plotting glucose concentration versus time, and integrated glucose concentration over 120 min was calculated as the area under the curve (AUC).

### *2.5. Plasma Levels of Metabolic Biomarkers Relevant to Diabetes and Obesity*

Plasma ghrelin, gastric inhibitory polypeptide (GIP), glucagon-like peptide-1 (GLP-1), insulin, leptin, resistin, and glucagon levels were analyzed using a mouse diabetes multiplex antibody assay kit (Bio-Plex Pro Mouse Diabetes Assay, Bio-Rad, Hercules, CA, USA) on the Bio-Plex 200 system (Bio-Rad, Hercules, CA, USA) according to the manufacturer's instructions.

### *2.6. Plasma Levels of Inflammation Cytokines*

Plasma interleukin-2 (IL-2), interleukin-4 (IL-4), interleukin-5 (IL-5), interleukin-10 (IL-10), interleukin-12 (IL-12), granulocyte-macrophage colony-stimulating factor (GM-CSF), interferon gamma (IFN-γ), and tumor necrosis factor alpha (TNF-α) levels were analyzed using a multiplex antibody kit following the manufacturer's instructions (Bio-Plex Pro Mouse Cytokine Assay 8-plex, Bio-Rad) with the Bio-Plex 200 system (Bio-Rad).

#### *2.7. Homeostatic Model Assessment of Insulin Resistance (HOMA-IR) Index Calculation*

HOMA-IR was calculated from fasting blood glucose and plasma insulin concentrations. HOMA-IR index was calculated according to Equation (1) [17]. Insulin and plasma glucose concentrations after 16 h fasting were used to calculate the HOMA-IR index.

$$\text{HOMA} - \text{IR} = \frac{\text{Fastig plasma insulin (mU/L)} \times \text{Fating plasma glucose (mmol/L)}}{22.5} \tag{1}$$

### *2.8. Fecal Microbiome Analysis*

Mice were placed in paper cups fecal pellets were immediately collected and, stored at −80 ◦C. DNA from feces was extracted using Qiagen DNeasy PowerSoil kits (Qiagen, Valencia, CA, USA) following the standard protocol. The V3–V4 domains of the 16S rRNA were amplified using primers 319F/806R (TCGTCGGCAGCGTCAGATGTGTATAAGA-GACAG (spacer) GTACTCCTACGGGAGGCAGCAGT and [GTCTCGTGGGCTCGGA-GATGTGTATAAGAGACAG (spacer) CCGGACTACNVGGGTWTCTAAT, respectively) containing an Illumina tag sequence, a variable-length spacer, a linker sequence, and the 16S target sequence. Each sample was barcoded with an Illumina P5 adapter sequence, a unique eight nucleotide (nt) barcode, and a partial matching sequence of the forward primer, as well as reverse primers with an Illumina P7 adapter sequence, unique 8 nt barcode, and a partial matching sequence of the reverse adapter. The final product was quantified on a Qubit 4.0 instrument using the dsDNA Broad Range DNA kit (Invitrogen, Carlsbad, CA, USA), and individual amplicons were pooled in equal amounts. The pooled library was cleaned with Ampure XP beads (Beckman Coulter, Brea, CA, USA), and bands of interest were further isolated by gel electrophoresis (Sage Science, Beverly, MA, USA). The library was quantified via qPCR then sequenced with 300 bp dual end sequencing with an Illumina MiSeq at the Genome Center DNA Technologies Core, UC Davis. The names of the repository/repositories and the accession numbers can be found at https://www.ncbi.nlm.nih.gov/ (accessed on 4 August 2021), PRJNA722496.

The Raw FASTQ files and adapter trimmings were demultiplexed with dbcAmplicons version 0.8.5 (https://github.com/msettles/dbcAmplicons (accessed on 4 August 2021)). Forward and reverse unmerged reads were imported into QIIME2 version 2020.2 (https: //qiime2.org (accessed on 4 August 2021)), and sequence variants were determined by utilizing the DADA2 analysis pipeline. Singletons and chimeras were removed as part of the quality filtering process, and the remaining sequences were clustered into amplicon sequence variants (ASVs). The clustered sequences were then compared against the Silva 132 reference database which was used for taxonomic assignment, meeting 99% identity.

Shannon's index was calculated and displayed using the R program through rarefactions to indicate alpha-diversity. Beta-diversity was used to evaluate differences by both weighted and unweighted UniFrac methods. Subsequently, a principal coordinate analysis (PCoA) based on Bray–Curtis distance was performed with an iterative algorithm. An online LEfSe analysis was adopted to search for the biomarkers of different

groups (http://huttenhower.sph.harvard.edu/galaxy (accessed on 4 August 2021)). Based on the LEfSe analysis, bacteria with *p*-values < 0.05 in LDA scores of 3.0 was plotted.

### *2.9. RT-PCR*

RNA was extracted from livers and adipose tissues by TRIzol and an RNA purification kit (Invitrogen, Life Technologies, Carlsbad, CA, USA). All primers and probes for ddPCR were designed by Invitrogen (Invitrogen, Life Technologies, Carlsbad, CA, USA) as per MIQE guidelines [18]. cDNA was synthesized using a GeneAmp RNA PCR kit (Applied Biosystems, Foster City, CA, USA). Synthesized cDNA was diluted 10 times with dH2O, and 1 μL of diluted cDNA was used in each real-time RT-PCR using SYBR green supermix (Bio-Rad, Hercules, CA, USA) with an Mx3000P instrument (Agilent, Cedar Creek, TX, USA). Cycle conditions were as follows: 5 min at 95 ◦C and 94 ◦C for 30 s, followed by 60 ◦C for 1 min, and then 72 ◦C for 30 s. Primers were validated by PCR product sizes, and no primer dimers were observed in gel electrophoresis of PCR products. Primer amplification efficiency was over 90% for every RT-PCR assay. Differences in mRNA expression in liver and adipose tissues were calculated after normalization to β-actin or 36B4 mRNA expression using the 2−ΔΔCt method [19]. The genes used in this study were Srebf1 (NCBI gene ID:78968), Cyp7a1 (NCBI gene ID:13122), Ppara (NCBI gene ID:19013), Cyp51 (NCBI gene ID:13121), Scd1 (NCBI gene ID:20249), Fasn (NCBI gene ID:14104), Slc2a4 (NCBI gene ID:20525), Adipoq (NCBI gene ID:11450), 36b4 (NCBI gene ID:11837), and β-actin (NCBI gene ID: 11461). Primers are shown in Supplementary Table S1.

### *2.10. Statistical Analysis*

Results were expressed as the mean ± SEM. The significance of differences between treatments was analyzed by ANOVA, followed by Tukey–Kramer HSD tests, using 2016 SAS (version 9.4, SAS Inc., Cary, NV, USA). The significance level was set at *p* < 0.05.

### **3. Results and Discussion**

### *3.1. Animal Metrics*

Mice fed an HF diet had almost three times higher weight gain compared to mice on the LF diet, confirming HF diet-induced obesity (Figure 1A). Published reports of weight gain of diet-induced obese (DIO) mice fed quercetin have not been consistent. In our study, mice fed the HFQ diet gained 69.7% less weight (*p* < 0.05) and had a 66.7% lower feed efficiency ratio (g gain/calories intake, *p* < 0.05) than mice on the HF diet. Weight gain and feed efficiency (Figure 1B) of mice fed HFQ diet were similar to the LF group (*p* > 0.05). The food intake data are shown in Figure 1C. The total food intake of mice on the HFQ diet is less than the HF group. However, the feed efficiency ratio (g gain/g feed) was decreased significantly with quercetin supplementation. Quercetin supplementation might affect appetite. Porras and coworkers [20] reported a similar lower weight gain (77% of the control) and lower feed intake in C57BL/6J mice fed a 60% fat calorie diet containing 0.05% quercetin for 12 weeks. However, others have reported that quercetin supplementation of the HF diet fed to C57BL/6J mice did not result in differences in body weight. Kuipers and coworkers [21] reported no differences in body weight or feed consumption in the same mice model fed a 45% fat diet supplemented with 0.1% quercetin for 12 weeks. The differences in outcomes may be due to the lower percentage of fat calories, 46% vs. 53%, or the low solubility of quercetin. A previous study of isorhamnetin, an Omethylated quercetin glycoside abundant in onions, suggested that the glucose component in the quercetin glucoside affects its bioavailability [22]. The application of quercetin is also limited due to stability and solubility issues [23]. In this study, quercetin solubility and bioavailability were optimized by dissolving in alcohol to a molecular form before dispersing on the dry ingredients. The fat content of diet might influence the bioavailability and metabolism of quercetin [24].

**Figure 1.** (**A**). Anthropometrics in mice fed HF, HFQ, and LF diets for 6 weeks. HF: high-fat control diet (46% kcal from fat, 16.5% kcal from protein, and 37.5% kcal from carbohydrate); HFQ: 0.05% quercetin in high-fat diet; LF: low-fat control diet (16% kcal from fat, 20% kcal from protein, and 64% kcal from carbohydrate) (**B**). Feed efficiency ratio of mice fed different diets. (**C**). Food intake of mice fed different diets. Values are means ± SEMs, *n* = 8/group. Bars with different letters were significantly different (*p* < 0.05).

Mice on the HFQ diet for 6 weeks had significantly (*p* < 0.05) lower liver and adipose weight, 19.6% and 58.3%, respectively, than those on the HF diet. The fasting blood glucose level of the mice on the HFQ diet was 25.4% lower than that of those on the HF diet. Final body weight, liver weight, adipose weight, and blood glucose levels of mice fed the HFQ and LF diets were not significantly different. This suggests that supplementation of quercetin in the HF diet has health-promoting effects in terms of preventing increases in in body, liver, and epididymal adipose tissue weights associated with mice on the HF diet. Previously, researchers reported that C57BL/6J mice fed a high-fat diet (39.9% energy from fat) supplemented with 0.05% quercetin had reduced body weight, blood glucose, insulin, cholesterol, TNF-α, and other markers of metabolic syndrome after 20 weeks of feeding but not after 8 weeks [7]. Since the present study did not first induce metabolic syndrome in mice followed by treatment with quercetin, this indicated that the supplementation with quercetin delays the development of obesity.

### *3.2. Plasma/Hepatic Lipid Content and Triglyceride (TG) Levels*

Quercetin supplementation of the HF diet lowered low-density lipoprotein (LDL) and plasma TG concentration by 37.6% and 62.9%, respectively, compared with the HF diet (*p* < 0.05) (Figure 2A). However, plasma very-low-density lipoprotein (VLDL) and high-density lipoprotein (HDL) cholesterol concentrations of mice fed the HFQ diet were not different from those on the HF diet (Figure 2A). The hepatic total lipid level of mice fed the HFQ diet was 35.5% lower than that of those on the HF diet (Figure 2B). The reduction in plasma TG and hepatic lipid levels was similar to that seen by Kobori and coworkers [7], who reported 27% lower plasma TG and about 30% lower hepatic lipids in C57BL/6J mice fed a Western diet (39.9% energy from fat) supplemented with 0.05% quercetin after 20 weeks but not at 4 or 8 weeks. The present study indicates that the HFQ diet may be responsible for the lowered total lipid content in liver after 6 weeks.

**Figure 2.** (**A**). Plasma lipoprotein cholesterol concentration and triglyceride level in mice fed a high-fat (HF) diet, high-fat diet containing 0.05% quercetin (HFQ), and low-fat (LF) diet for 6 weeks. VLDL: very-low-density lipoprotein; LDL: low-density lipoprotein; HDL: high-density lipoprotein; TC: total cholesterol; TG: triglyceride. Data are expressed as means ± SEMs, *n* = 8/group. Bars with different letters within the same plasma lipoprotein were significantly different (*p* < 0.05). (**B**). Hepatic total lipid content in mice fed with high-fat (HF) diet, high-fat diet containing 0.05% quercetin (HFQ), and low-fat (LF) diet for 6 weeks. Data are expressed as means ± SEMs, *n* = 5/group. Bars with different letters were significantly different (*p* < 0.05).

### *3.3. Glucose Tolerance Test (GTT) and Insulin Resistance*

The oral glucose tolerance test (GTT) curve, the area under the curve (AUC), and the HOMA-IR index are shown in Figure 3A–C, respectively. Compared to the HF diet, HFQ significantly lowered the fasting blood glucose level (*p* < 0.05, Figure 1A). The blood glucose levels of the GTT for the HFQ-fed mice were lower at 15, 30, and 60 minutes but were not significantly different from HF. The HFQ diet significantly improved glucose metabolism (*p* < 0.05), lowered AUC by 10.9% compared to the HF diet. The HOMA-IR index of mice fed the HF diet was 82.3% higher than that of mice on the HFQ diet, suggesting that quercetin supplementation contributed to improved HF diet-induced IR. In the Figure 3C, HOMA-IR index of LF groups was 0, but not included in Figure 3C. In B6 mice fed an HF diet containing 0.4% quercetin for 26 weeks, it was reported that fasting blood glucose levels was lower but not significant [25]. Vessal and coworkers [26] reported that the glucose tolerance of STZ-induced diabetic rats returned to normal levels after intraperitoneal (ip) administration of quercetin for 10 days due to the regeneration of pancreatic islets. This experiment suggests that quercetin itself is bioactive as opposed to metabolites of gut bacterial metabolism. However, quercetin or its glycosides may be excreted through the bile, recirculated into the intestinal lumen, and made accessible to gut bacteria for metabolism into phenolic acids [27]. Kobori and coworkers [7] suggested that, in mice fed high-fat Western type diets, quercetin enables the recovery of cell functions in the liver and pancreas by reducing oxidative stress.

**Figure 3.** (**A**). Glucose tolerance in mice fed a high-fat (HF) diet, high-fat diet containing 0.05% quercetin (HFQ), and low-fat (LF) diet for 6 weeks. (**B**). Area under glucose tolerance test (GTT) curve values. (**C**). Homeostatic model assessment of insulin resistance (HOMA-IR) index of HF and HFQ diet-fed mice; the HOMA-IR index for the LF group was 0. Data are expressed as means ± SEMs, *n* = 8/group. Bars with different letters were significantly different (*p* < 0.05).

### *3.4. Plasma Biomarkers of Diabetes and Obesity*

Plasma ghrelin, GIP, GLP-1, insulin, leptin, resistin, and glucagon levels were analyzed using a multiplex immunoassay method (Figure 4). Mice fed the HFQ diet had a 34.5% higher ghrelin concentrations and 88%, 92%, 27%, and 97% lower insulin, leptin, resistin, and glucagon levels, respectively, than those on the HF diet. Plasma GIP and GLP-1 levels of HFQ diet fed mice were not significantly different from those on the HF (*p* > 0.05).

**Figure 4.** Boxplot of plasma biomarker concentrations related to diabetes and obesity, *n* = 8/group. ND = not detected. Top edge of the box, 75th percentile; bottom edge, 25th percentile; horizontal bar within box, median; top horizontal bar outside box, maximum concentration; bottom horizontal bar outside box, minimum concentration. Boxes with different letters were significantly different (*p* < 0.05). ND represents not detected.

Ghrelin is an orexigenic hormone secreted mainly by the stomach preprandially and is responsible for regulating appetite and energy hemostasis. Ghrelin stimulates appetite; thus, our finding that the ghrelin level in mice fed the HFQ diet was higher but the feed intake and feed efficiency ratio were lower than in the HF diet-fed mice was unexpected. Ghrelin levels are higher before meals and lower between meals. In the current study, mice were sacrificed after a 16 h fast; therefore, high ghrelin levels in mice were expected. However, Moesgaard and coworkers [28] reported that the feeding status (fasting or nonfasting) does not affect ghrelin levels, and the reduced expression of ghrelin in obese C57BL/6J mice was due to decreasing numbers of ghrelin-producing cells. After gastric bypass, DIO mice had higher levels of ghrelin than mice without bypass, suggesting that lower feed intake may be associated with ghrelin levels [29]. In this study, ghrelin levels in mice fed LF diet tended to be higher, supporting the hypothesis that a HF diet reduces ghrelin-producing cells. Previous studies have not reported the effect of quercetin feeding with HF diets on the secretion of ghrelin.

Plasma insulin was lower in mice fed a HFQ diet compared to HF diet-fed mice. A high fasting plasma insulin level is an indicator of IR, and IR often precedes T2D. Plasma insulin and blood glucose levels of mice fed the HF diet were 83.3% and 25.4% higher (*p* < 0.05), respectively, than those on the 0.05% HFQ diet after 6 weeks. Kobori et al. [7] reported higher plasma insulin concentrations of C57BL/6J mice fed a Western diet (39.9% energy from fat) compared to the Western diet containing 0.05% quercetin after a 20 week feeding study but not at 8 weeks. However, the same researchers reported that, in STZ-induced diabetic BALB/c mice, low-fat (AIN93) diets containing 0.5% quercetin restored plasma insulin compared to STZ diabetic controls (*p* < 0.05) but did not reach the levels of the normal nondiabetic mice [30]. These results indicated that quercetin supplementation lowers insulin secretion compared to the HF diet.

Leptin is an adipokine that regulates energy balance by inhibiting hunger. In humans, obesity is also associated with higher serum leptin levels, indicating leptin resistance. In this study, serum leptin levels of mice fed the HFQ and LF diets were significantly (*p* < 0.05) lower compared to HF, suggesting normal leptin sensitivity (Figure 4). Few studies have reported the effect of quercetin intake on plasma leptin level. Hoek-van den Hil and coworkers [31] reported a lower body weight gain in C57bl/6JOlaHsd mice fed a HF diet containing 0.33% quercetin for 12 weeks. They reported that serum leptin and its adipose gene expression were lower compared to HF controls. In rats, Wein and coworkers [32] reported no differences in leptin levels between HF or 0.03% HFQ diets for 4 weeks, possibly due to the lower level of quercetin and the shorter feeding period.

Resistin is a proinflammatory adipokine related to insulin resistance and obesity. Zhang and coworkers [33] reported that quercetin (75 mg/kg/day) decreased serum resistin in a rat model of nonalcoholic fatty liver disease induced by a HF diet for 8 weeks. Studies of quercetin on plasma resistin are scarce, however, related flavonoids from herbs have shown effects on resistin concentration. The infusion of quercetin and *Ruta graveolens*, a traditional medicinal plant native to Europe, inhibited resistin expression in adipose tissue through downregulation of the resistin-encoding gene [34].

Glucagon is a hormone produced by the α cells of the pancreas that promotes glucose synthesis and fatty-acid oxidation in the liver. Glucagon's action opposes that of insulin, and higher levels of glucagon in humans have been associated with insulin resistance [35]. The lower level of plasma glucagon in HFQ fed mice in this study supports the HOMA-IR index for increased insulin sensitivity [36]. To the best of our knowledge, the present study is the first to report the inhibitory effect of a HFQ diet on plasma resistin and improvement of HOMA-IR by reducing glucagon secretion.

### *3.5. Plasma Levels of Inflammatory Cytokines*

There were no differences in all plasma cytokines, IL-2, IL-4, IL-5, IL-10, IL-12, GM-CSF, IFN-g, and TNF-α, between HFQ and HF or LF treatments analyzed using the multiplex immunoassay, as shown in the Supplementary Figure S1. However, despite this lack of

difference in cytokine levels, several studies have found that quercetin decreases TNF-α. C57BL/6 mice fed quercetin at the same concentration (0.05% of diet) as this study but in a diet with higher fat content (60% energy from fat) for 9 weeks had a lower TNF-α level [37]. Kobori and coworkers [7] reported a decrease in TNF-α at 20 weeks but not at 8 weeks in C57BL/6J mice fed a diet with a similar fat content to the present study. The lack of a difference between treatments in our study may be due to a need for longer high-fat feeding time required to develop inflammation and the fat level of the diet.

### *3.6. Fecal Microbiota Analysis*

Quercetin supplementation of the HF diet resulted in beneficial changes in the microbiome compared with HF. Quercetin lowered the relative abundance of the phyla Actinobacteria and Firmicutes (87.4% and 14.9%, respectively) but increased the abundance of Bacteroidetes (49.3%) (Figure 5A). The ratio of Firmicutes/Bacteroidetes (F/B) (Figure 5B) was significantly reduced by 65.6% by the HFQ diet (*p* < 0.05), and it was not different from LF diet. The ratio of Firmicutes to Bacteroidetes, F/B, is often cited as a marker of microbiota-associated obesity [38]. Firmicutes and Bacteroidetes account for 90% of the phyla in humans, and other phyla such as Actinobacteria are present at about 5–6%. Porras and coworkers [22] found that an HF diet induced nonalcoholic fatty liver disease in mice, which had an increased relative abundance of Bacteroidetes. However, Firmicutes were not significantly affected by 0.05% quercetin supplementation in the HF diet (*p* > 0.05). Turnbaugh and coworkers [39] claimed that the gut microbiota in ob/ob mice were more efficient at releasing calories from the diet than their lean siblings. The present study is the first to report that mice fed HF diets supplemented with quercetin significantly lowered the relative abundance of Firmicutes and increased the relative abundance of Bacteroidetes (*p* < 0.05). The alpha-diversity (Figure 5C) of the gut microbiota in the HFQ group was similar to that of the HF and LF groups, suggesting that quercetin supplementation did not affect the diversity of the gut microbiota. The results of the beta-diversity analysis illustrated that the taxonomic composition was distinctly different between the HF and LF groups (Figure 5D). To further characterize the changes to the gut microbiota, an LEfSe analysis was used to gain insight into the differences among the groups (Figure 5E,F). One significantly different class and one family were identified; Bacilli was higher in the LF group, and Peptostreptococcaceae was higher in the HFQ group. According to the LEfSe analysis, these abundant taxa could be considered as potential biomarkers (LDA score > 3.0, *p* < 0.05).

**Figure 5.** *Cont*.

**Figure 5.** (**A**) Effect of 0.05% quercetin on the relative abundance of the four most abundant bacterial phyla (*n* = 5) in LF (low−fat), HF (high−fat), and HFQ (high−fat with 0.05% quercetin) diets. (**B**) Firmicutes/Bacteroidetes ratio of HF, HFQ, and LF diets for 6 weeks. (**C**). The alpha−diversity was assessed by calculating the Shannon index. (**D**) The beta−diversity was calculated by principal coordinate analysis (PCoA) for the visualization of pairwise community dissimilarity (Bray−Curtis index) of the microbial community. (**E**) Results of linear discriminative analysis (LDA). (**F**) Effect size (LefSe) analysis among three groups. Cardiogram showing differentially abundant taxonomic clades with an LDA score of 3.0 among groups with a *p*−value of 0.05; *n* = 5/group. An asterisk (\*) indicates a significant difference.

At the family level, the relative abundance of Akkermansiaceae, Bacteroidaceae, Eggerthellaceae, and Peptostreptococcaceae in the HFQ diet-fed mice was greater than that of HF diet alone. The HFQ diet significantly reduced the relative abundance of Atopobiaceae and Erysipelotrichaceae compared to the HF diet (*p* < 0.05). Erysipelotrichaceae

has associated with diet-induced obesity [14]. Rabot [40] reported that diet-induced obese mice had higher levels of the Erysipelotrichaceae family. The Pearson correlation analysis between leptin and Erysipelotrichaceae suggested a strong correlation *r* = 0.99, *p* < 0.02. In contrast to Erysipelotrichaceae, the relative abundance of the Bacteroidaceae family in HFQ diet-fed mice was higher in the present study and supports the work of Etxeberria et al. who reported an increasing relative abundance of the Bacteroidaceae family induced by the HFQ diet, although the changes were not statistically significant (*p* > 0.05) [14]. The enrichment of the Bacteroidaceae family was found to be negatively associated with HF diets in mice. Thus, the result from the current investigation suggested that the HFQ diet enrichment of the Bacteroidaceae family may have contributed to a gut microbiome with a higher level of Bacteroidaceae family. Tan et al. (2018) reported that quercetin fed for 12 weeks had a similar effect on body weight, metabolic features, and the gut microbiome, particularly when given with a soluble fiber [41]. It was reported that quercetin was metabolized in the gut after administration, and its methylated metabolite isorhamnetin was the dominant form in the serum; however, the metabolites of quercetin in the gut remained unclear [42]. Moreover, the authors also evaluated the short-chain fatty acids (SCFAs) in the fecal samples, and results suggested that quercetin-treated mice exhibited significantly (*p* < 0.05) higher levels of butyrate in their feces compared with the antibiotic-treated mice [42]. Moreover, quercetin supplementation did not seem to modify the intestinal barrier permeability-associated markers TJP-1, TJP-2, and Ocln gene [12]. Jin et al. reported that quercetin increased the expression of ZO-1 in rats treated with quercetin [43]. These two studies indicated that quercetin might affect the tight junction protein level of ZO-1but not Ocln. Weight gain was reported to be accompanied by the release of inflammatory cytokines from adipose tissues in response to lipopolysaccharides (LPS), cell-wall fragments of Gram-negative bacteria that pass from the intestinal lumen through the intestinal wall [44]. While Gram-negative bacteria are the sources of LPS and initiator of inflammation, other commensal and probiotic bacteria were shown to prevent or reduce the severity of diabetes and other metabolic diseases [45]. Correlation between glucagon and Peptostreptococcaceae was strong (*r* = 0.99, *p* < 0.04). The present study is the first to report that the relative abundance of Atopobiaceae, Eggerthellaceae, and Peptostreptococcaceae families was significantly affected by the HFQ diet compared to the HF fed mice.

At the genus level, the relative abundances of *Akkermansia*, *Bacteroides*, *Marvinbryantia*, and *Romboutsia* genera were significantly increased (64.2%, 67.9%, 65.1%, and 75.2%, respectively) by the HFQ diet compared to the HF alone. There have been few reports of the presence of *Romboutsia* in feces, but a recent report suggested that this genus was indicative of a healthy status of patients [46]. The relative abundance of *Blautia*, *Clostridium sensu stricto 1*, *Erysipelatoclostridium*, *Lactobacillus*, and *Turicibacter* genera was lowered by the HFQ diet compared to the HF diet, but this did not reach statistical significance. *Akkermansia muciniphila* is one of the most abundant species in the intestine, which has been found to be lower in obesity and lowered in individuals treated with metformin [47]. *Akkermansia* enrichment was inversely correlated with obesity [14]. We observed an increase by 64.2% in the relative abundance of *Akkermansia* in the feces of HFQ-fed mice, as well as a significant increase by 68.0% in the relative abundances of *Bacteroides*, *Marvinbryantia*, and *Romboutsia* genera in feces of the HFQ-fed mice compared to the HF diet. More propionate was produced by *Bacteroides*, and the lipid synthesis from acetate was inhibited by propionate [48]. The increased abundance of *Bacteroides* may contribute to weight loss via propionate inhibition of lipid synthesis. In addition, a significantly greater relative abundance of *Marvinbryantia* genus was observed in the feces of HFQ-fed mice. The relative abundance of *Marvinbryantia* was positively correlated with body weight [49]. However, the mechanism is unknown. The current results show that a HFQ diet reshapes the gut microbiome of mice in the HF diet group withtaxa associated with a lean phenotype and less metabolic dysfunction.

### *3.7. RT-PCR Analysis*

Plasma TG, LDL and total hepatic lipid contents, fasting glucose, GTT, and HOMA-IR were significantly lower in mice fed HFQ diets compared to HF diet-fed mice. We, therefore, compared the expression of selected genes for fat and glucose metabolic pathways in liver and adipose tissue (Figure 6A,B, respectively). The mRNA levels of hepatic genes for major enzymes for cholesterol and bile acid synthesis, *Cyp51* (cytochrome P450, family 51, encoding lanosterol 14α-demethylase) was 0.35-fold lower than the HF diet, but the expression of *Cyp7a1* (cytochrome P450, family 7, subfamily a, polypeptide 1, encoding cholesterol 7-alpha-monooxygenase) was not changed (0.93 compared to HF). Lanosterol demethylase is considered to be the first committed step for the synthesis of sterols from lanosterol. *Cyp7a1* encodes the gene for cholesterol 7-alpha hydroxylase, the enzyme responsible for the rate-limiting step of bile acid synthesis. These results suggest that the observed lower LDL cholesterol was due to reduced hepatic cholesterol synthesis rather than excretion of bile acids. Fatty-acid metabolism-related genes, *Scd1* (stearoylcoenzyme A desaturase 1) and *Srebf1* (sterol regulatory element binding transcription factor 1), were reduced by 0.52 and 0.35-fold, respectively. *Srebf1* is a gene regulated by insulin and codes for transcription factor that increase glycolysis and lipogenesis. Its reduced expression relative to HF suggests that the effect may have been due to lower plasma insulin. *Scd1* catalyzes the synthesis of monounsaturated fatty acids that serve as substrates for the synthesis and storage of triglycerides. The relative expression of *Ppara* (peroxisome proliferator activated receptor alpha), a nuclear transcription factor regulating hepatic fat metabolism, was 0.75 of the control. The lower expression of *Pparα* and *Scd1* suggests a lower uptake and oxidation of fatty acids and may be related to the lower liver TG levels (Figure 2B).

In adipose, fatty-acid synthase (Fasn) is a multienzyme protein that metabolizes acetyland malonyl-CoA derived from glucose into fatty acids. The downregulation of the *Fasn* gene may be related to the lower liver lipid content (Figures 2B and 6B). *Slc2a4* (solute carrier family 2, facilitated glucose transporter member 4, also known as *Glut4*) is an insulin-regulated glucose transporter, and downregulation of Slc2a4 in adipose tissue indicated less glucose intake and, therefore, less fat storage in adipocytes in mice fed an LF diet. However, overexpression of Slc2a4 increases insulin sensitivity and glucose tolerance in obese mice [50]. Slc2a4 was significantly increased (Figure 6B) and supports the lower HOMA-IR and improved insulin sensitivity by the HFQ diet compared to the HF diet alone. Adiponectin is a protein hormone coded by the *Adipoq* gene by adipocytes which regulates glucose level and fatty-acid oxidation. *Adipoq* expression was almost three times higher than HF control (Figure 6B), confirming the plasma multiplex immunoassay results. Higher adiponectin levels are also associated with improved insulin sensitivity, as indicated by lower HOMA-IR. Decreased *Adipoq* is related to the development of IR [51]. The proposed mechanism responsible for ameliorating insulin resistance in mice fed an HF diet supplemented with quercetin is shown in Figure 6C. The increased level of *Akkermansia* was related to reduced insulin resistance [52]. The *Lactobacillus* level was reported to be positively associated with weight loss [53]. Gut microbiome changes are associated with insulin resistance; however, the mechanism behind the association remains unclear.

**Figure 6.** (**A**) Relative hepatic gene expression of *Srebf1*, *Cyp7a1*, *Ppara*, *Cyp51*, and *Scd1* in mice fed an HFQ diet compared to the HF diet (red dotted line), with the LF diet group shown as a reference. (**B**) Relative expression of *Fasn*, *Slc2a4*, and *Adipoq* genes in epididymal adipose tissue of mice fed an HFQ diet compared to the HF diet, with the LF diet group shown as a reference. Data are expressed as the mean ± SE, *n* = 4/group. Differences in mRNA expression for livers and adipose tissue were calculated after normalizing to 35B4 mRNA expression. An asterisk (\*) indicates a significant difference (*p* < 0.05) compared to the HF diet. The dotted line (x = 1) represents the expression of control gene. (**C**) Proposed mechanism responsible for ameliorating insulin resistance in mice fed an HF diet supplemented with quercetin, involving the glucose signaling pathway, and ChREBP (dashes outline), as extrapolated from the literature [54].

### **4. Summary**

The present study demonstrated that chronic quercetin supplementation for 6 weeks significantly lowered plasma triglyceride level, decreased body weight gain, reduced liver fat accumulation, ameliorated insulin resistance, and decreased F/B ratio (associated with lean phenotype) of HF diet-induced obese mice. Plasma ghrelin, leptin, resistin, and glucagon changes were associated with improvements in metabolic health. The HFQ diet attenuated the increase in Firmicutes/Bacteroidetes ratio and modulated gut bacteria composition at both family and genus levels. Furthermore, chronic supplementation of quercetin resulted in a significant alternation in genera related to obesity (*Bacteroides* and *Akkermansia*). The present study suggested that chronic quercetin supplementation prevented the increase of many biomarkers of obesity-related metabolic dysfunction, and microbiota associated with lean phenotypes and healthy metabolic profiles were restored by quercetin supplementation.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/antiox10081251/s1, Figure S1: Concentrations of plasma inflammatory cytokines, *n* = 8/group. Top edge of the box, 75th percentile; bottom edge, 25th percentile; horizontal bar within box, median; top horizontal bar outside box, maximum concentration; bottom horizontal bar outside box, minimum concentration. Boxes with different letters were significantly different (*p* < 0.05)., Table S1: gene primers used in the PCR assay.

**Author Contributions:** Conceptualization, Y.T., W.Y. and H.H.; methodology, Y.T. and W.Y.; software, Y.T., H.H. and M.R.; validation, C.C.T., P.A. and L.C.; formal analysis, Y.T.; investigation, Y.T. and C.C.T.; resources, W.Y.; writing—original draft preparation, Y.T.; writing—review and editing, Y.T., W.Y., S.K.C.C. and S.M.; visualization, Y.T. and H.H.; supervision, W.Y., S.M. and H.H.; project administration, W.Y. All authors have read and agreed to the published version of the manuscript.

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

**Institutional Review Board Statement:** This study was approved (13 August 2018) by the Institutional Animal Care and Use Committee, Western Regional Research Center, USDA, Albany, CA, USA (protocol No. 18-4).

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data are contained within the article.

**Acknowledgments:** The authors thank Jackie Miller and Ezekial Martinez for their diligent attention to animal care, and James Pan for helping with liver fat extraction. My daughter Zhiyan Meng (juanjuan) accompanied me revising, editing and proofread the manuscript.

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

### **References**


## *Article Bacillus amyloliquefaciens SC06* **Induced AKT–FOXO Signaling Pathway-Mediated Autophagy to Alleviate Oxidative Stress in IPEC-J2 Cells**

**Li Tang 1,2, Zihan Zeng 1,2, Yuanhao Zhou 1,2, Baikui Wang 1,2, Peng Zou 1,2, Qi Wang 1,2, Jiafu Ying 1,2, Fei Wang 1,2, Xiang Li 1,2, Shujie Xu 1,2, Pengwei Zhao 3,\* and Weifen Li 1,2,\***


**Abstract:** Autophagy is a conserved proteolytic mechanism, which degrades and recycles damaged organs and proteins in cells to resist external stress. Probiotics could induce autophagy; however, its underlying molecular mechanisms remain elusive. Our previous study has found that *BaSC06* could alleviate oxidative stress by inducing autophagy in rats. This research aimed to verify whether *Bacillus amyloliquefaciens SC06* can induce autophagy to alleviate oxidative stress in IPEC-J2 cells, as well as explore its mechanisms. IPEC-J2 cells were first pretreated with 108 CFU/mL *BaSC06*, and then were induced to oxidative stress by the optimal dose of diquat. The results showed that *BaSC06* significantly triggered autophagy, indicated by the up-regulation of LC3 and Beclin1 along with downregulation of p62 in IPEC-J2 cells. Further analysis revealed that *BaSC06* inhibited the AKT–FOXO signaling pathway by inhibiting the expression of p-AKT and p-FOXO and inducing the expression of SIRT1, resulting in increasing the transcriptional activity of FOXO3 and gene expression of the ATG5–ATG12 complex to induce autophagy, which alleviated oxidative stress and apoptosis. Taken together, *BaSC06* can induce AKT–FOXO-mediated autophagy to alleviate oxidative stress-induced apoptosis and cell damage, thus providing novel theoretical support for probiotics in the prevention and treatment of oxidative damage.

**Keywords:** *Bacillus amyloliquefaciens SC06*; IPEC-J2; oxidative stress; autophagy; apoptosis; AKT–FOXO

### **1. Introduction**

The gastrointestinal tract is an important primary digestive organ, whilst intestinal health directly affects the health and growth of animals. While exposed to exogenous stimulators, the gastrointestinal tract is susceptible to oxidative stress [1,2]. The mucosal barrier, composed of intestinal epithelium, the mucus layer, and cells involved in local immune responses [3], plays a key role in maintaining intestinal homeostasis. The epithelium is composed of a single layer of columnar intraepithelial cells. IPEC-J2 cells, derived from the columnar epithelial cells of a piglet's jejunum, were first isolated from the middle jejunum of neonatal piglets by A.J. Brosnahan et al. at the University of North Carolina [4]. IPEC-J2 cells were initially used as a porcine small intestine model to study the pathogenic bacteria that induce porcine hyperplastic bowel disease [5], and then gradually applied to various studies, including oxidative stress [6], intestinal microorganism [7], and intestinal immune response [8]. IPEC-J2 cells are highly similar to normal intestinal epithelial cells due to their

**Citation:** Tang, L.; Zeng, Z.; Zhou, Y.; Wang, B.; Zou, P.; Wang, Q.; Ying, J.; Wang, F.; Li, X.; Xu, S.; et al. *Bacillus amyloliquefaciens SC06* Induced AKT–FOXO Signaling Pathway-Mediated Autophagy to Alleviate Oxidative Stress in IPEC-J2 Cells. *Antioxidants* **2021**, *10*, 1545. https://doi.org/10.3390/antiox10101545

Academic Editors: Baojun Xu and Stanley Omaye

Received: 19 July 2021 Accepted: 25 September 2021 Published: 28 September 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/).

tumor-free genes, and are derived from piglets. The high similarity between IPEC-J2 cells and the real pig and human small intestine has gradually attracted researchers' attention in recent years.

Autophagy is an evolutionary conserved cell procedure that participates in the lysosomal degradation of proteins and organelles, contributing significantly to cell growth maintenance, differentiation, and homeostasis in some adverse conditions, such as hypoxia and starvation [9]. It plays an integral role in maintaining cellular homeostasis and promoting cell survival, while apoptosis exerts a defensive role, selectively removing cells and renewing cells and tissues [10,11]. Therefore, an equilibrium of autophagy and apoptosis in intestinal epithelial cells affects intestinal health. During the process of autophagy, portions of the cytoplasm or the entire organelle are sequestered in double-membranous vesicles, called autophagy vacuoles (AV) or autophagosomes. The autophagosome then fuses with the lysosome to form a monofilm autophagosome that degrades its contents [12].

In vivo and in vitro models, increasing evidence reveals that probiotics have the potential to alleviate oxidative stress in livestock and prevent them from oxidative stress. Some probiotics, such as *Lactobacillus and Bifidobacterium*, exhibit a good therapeutic effect on alleviating intestinal oxidative stress [13–15]. Probiotic *Bacillus* spp. are widely used to improve animal growth performance and prevent gastrointestinal disorders [16,17] and significantly alleviated intestinal oxidative stress in aquatic products and piglets [18–20]. In addition, *Bacillus* spp. could prevent oxidative stress and LPS-induced inflammatory responses in Raw 264.7 macrophages [21]. Furthermore, in chickens, *Bacillus subtilis* increased their antioxidant capacity and oxidative stability [22]. Studies have elucidated that probiotic can protect intestinal epithelial cells from damage and necrotizing apoptosis by regulating the autophagy and apoptosis signaling pathways, recruiting immune cells, and anti-inflammatory factors [23,24]. Some recent studies have shown that probiotics can regulate autophagy to alleviate oxidative stress. For example, metronidazole and *L. reuteri* combination treatment could decrease oxidative stress and inflammatory and autophagic pathways to prevent NAFLD progression [25]. *Lactobacillus reuteri ZJ617* and *Lactobacillus rhamnosus GG* supplementation suppressed lipopolysaccharide-induced oxidative stress by attenuating apoptosis and autophagy via the mTOR signaling pathway [26].

Currently, studies on the effect of probiotics on oxidative stress by regulating autophagy are still rare. In our previous experiments involving rats, it was uncovered that *Bacillus amyloliquefaciens SC06* (*BaSC06*) alleviated oxidative stress through autophagy via the p38 signaling pathway, but the other specific signaling pathway was inadequately investigated [27]. This study therefore sought to verify whether *BaSC06* can induce autophagy to alleviate oxidative stress in IPEC-J2 cells, as well as explore the other related signaling pathways.

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

### *2.1. BaSC06 Bacterial Strain Preparation*

For this study, the probiotic *BaSC06* (CCTCC No: M2012280) was isolated from soil by the Laboratory of Microbiology, Institute of Feed Sciences, Zhejiang University, and preserved at the China Center for Type Culture Collection Afterward, the *BaSC06* strains were cultured at 37 ◦C in Luria-Bertani (LB) broth overnight, and then gathered by centrifugation (8000 rpm for 5 min). After that, the *BaSC06* strains were washed twice with PBS (pH = 7.4) and suspended at 10<sup>8</sup> CFU/mL in the cell culture media. The fresh bacteria suspensions were prepared for cell incubation.

### *2.2. IPEC-J2 Cell Culture*

IPEC-J2 cells were provided by Northwest Sci-Tech University of Agriculture and Forestry, which was then incubated at 37 ◦C in a humidified 5% CO2 with DMEM/F12 (HyClone, Logan, UT, USA) media, containing 10% FBS (Gibco, Grand Island, NE, USA) and 1% antibiotics (100 mg/mL of streptomycin and 100 U/mL of penicillin G).

### *2.3. Establishing Oxidative Stress Model in IPEC-J2 Cells*

The diquat (DQ)-induced oxidative stress model was evaluated utilizing an MTT cell assay kit for cell proliferation and cytotoxicity (Nanjing Jiancheng Bioengineering Institute, Nanjing, China). According to instructions, 10<sup>4</sup> cells per well were seeded in 96-well plates and cultured for 12 h, followed by DQ treatment at various concentrations (0, 250, 500, 750, 1000, and 1250 μmol/L) for 6 h, with nine parallel holes in each group. Thereafter, to each well was added 50 μL of MTT assay solution, and then incubated for 4 h. Afterward, a Spectra Max M5 microplate reader (Sunnyvale, CA, USA) was used to determine the absorbance of the plate at 570 nm. In order to set up the oxidative stress model for IPEC-J2 cells, the optimal DQ concentration was selected according to the IC50, calculated using a probability unit based on the MTT assay. The optimal concentration of *BaSC06* was determined using a Cell Counting Kit-8 (CCK-8 kit, Nanjing Jiancheng Bioengineering Institute, Nanjing, China) as per the instructions of the manufacturer. IPEC-J2 cells were treated with *BaSC06* at various concentrations (0, 105, 106, 107, 108, and 109 CFU/mL) using the same seeding method as MTT, with nine parallel holes in each group, for 6 h. A total of 10 μL of CCK-8 solutions were added to every well, and the plate incubated for 1 h. The optimal *BaSC06* concentration was selected and calculated according to the viability of the cells based on the CCK-8 assay. The Spectra Max M5 microplate reader (Sunnyvale, CA, USA) was used to determine the absorbance of the plate at 450 nm.

IPEC-J2 cells were further divided into four groups: CK (PBS treatment only), DQ (diquat (DQ) treatment only), Ba (*BaSC06* treatment only), and Ba+DQ (*BaSC06* combined with diquat treatment) groups. Notably, IPEC-J2 cells in the CK group were treated with PBS, while those in the DQ and Ba groups were treated with 950 μmol/mL diquat and 108 CFU/mL *BaSC06*, respectively. In the Ba+DQ group, IPEC-J2 cells was pretreated with 108 CFU/mL *BaSC06* for 6 h and then treated with 950 μmol/mL diquat for the same period. Before *BaSC06* or DQ treatment, each group was washed twice with PBS at the same time.

Furthermore, 25 μM AKT phosphorylation inhibitor Perifosine (Biotime Biotechnology) was used to pre-incubate with IPEC-J2 cells for 48 h according to the instruction, and we then repeated the former *BaSC06* and DQ treatment methods.

#### *2.4. ROS Generation Analysis*

A reactive oxygen species (ROS) assay kit (Nanjing Jiancheng Bioengineering Institute, Nanjing, China) was utilized to detect the ROS level in treated IPEC-J2 cells in 96-well plates, 10<sup>4</sup> cells in one well, with nine parallel holes in each group. The 2 ,7 -dichlorohydrofluorescein diacetate (DCFH-DA) is the most sensitive and commonly used probe for detecting intracellular ROS. Specifically, all cell samples were treated with 10 μM DCFH-DA solution, after pretreated with *BaSC06* and DQ at 37 ◦C for 30 min and subsequently washed with FBS-free media and PBS. DCFH is oxidized into a strong green fluorescence substance DCF (dichlorofluorescein) in the presence of ROS in cells, and its emission wavelength is 502 nm, while the fluorescence peaks at the wavelength of 530 nm. Its fluorescence intensity is proportional to the ROS levels in cells. The fluorescence signals were monitored using the microplate reader SpectraMax M5 (Sunnyvale, CA, USA) and an Olympus BX61W1-FV1000 laser scanning confocal microscope (Tokyo, Japan).

### *2.5. Detection of Antioxidant Capacities*

IPEC-J2 cell lysates were gathered to determine oxidative stress indexes including the level of methane dicarboxylic aldehyde (MDA), and the activities of superoxide dismutase (SOD), catalase (CAT), as well as the glutathione peroxidase (GSH-Px), each group with 6 replications. All assays were performed following the manufacturer's guidelines (Jiancheng Bioengineering Institute, Nanjing, China) [28].

### *2.6. Apoptosis Cell Analysis by TUNEL Assay*

For this experiment, an Apoptosis Detection Kit named TUNEL BrightGreen (Vazyme, Nanjing, China) was used to detect the apoptosis levels in IPEC-J2 cells (*n* = 6). The 3 hydroxyl terminus of the broken DNA can bind to FITC-12-DUTP, which is activated by Terminal Deoxynudleotidyl Transferase (TdT). Based on the manufacturer's instruction, 4% paraformaldehyde was used to fix the slides. Then all the slides were incubated with proteinase K solution at a concentration of 20 μg/mL, followed by 50 μL BrightGreen labeling mix as well as 50 μL recombinant TDT enzyme. Then, the stained IPEC-J2 cells were instantly examined under an Olympus BX61W1-FV1000 laser scanning confocal microscope (Tokyo, Japan). Furthermore, IPEC-J2 cell were collected to be measured by a FC500 flow cytometer (Beckman Coulter, Fullerton, CA, USA).

### *2.7. Annexin V-FITC/PI Apoptosis Assay*

Annexin V has been used as a sensitive indicator of early apoptosis because it binds to the membrane of early apoptotic cells via phosphatidylserine exposed externally. Propidium iodide (PI) is a nucleic acid dye that cannot penetrate the entire cell membrane, but PI can penetrate the cell membrane and make the nucleus red due to increased membrane permeability in the middle and late apoptotic cells as well as dead cells. Therefore, annexin V matched with PI could be used to distinguish cells at different stages of apoptosis. IPEC-J2 cells were collected and washed in PBS. Then they were suspended in a 1× annexin binding buffer, mixed, and incubated with 5 μL annexin V-FITC V and 5 μL PI for 10 min. After that, a FC500 flow cytometer (Beckman Coulter, Fullerton, CA, USA) was utilized to measure the stained cells.

### *2.8. Detection of Caspase-3 Activity*

The activity of caspase-3 was measured by the Caspase-3 Activity Kit (Beyotime, Shanghai, China). After treatment, IPEC-J2 cells were collected after rinsed with cold PBS and centrifugation. Caspase-3 activity in the supernatant were assayed using the kit following the instruction. The caspase activity was expressed as the percentage of enzyme activity compared to the control.

### *2.9. FOXO3a siRNA and Transfection*

Three small interfering RNAs targeting pig FOXO3a and negative control siRNA were synthesized by Sangon Biotech, Shanghai, China, the sequences are summarized in Table 1. IPEC-J2 cells were cultured in antibiotic-free medium, then treated with Lipo-2000 and siRNA mixture. All the cells were collected for western blotting.


#### **Table 1.** List of siRNA.

### *2.10. Western Blotting*

We refer to the protocol of Xiao et al. for our Western blot assay [29]. The cell lysis buffer produced by Western and American Psychological Association (RIPA, Biotime Biotechnology, Xiamen, China) was used to prepare the total IPEC-J2 cell lysates (*n* = 3), while the nuclear protein extraction kit (Biotime Biotechnology, Xiamen, China) was utilized to gather nuclear proteins (*n* = 3). In accordance with the manufacturer's protocol, a BCA Protein assay kit (Biotime Biotechnology, Xiamen, China) was used to measure protein concentrations. After the SDS-PAGE, proteins were electrophoretically transferred to nitrocellulose membranes (Sangon Biotech, Shanghai, China). Subsequently, the membranes

were incubated with first antibodies at 4 ◦C overnight, and thereafter incubated with secondary antibodies goat anti-mouse IgG-HRP and goat anti-rabbit IgG-HRP after washing by TBST. The blots were then developed with an ECL detection system (Tanon 5200, Shanghai, China). Anti-Bcl2 and anti-LC3II primary antibodies were purchased from Abcam (Cambridge, MA, USA), while primary antibodies such as anti-Bax, anti-Beclin1, anti-SQSTM/p62, anti-phosphor-mTOR (Ser2448), anti-mTOR, anti-phosphor-AKT (Ser473), and anti-AKT used in this study were obtained from CST (Danvers, MA, USA). Additionally, anti-FOXO3, anti-phosphor-FOXO3 (Ser253), and anti-Histone1 were acquired from Huabio (Hangzhou, China). Mouse anti-β-actin monoclonal antibody was obtained from Biotime Biotechnology, Xiamen, China. For all the proteins mentioned above, the relative density was analyzed with ImageJ software.

### *2.11. Immunofluorescence Analysis*

IPEC–J2 cells were seeded and cultured in 12-well plates for 12 h to reach 70% confluence. Cell samples (*n* = 6) were fixed with cold methanol for 5 min, and after that, all the samples were blocked with 2.5% BSA at room temperature for 2 h, followed by incubating with anti-LC3II antibody (Abcam, Cambridge, MA, USA) for 12 h at 4 ◦C. Thereafter, Alexa Fluor 488 conjugated antibody (Biotime Biotechnology, Xiamen, China) and DAPI solution (Biotime Biotechnology, Xiamen, China) were used to stain cells and nucleus. An Olympus BX61W1-FV1000 laser scanning confocal microscope (Tokyo, Japan) was utilized to obtain images.

### *2.12. RNA Extractions and Quantitative Real-Time PCR (qPCR) Analysis*

Total RNA was extracted utilizing RNAiso plus (Takara, Japan) from IPEC-J2 cells. The Nanodrop was used to examine the concentration of RNA. We utilized the Prime-Script II 1st Strand cDNA Synthesis Kit (Vazyme, Nanjing, China) to synthesize cDNA based on the manufacturer's manual. The qRT-PCR was conducted with the use of the HiScript II One Step qRT-PCR SYBR Green Kit (Vazyme, Nanjing, China) based on the manufacturer's manual. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used to normalize the amount of total RNA as an endogenous control. The Primer 5.0 as well as the Oligo 7.0 software were utilized to design and validate the primers. The primer sequences are summarized in Table 2. We estimated the abundance of the mRNA using the 2−ΔΔCt method.



### *2.13. RNA Extraction and RNA-Seq Analysis*

After the total RNA extraction, mRNA was purified from total RNA using poly-T oligo-attached magnetic beads. First-strand cDNA was synthesized using random primers. Second-strand cDNA was synthesized by DNA polymerase I, Rnase H, dNTP, and buffer. Then, the cDNA fragments were purified with an AMPure XP system (Beckman Coulter, Beverly, GA, USA). Afterwards, PCR was performed with Phusion High-Fidelity DNA polymerase, Universal PCR primers, and Index (X) Primer. Lastly, the PCR products were purified (AMPure XP system) and the library quality assessed on an Agilent Bioanalyzer 2100 system. The ligated products were size-selected by agarose gel electrophoresis and PCR, and then sequenced using Illumina HiSeqTM 4000. For the bioinformatics analysis, the original reading containing the adapter or low quality (Q value ≤ 10) was removed and mapped to the pig reference genome (assembly SSCROFA 11–1). We further analyzed the differentially expressed genes (DEGs) between different samples or groups with the DeSeq2 R package (1.16.1). Genes with a fold change of ≥1, at a false discovery rate of *p* ≤ 0.05, were considered significantly differentially expressed. The enrichment of DEGs was performed using the KEGG pathway database.

### *2.14. Statistical Analysis*

All data are showed in the form of the mean ± standard deviation (SD). Statistically significant differences between means were calculated with a one-way ANOVA with a Tukey test and analyzed in SPSS statistical software, version 23.0 (IBM®, Chicago, IL, USA). The IC50 of the DQ to IPEC-J2 cells was calculated utilizing the probit method in SPSS statistical software, version 23.0 (Chicago, IL, USA). *p* < 0.05 was considered statistically significant. Figures were prepared utilizing Prism 9.0 software (GraphPad Software Inc., San Diego, CA, USA) and Origin 8.0 software (Origin Lab Corporation, Northampton, MA, USA).

### **3. Results**

### *3.1. Establishment of Oxidative Stress Model Induced by Diquat in IPEC-J2 Cells*

IC50 represents 50% of the inhibitor concentration required for inhibition of enzymes, cells, cell receptors, or microorganisms. DQ, which is widely utilized as an herbicide in agriculture, was used to establish an oxidative stress model. Notably, DQ reduced the IPEC-J2 cells' viability in a dose-dependent manner (Figure 1a). Utilizing the probit method, the IC50 for DQ in IPEC-J2 cells was 932.2 μmol/mL. Therefore, 950 μmol/mL DQ was used in the following experiment. IPEC-J2 cells were exposed to *BaSC06* at different concentrations (0, 105, 106, 107, 108, and 109 CFU/mL) for 6 h, and their cell viability was further detected. As shown in Figure 1b, the cell viability at the 10<sup>8</sup> CFU/mL treatment was the closest to 100% (105.384%), showing no significant change compared to the cells in the CK group (*p* > 0.05). Hence, 108 CFU/mL *BaSC06* was used in further experiments.

**Figure 1.** The establishment of a diquat-induced oxidative stress model. (**a**) IPEC-J2 cells were treated with DQ at various concentrations (0, 250, 500, 750, 1000, 1250 μmol/L) for 6 h. The IC50 was calculated using the probit method. (**b**) IPEC-J2 cells were treated with *BaSC06* for 6 h at different concentrations (0, 105, 106, 107, 108, and 109 CFU/mL). Cell viability was calculated using an CCK-8 kit. Data are exhibited in the form of the mean ± SD (*n* = 9), and significance was measured by one-way ANOVA with a Tukey test: \* *p* < 0.05, \*\* *p* < 0.01, and ns = no significance (*p* > 0.05).

### *3.2. BaSC06 Alleviated Oxidative Stress Induced by DQ in IPEC-J2 Cells*

Consequent ROS generation and redox cycling are believed to be the key causes of oxidative stress induced by DQ [30]. The production of ROS in IPEC-J2 cells was examined using a DCFH-DA fluorescence assay and Microplate Reader. Compared with the CK group, DQ treatment significantly increased the generation of ROS (32.73 ± 4.755%, *p* < 0.001), which was markedly inhibited by *BaSC06* pretreatment (10.70 ± 1.588%, *p* < 0.01) (Figure 2a,b). A similar result detected by microplate reader was obtained (Figure 2c). In addition, MDA content was significantly increased in the DQ group (14.12 ± 1.07 nmol/mg protein, *p* < 0.01) (Figure 2d), but markedly decreased in the Ba+DQ group (4.03 ± 1.07 nmol/mg protein, *p* < 0.01). However, no significant difference was observed in the GSH-Px activity between CK and DQ groups, which was significantly increased in *BaSC06*-treated cells (*p* < 0.05) (Figure 2e). At the same time, compared with the CK group, the activity of T-SOD decreased by 48.2% (*p* <0.05) in the DQ group, which significantly increased by 26.3% in *BaSC06* pretreatment group (*p* < 0.05) (Figure 2f). These results suggest that *BaSC06* could enhance the antioxidant capacity of IPEC-J2 cells by decreasing the production of ROS and increasing the activities of antioxidant enzymes.

**Figure 2.** *BaSC06* alleviated oxidative stress induced by DQ in IPEC-J2 cells. (**a**–**c**) Levels of ROS measured by DCFH-DA fluorescence assay, and the data are presented as the ratio of green fluorescence and the percentage of ROS production by DHE. Scale bar: 100 μm; *n* = 9 in each group. (**d**–**f**) Antioxidant capacity in cell lysates indicated by MDA, T-SOD, and GSH-Px; *n* = 6 in each group. All data were analyzed utilizing one-way ANOVA with Tukey test: \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001, and ns = no significance (*p* > 0.05).

#### *3.3. BaSC06 Alleviated Oxidative Stress-Induced Apoptosis in IPEC-J2 Cells*

The results of the expression of apoptosis-related proteins showed that *BaSC06* markedly upregulated the protein level of Bcl2 (*p* < 0.05), while that in the DQ treatment group indicated no significant changes compared with untreated cells (*p* > 0.05) (Figure 3a,c). The expression of Bax and the ratio of Bax/Bcl2 was significantly increased in cells treated with DQ, but *BaSC06* pre-treatment altered this trend significantly (*p* < 0.05) (Figure 3a,b,d). DQ also increased the activity of caspases-3 (Figure 3f), the relative intensity of cleaved caspase-3 and the mRNA levels of caspase-8 considerably (*p* < 0.01), which were significantly reversed by BaSC06 pre-treatment (*p* < 0.01, Figure 3e,g). In addition, compared with the CK group, the number of apoptotic bodies (green puncta) in the DQ group, detected using a TUNEL kit, was considerably increased by 31.40 ± 7.19% (*p* < 0.01), but no significant changes and a significant decrease (18.25 ± 5.843%) were observed in the Ba group and Ba+DQ group (*p* < 0.01), respectively (Figure 3h,i). Furthermore, the similar results were obtained using TUNEL assay by flow cytometry (Figure 3j,k). *BaSC06* pretreatment dramatically downregulated DQ-triggered apoptosis (*p* < 0.01). The percentages of early and late apoptotic cells in the CK, DQ, Ba, and Ba+DQ groups were 5.05 ± 1.31%, 12.95 ± 0.64%, 5.67 ± 0.45%, and 6.48 ± 0.40%, respectively (Figure 3l,m). Overall, these findings demonstrate that pre-treatment with *BaSC06* might significantly alleviate apoptosis induced by DQ in IPEC-J2 cells.

**Figure 3.** *Cont*.

**Figure 3.** *BaSC06* pretreatment inhibited apoptosis in IPEC-J2 cells during diquat exposure. (**a**–**e**) The ratio of Bcl2/β-actin, Bax/β-actin and Bax/Bal2, and cleaved caspase-3/β-actin were analyzed using ImageJ software. Data were analyzed utilizing one-way ANOVA with a Tukey test; *n* = 3 in each group; \* *p* < 0.05, \*\* *p* < 0.01, and ns = no significance (*p* > 0.05). (**f**) Caspase-3 activity was evaluated using a caspase-3 activity assay kit. (**g**) The mRNA expression levels of caspase-8 were determined using quantitative real-time PCR. (**h**,**i**) TUNEL assay. IPEC-J2 cells were stained with a BrightGreen apoptosis detection kit, after being treated with 108 CFU/mL BaSC06 and 950 μmol/mL diquat. Scale bar: 5 μm. (**j**,**k**) TUNEL assay by flow cytometry. (**l**,**m**) Annexin V-FITC/PI apoptosis assay. Apoptotic cell rates were detected with a FITC annexin V-FITC/PI apoptosis kit, and then analyzed by flow cytometry. All data were analyzed using one-way ANOVA with a Tukey test; *n* = 6 in each group; \* *p* < 0.05, \*\* *p* < 0.01, and ns = no significance (*p* > 0.05).

### *3.4. BaSC06 Triggered Autophagy during Oxidative Stress in IPEC-J2 Cells*

We further evaluated whether *BaSC06* can induce autophagy in IPEC-J2 cells under oxidative stress. As expected, 10<sup>8</sup> CFU/mL *BaSC06* upregulated the LC3II/LC3I ratio dramatically in a time-dependent manner, particularly from 6 h to 10 h (*p* < 0.05) (Figure 4a,b), but decreased p62 expression (Figure 4a,c). A significant decreased expression of LC3-II in the DQ and Ba+DQ group was found (*p* < 0.05) (Figure 4d,e), and no significant change was observed between the two groups (*p* > 0.05). Further, DQ treatment substantially inhibited the degradation of p62 (*p* < 0.05); however, *BaSC06* pretreatment dramatically blocked this trend (*p* < 0.05) (Figure 4d,f). The ratio of LC3II/LC3I markedly increased in the *BaSC06* group compared to the DQ group (*p* < 0.01) (Figure 4d,e). Additionally, compared with the CK group, *BaSC06* pretreatment markedly upregulated the expression of Beclin1 (*p* < 0.05), while that in the DQ group exhibited no significant change (*p* > 0.05) (Figure 4d,g). The results of autophagy based on the percentage of LC3-positive cells demonstrated that, compared with the CK group, DQ significantly downregulated LC3-positive cells (45.92 ± 3.23%, *p* < 0.01), while *BaSC06* significantly increased LC3 puncta (62.81 ± 4.90%, *p* < 0.001) (Figure 4h,i). To sum up, these findings suggest that *BaSC06* treatment can induce autophagy in IPEC-J2 cells.

**Figure 4.** Effects of *BaSC06* on autophagy during oxidative stress in IPEC-J2 cells. (**a**–**c**) *BaSC06* triggered autophagy in IPECJ-J2 cells in a time-dependent manner. Cell lysates were collected to detect the protein levels of LC3-II/LC3-I and p62/β-actin, and data were analyzed using ImageJ software. (**d**–**g**) IPEC-J2 cells pretreated with *BaSC06* in the concentration of 10<sup>8</sup> CFU/mL for 6 h, subsequently treated with DQ in the concentration of 950 μmol/mL for another 6 h. The LC3II/LC3I, p62/β-actin, or Beclin1/β-actin ratio was analyzed by ImageJ software. Data were analyzed using one-way ANOVA with a Tukey test; *n* = 3 in each group; \* *p* < 0.05, \*\* *p* < 0.01, and ns = no significance (*p* > 0.05). (**h**,**i**) *BaSC06* increased LC3 puncta in IPEC-J2 cells. After the same treatment with (**d**), IPEC-J2 cells were stained and visualized under confocal microscopy for immunofluorescence analysis. Scale bar: 5 μm; *n* = 6 in each group. The number of LC3-positive cells was statistically analyzed: \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001, and ns = no significance (*p* > 0.05).

### *3.5. KEGG Pathway Analysis of DEGs*

The DEGs' numerical analysis is depicted in Figure 5a,c,e. It was uncovered that compared to the CK group, *BaSC06* induced the upregulation of 115 and downregulation of 148 genes, while DQ induced the upregulation of 546 and downregulation of 500 genes. In addition, *BaSC06* upregulated 215 and downregulated 309 genes compared to the Ba+DQ group (fold change > 1 and *p* < 0.05). Then, a KEGG enrichment pathway analysis was performed and obtained the autophagy-related pathways, including MAPK, AMPK, PI3K-AKT, P53, FOXO, JAK-STAT, NF-kappa B, TNF, TGF-beta, and mTOR signaling pathways (Figure 5b,d,f). Among them, the FOXO signaling pathway was enriched with the most significant differential genes, which implies that this key pathway may regulate *BaSC06* or diquat-induced autophagy in IPEC-J2 cells. In addition, *BaSC06* was also found to have an effect on Akt in rats in our previous study [27], so we speculated that autophagy may be regulated by the Akt–FOXO signaling pathway.

Next, differential genes in the FOXO signaling pathway were examined, and 83 genes with significant differences were identified (Figure 5g). To verify the identified DEGs by qRT-PCR, we randomly selected 5 genes, SGK1, Beclin1, Raf1, MDM2, and STAT3, from the significant DEGs. The results revealed that the sequencing results were consistent with the RT-qPCR data, which confirmed the reliability of the sequencing data (Figure 5h–l).

**Figure 5.** *Cont*.

**Figure 5.** *Cont*.

**Figure 5.** Upregulation and downregulation genes, and differential genes in the FOXO signaling pathway and validation of the DEGs data by RT-qPCR. (**a**–**f**) Volcano plots of the DEGs. The x-axis indicates the difference in expression level on a log2 scale (fold change), while the y-axis represents the *p*-value. Red represents the upregulation gene whereas green denotes downregulation genes. Bubbles of KEGG pathways for differential gene enrichment. The circle presents the gene number. The color of the circles indicates the *p*-value. (**g**) Heatmaps of all differential genes in the FOXO signaling pathway. (**h**–**l**) Validation of the RNA-Seq expression profiles of genes randomly selected from DEGG by RT-qPCR; *n* = 9 in each group. Black bars represent FPKM (Fragments Per Kilobase Million), while grey bars represent the fold change. FPKM normalized values as gene expression in RNA-seq. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001, and ns = no significance (*p* > 0.05). Values are the mean ± SD.

### *3.6. BaSC06 Can Regulate the Transcriptional Activity of FOXO3 Transcription Factor in IPEC-J2 Cells*

It is well known that the subcellular localization and transcriptional activity of FOXO proteins are mainly regulated by posttranslational modifications, including phosphorylation, acetylation, and ubiquitination [31]. Phosphorylation of FOXO3 can transfer it from the nucleus to the cytoplasm, thereby inactivating FOXO3. Currently, the expression of the p-FOXO3 protein in the DQ group was significantly increased (*p* < 0.01), which was reversed by the pretreatment of *BaSC06* (*p* < 0.01) (Figure 6a,b). More importantly, the expression of Sirt1 is crucial due to its deacetylation function. Noticeably, *BaSC06* significantly elevate the reduced expression of SIRT1 induced by DQ (*p* < 0.05) (Figure 6a,c), which implied that *BaSC06* may decrease acetylation of FOXO3. The protein expression FOXO3 in the nucleus decreased significantly in the DQ group (*p* < 0.05), while pretreatment with *BaSC06* inhibited this change in IPEC-J2 cells (Figure 6a,d). In summary, these outcomes indicate that *BaSC06* can increase the FOXO3 expression in nuclear by inhibiting its phosphorylation and increasing its deacetylation, thereby increasing the transcriptional activity of the FOXO3.

**Figure 6.** Effects of *BaSC06* on FOXO3 transcription factor in IPEC-J2 cells. (**a**–**d**) The expression of phosphate-FOXO3, FOXO3, and Sirt1 protein in IPEC-J2 cells. Cell lysates were collected to detect the protein levels of p-FOXO3/FOXO3, Nu-FOXO3/H1, and SIRT1/β-actin, and data analyses were executed using ImageJ software. Data were analyzed using one-way ANOVA with a Tukey test; *n* = 3 in each group; \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001, and ns = no significance (*p* > 0.05).

### *3.7. BaSC06 Mediated Autophagy by the AKT–FOXO Signaling Pathway Independent of mTOR*

The ratio of p-AKT/AKT exhibited significantly reduced in the *BaSC06* pretreatment groups compared to the DQ group (*p* < 0.05) (Figure 7a, b). However, no significant change for the ratio of p-mTOR/mTOR was observed in all groups (*p* > 0.05) (Figure 7a,c). The results showed that the addition of perifosine (the inhibitor of AKT phosphorylation) significantly reduced the phosphorylation of FOXO3a (*p* < 0.05) (Figure 7d,e), confirming that the transcriptional activity of FOXO3a was upregulated. Then we used siRNA to reduce FOXO3 expression, and further detected the expression of autophagy related proteins in IPEC-J2 cells. SscFOXO3a-608 had the best silence effect (34.53 ± 0.16%), the protein expression level of FOXO3a was significantly reduced, so we used it in subsequent experiments (Figure 7h,i). The FOXO3/β-actin and the LC3II/LC3I ratio decreased significantly in sscFOXO3a-608-added groups (*p* < 0.05) compared to the NC group, but no significant difference was observed among all sscFOXO3a-608-added groups (*p* > 0.05) (Figure 7j–l). Although the expression level of P62 showed the increasing tend when Ba was added, there was no significant difference compared with NC group (*p* > 0.05). These results demonstrated that the decrease of FOXO3 expression significantly inhibited autophagy in IPEC-J2 cells, and that *BaSC06* induced autophagy by the AKT-FOXO signaling pathway, not by the AKT/mTOR signaling pathway.

**Figure 7.** *BaSC06* mediated autophagy was triggered by the AKT–FOXO signaling pathway independent of mTOR, and the effect of *BaSC06* on FOXO downstream autophagy-related target genes. (**a**–**c**) The expression of p-AKT, AKT, p-mTOR, and mTOR protein in IPEC-J2 cells. Cell lysates were gathered to detect the ratio of *p*-AKT/AKT and p-mTOR/mTOR.

(**d**–**g**) We added 25 μM perifosine into the *BaSC06* pretreated groups and collected the cell lysates to detect the protein levels of p-FOXO3/ FOXO3, LC3II/LC3I, and p62/β-actin. (**h**,**i**) The efficacy of all siRNAs. (**j**–**m**) The expression of FOXO3//β-actin, LC3II/LC3I, and p62 /β-actin in siRNA treatment groups. All data analyses were implemented using ImageJ software. Data were analyzed using one-way ANOVA with a Tukey test; *n* = 3 in each group; \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001, and ns = no significance (*p* > 0.05). (**n**–**r**) The relative mRNA expression levels of ATG5, ATG12, ATG8, ATG14, and ATG16L1. Data were evaluated using one-way ANOVA with a Tukey test; *n* = 9 in each group; \* *p* < 0.05, \*\* *p* < 0.01, and ns = no significance (*p* > 0.05).

> We further explored the expression of downstream autophagy-related target genes of FOXO. Our results demonstrated that compared with the CK group, the mRNA expressions level of ATG5, ATG12, ATG8, as well as ATG14 of the DQ group were substantially decreased, whereas these genes expression level significantly up-regulated in *BaSC06* treatment group (*p* < 0.05) (Figure 7n,o,q,r). Although treatments of DQ and *BaSC06* alone insignificantly affected the mRNA expression level of ATG16L1 (*p* > 0.05), pretreatment with *BaSC06* still markedly increased the mRNA expression level of ATG16L1 when it was compared with the DQ group (*p* < 0.05) (Figure 7p). These results verified that *BaSC06* could activate the AKT-FOXO signaling pathway to induce autophagy.

### **4. Discussion**

Diquat, which is widely used as an herbicide in agriculture, can be absorbed by green plants that turn it into peroxide-free radicals. Toxicological studies have demonstrated that diquat can cause damage to the digestive system as well as to the lung, liver, and other organs, eventually leading to death [30] and can induce oxidative stress and injuries in intestinal epithelial cells [32]. Therefore, Diquat has been largely applied to induce oxidative stress in vivo, because it can significantly increase the serum MDA level as well as inhibit the activities of antioxidant enzymes (including SOD and GSH-Px) [33–35]. But different cells have different sensitivities to it. 80 μmol/L DQ was used in our previous study in IEC-6 cells [27]. Current study found that the IC50 for DQ in IPEC-J2 cells was 932.2 μmol/ml, which is consistent with the results by Xu C et al. (2018), who used 1 mmol/L diquat to successfully establish an oxidative stress model on IPEC cells [36].

Elevated ROS production and low antioxidant capacity has been identified to induce oxidative stress in cells. MDA is an important biomarker of lipid peroxidation. Importantly, GSH-Px, CAT, and T-SOD are the main antioxidant enzymes in cells [37]. SOD terminates lipid peroxidation and eliminates ROS; however, its activity might be defected during acute injuries, leading to DNA damage, aggregated lipid peroxidation, and cell dysfunction [38]. The results of the present study showed that DQ exposure significantly increased MDA level in IPEC-J2 cells, while *BaSC06* pretreatment dramatically decreased the levels of ROS and MDA by reversing the decreased activities of SOD and GSH-Px induced by DQ, suggesting the potential role of *BaSC06* on alleviating oxidative stress, which corresponded with previous reports [27,36,39,40].

Evidence suggests that apoptosis can be caused by excessive oxidative stress, while excessive ROS production results in cell injury and death. Mitochondria are the major intracellular source of ROS [41]. The Bcl-2 protein family consists of Bcl-2, Bax, and Bak, among others. Bcl-2 is primarily located in the inner mitochondria membrane, the endoplasmic reticulum, and the perinuclear membrane and its function is regulated by its protein products, Bax and Bcl-xl [42]. What's more, the Caspase family is a key protein of apoptosis [43]. Our results also revealed that DQ can induce cell apoptosis and damage in IPEC-J2 cells, which was indicated by the increased caspase-3 activity and the expression level of cleaved caspase-3 protein and the mRNA expression of caspase-8 gene, which is consistent with the results of our previous research on rats, *BaSC06* also decreased apoptosis of rats caused by diquat both in vivo and in vitro, which was verified by the reversal of the upregulated caspase-3 and Bax expression, and the decreased expression of Bcl2 [27]. Besides, DNA fragmentation has been found to occur during cell apoptosis. A study has reported that ROS reacts with cellular macromolecules through oxidation

causing the cells to undergo an active process of cell death, which is set in motion by a high-conserved genetic program and culminated in DNA fragmentation and the formation of apoptotic bodies [44]. In this study, we verify that *BaSC06* can decrease apoptosis caused by diquat-induced oxidative stress via regulating the expression of apoptotic proteins and reducing the production of apoptotic bodies.

Autophagy is a catabolic pathway that is activated in response to different cellular stressors, such as damaged organelles, accumulation of misfolded or unfolded proteins, ER stress, accumulation of ROS, and DNA damage [45]. Several studies have reported that autophagy activation could alleviate oxidative stress. For instance, spermidine provides neuroprotection against oxidative stress and apoptosis by activating autophagy in aging male rats [46]. Tetrahedral framework nucleic acid inhibits chondrocyte apoptosis and oxidative stress through activation of autophagy [47]. In addition, quercetin helps the retina external barrier avoid oxidative stress injury by promotion of autophagy [48]. However, little information is available on the effect of probiotics on oxidative stress by regulating autophagy. Our previous study demonstrated that the autophagy induced by *BaSC06* was involved in decreasing oxidative stress in the rat [27]. Interestingly, we herein obtained similar results that *BaSC06* can upregulate the expression of LC3 and Beclin1, degrade p62, and increase LC3 puncta during diquat-induced oxidative stress in IPEC-J2 cells. Thus, it was suggested that the probiotic *BaSC06* contributes to alleviating oxidative stress by inducing autophagy in IPEC-J2 cells.

The results of KEGG enrichment pathway analysis revealed that DEGs and enriched major autophagy-related pathways contained P53, FOXO, JAK-STAT, NF-kappa B, TNF, TGF-beta, MAPK, AMPK, PI3K-AKT, and mTOR signaling pathways. In the Ba vs. CK, DQ vs. CK, and Ba+DQ vs. CK comparisons, FOXO was the primary signaling pathway that enriched the most significant DEGs regulating the autophagy process. The main DEGs in the FOXO signaling pathway are depicted in the heatmap included FOXO3, among others. Transcriptional factor FOXO3 has been uncovered to be extensively involved in autophagy and apoptosis, regulating the cell cycle, and participating in antioxidant stress response [49]. Moreover, FOXO3 activates autophagy by upregulating autophagy regulatory genes, including LC3, ATG12, γ-GABA receptor-associated protein 1 gene, yeast ATG8, and BNIP3 [50,51]. The acetylation, phosphorylation, and other post-transcriptional modification sites in FOXO significantly affect its DNA-binding activity, as well as its subcellular localization. The acetylation/deacetylation of FOXO regulates its transcriptional activity. Oxidative stress increase FOXO acetylation, and then acetylated-FOXO (Ac-FOXO) accumulates in the nucleus and binds to nucleosomes to block its transcriptional activity [52,53]. Furthermore, SIRT1 indirectly regulates autophagy by deacetylation of FOXO3, leading to increased expression of autophagy-related genes, including Bnip3, which are critical for autophagy induction [54]. The signal crossovers of SIRT1 and ROS can cause the decrease of autophagy and reduce the occurrence of the inflammatory response [55]. SIRT1 might deacetylate FOXO3 for oxidative stress response, thus improving the anti-stress ability of cells [56]. Overwhelming evidence suggests that SIRT1 deacetylates the FOXO factors, including FOXO1, FOXO3a, and FOXO4, and subsequently stimulate the expression of antioxidants, such as CAT, MnSOD, and Trx. Besides, an automatic feedback loop also potentiates SIRT1 expression [56–60]. In this work, *BaSC06* pretreatment significantly slowed the decline of SIRT1 induced by DQ, implying the increased deacetylation of FOXO3.

Phosphorylation/dephosphorylation of FOXO determines its subcellular localization. Phosphorylation of FOXO is mainly affected by protein kinase B (PKB or AKT). Studies have shown that p-AKT can phosphorylate FOXO3 at Thr32 /Ser315 /Ser253, thus preventing FOXO3 from entering the nucleus and inhibiting autophagy gene transcription, thus down-regulating autophagy level [61]. Hence, AKT-FOXO3 primarily exerts a negative regulatory role on autophagy regulation. Reduced nucleation of FOXO3a leads to decreased expression of reactive oxygen scavenging enzymes (superoxide dismutase and catalase), which results in increased intracellular ROS. However, some studies have suggested that PI3K-AKT-FOXO3 can promote the level of autophagy flow. The underlying mechanism is

that FOXO3, activated by phosphorylation, promotes the synthesis of glutamine synthase, and also prevents mTOR translocation to the lysosomal membrane in a glutamine synthasedependent manner, causing mTOR inhibition, thus promoting autophagy [62]. Our results revealed that in the DQ group, the significant increased expression of p-AKT, p-FOXO3 and the significant decreased nuclear FOXO expression were observed, although no significant changes were found in the expression of mTOR. Nevertheless, pretreatment with *BaSC06* reduced the expression of p-FOXO3 and restored it to the same level as in the CK group. We therefore hypothesized that *BaSC06* could increase the expression of FOXO3 by inhibition of p-AKT, thereby inducing the autophagy of IPEC-J2 cells. To further verify this result, we treated IPEC-J2 cells with perifosine, an AKT inhibitor. Perifosine can significantly reduce the phosphorylation of AKT and then reduce the extracellular signal-regulated kinase (ERK) 1/2, inducing cell cycle stagnation in G1 and G2 [63]. Prolong treatment of breast cancer cells with AKT inhibitors (which inhibit AKT phosphorylation) induces dephosphorylation of FOXO3a, nuclear translocation, and destruction of its binding to SIRT6, resulting in FOXO3a acetylation and BRD4 recognition [64]. Our results showed that *BaSC06* and the inhibitor perifosine exhibited almost the same effect. This suggests that both *BaSC06* and the inhibitor perifosine might significantly reduce p-FOXO3 expression, increase LC3 expression, and promote the degradation of p62. In addition, our results showed that *BaSC06* could not upregulate the ratio of LC3II/LC3I after FOXO3 knockdown, suggesting that FOXO3 is necessary for *BaSC06* to induce autophagy in IPEC-J2 cells. Overall, these outcomes indicate that *BaSC06* can promote autophagy and antioxidant enzyme production in IPEC-J2 cells by inhibiting the AKT-FOXO signaling pathway, thus alleviating oxidative stress.

The extension of autophagic vesicles leads to the formation of autophagosomes, usually bilayer organelles. During the process of autophagy, there are two ubiquitination processes that occur in the ATG5-ATG12 and LC3 systems, which are essential for the extension of autophagic vesicles and the maturation of autophagosomes [65]. In particular, ATG5 is a key protein involved in phagocytic membrane elongation in autophagic vesicular, which forms a constituent complex with ATG12. In this process, ATG7 activates ATG12 as an E1-like ubiquitin activase. Then, ATG12 is delivered to the E2-like ubiquitin transferase ATG10. Eventually, ATG12 binds to ATG5 for forming a complex [66]. Afterward, the ATG12-ATG5 complex and ATG16L further form the ATG12-ATG5-ATG16L complex, which is found to be located on the outer membrane of the autophagosome. Of note, the ATG5-ATG12-ATG16L complex exhibits E3 ligase-like activity, essentially by activating ATG3 enzyme activity, promoting LC3 (namely ATG8) transfer from ATG3 to the bottom phosphatidylethanolamine (PE) [65]. The ATG5-ATG12-ATG16L complex is disintegrated from the membrane, once the autophagosome formed. Our results suggest that *BaSC06* may induce autophagy in IPEC-J2 cells by promoting the formation of the ATG12-ATG5 complex, while the increase of ATG12 may be activated by FOXO deacetylation. These outcomes indicated that *BaSC06* can promote the expression of autophagy-related genes by increasing FOXO activity.

### **5. Conclusions**

Collectively, the possible molecular mechanisms underlying *BaSC06*-induced autophagy to attenuate oxidative stress are summarized as follows: *BaSC06* inhibited the AKT–FOXO signaling pathway by decreasing the expression of p-AKT and p-FOXO and increasing the expression of SIRT1, thereby increasing the transcriptional activity of FOXO3 and gene expression of the ATG5–ATG12 complex, which induce autophagy to alleviate oxidative stress in IPEC-J2 cells. Besides, *BaSC06* attenuates apoptosis by modulating the activities of antioxidant enzymes and the expression of the apoptotic proteins Bcl2, Bax, Caspase 3 and Caspase 8. All the above findings provide a valuable theoretical basis for the application of probiotics in preventing and treating diseases caused by oxidative stress and improving human and animal health (Figure 8).

**Figure 8.** Proposed model of the protective effect of *BaSC06* in DQ-induced oxidative stress and apoptosis. (1) *BaSC06* inhibited the AKT-FOXO signaling pathway by inhibiting the expression of p-AKT, p-FOXO and increasing the expression of SIRT1, thereby increasing the gene expression of the ATG5-ATG12 complex to induce autophagy. (2) *BaSC06* attenuate apoptosis by modulating the activities of antioxidant enzymes, the expression of the apoptotic proteins Bcl2, Bax, Caspase 3 and Caspase 8 to alleviate oxidative stress.

**Author Contributions:** Conceptualization, W.L. and L.T.; methodology, L.T.; software, L.T. and B.W.; validation, L.T., Z.Z., Y.Z. and P.Z. (Pengwei Zhao); formal analysis, L.T.; resources, Q.W. and P.Z. (Peng Zou); data curation, L.T. and B.W.; writing—original draft preparation, L.T.; writing—review and editing, L.T. and Z.Z.; visualization, S.X.; supervision, F.W. and J.Y.; project administration, X.L.; funding acquisition, W.L. and P.Z. (Pengwei Zhao). All authors have read and agreed to the published version of the manuscript.

**Funding:** This study is supported by the National Natural Science Foundation of China (No. 31672460, 32072766 and 31472128), the Natural Science Foundation of Zhejiang province (No. LZ20C170002), National High-Tech R&D Program (863) of China (No. 2013AA102803D), the Major Science and Technology Project of Zhejiang Province (No. 2006C12086), PRC.

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

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** All data that support the findings of this study are available in the submitted article except the RNA-seq, and the RNA-seq data are not publicly available due to privacy. The data are available from the corresponding author upon reasonable request.

**Conflicts of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

#### **References**

1. Qiao, Y.; Sun, J.; Ding, Y.; Le, G.; Shi, Y. Alterations of the gut microbiota in high-fat diet mice is strongly linked to oxidative stress. *Appl. Microbiol. Biotechnol.* **2012**, *97*, 1689–1697. [CrossRef] [PubMed]


MDPI St. Alban-Anlage 66 4052 Basel Switzerland www.mdpi.com

*Antioxidants* Editorial Office E-mail: antioxidants@mdpi.com www.mdpi.com/journal/antioxidants

Disclaimer/Publisher's Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Academic Open Access Publishing

www.mdpi.com ISBN 978-3-0365-8533-8