The gut microbiota composition affects dietary polyphenols-mediated cognitive resilience in mice by modulating the bioavailability of phenolic acids

Dietary polyphenols promote memory in models of sleep deprivation (SD), stress, and neurodegeneration. The biological properties of dietary polyphenols greatly depend upon the bioavailability of their phenolic metabolites derivatives, which are modulated by gut microbiota. We recently demonstrated that supplementation with grape-derived bioactive dietary polyphenol preparation (BDPP) improves SD-induced cognitive impairment. This study examined the role of the gut microbiota in the ability of BDPP to prevent memory impairment in response to SD. C57BL6/J mice, treated with antibiotics mix (ABX) or BDPP or both, were sleep-deprived at the end of a fear conditioning training session and fear memory was assessed the next day. Gut microbiota composition was analyzed in fecal samples and BDPP-driven phenolic acid metabolites extraction was measured in plasma. We report that the beneficial effect of BDPP on memory in SD is attenuated by ABX-induced dysbiosis. We identified specific communities of fecal microbiota that are associated with the bioavailability of BDPP-derived phenolic acids, which in turn, are associated with memory promotion. These results suggest the gut microbiota composition significantly affects the bioavailability of phenolic acids that drive the dietary polyphenols’ cognitive resilience property. Our findings provide a preclinical model with which to test the causal association of gut microbiota-polyphenols, with the ultimate goal of potential developing dietary polyphenols for the prevention/treatment of cognitive impairment.


Results
Gut microbiota contributes to dietary polyphenol-mediated promotion of resilience. To examine the role of gut microbiota in modulating the effect of BDPP treatment on promotion of cognitive resilience, male C57BL6 mice were randomly grouped into 6 treatment groups receiving either vehicle (Ctrl, SD) or antibiotic mix (ABX, SD + ABX) for 23 (Days 10-23) days or supplemented with BDPP (SD + BDPP, SD + ABX + BDPP) for 13 days (Days 10-23). On Day 22, we performed behavioral training for fear conditioning. At the end of the training session, Mice in Ctrl and ABX groups were left undisturbed in their home cages and mice in SD, SD + BDPP, SD + ABX, SD + ABX + BDPP were kept awake for 5 h. The next day (Day 23), mice were tested for memory performance (Fig. 1a). We did not find that any treatment had an effect on food intake or liquid consumption. In accordance to previous report 23 , mice in ABX treatment groups showed a small reduction of ~5% in body weight ( Supplementary Fig. S1).
We then assessed the gut microbial diversity within each treatment group (α-diversity) by calculating Shannon index at 6092 sequences/sample (Supplementary Table S4), which assumes that all microbial species are represented in our samples and that random sampling occurs. Using this index, we confirm a significant ABX-induced decrease in microbial α-diversity in vehicle treated mice ( Fig. 3b; ABX, 1.59 ± 0.05 vs. Ctrl, 2.81 ± 0.13, p < 0.0005) or BDPP treated mice ( Fig. 3b; ABX + BDPP, 1.38 ± 0.11 vs. BDPP, 3.27 ± 0.09, p < 0.0005). Interestingly, we found an increase in microbial α-diversity in mice that were treated with BDPP compare to vehicle-treated mice ( Fig. 3b; BDPP, 3.27 ± 0.09 vs. Ctrl, 2.81 ± 0.13, p < 0.05). We then assessed the dissimilarity of individual microbial communities between treatment groups (β-diversity). PCA plotting of weighted β-diversity PCA dissimilarity matrix data, accounting for presence and abundance of OTUs, revealed a distinct clustering effect in C57BL6 mice treated with either ABX (ABX) or ABX supplemented with BDPP (ABX + BDPP) compared to vehicle-treated (Ctrl) and BDPP-treated (BDPP) mice and clustering effect in BDPP compared to Ctrl (Fig. 3c). Collectively these data highlight that microbial diversity is altered by both ABX and BDPP treatment and suggest that, in the context of ABX treatment, the BDPP's impact on the microbiome is abolished.
ABX treatment in the context of BDPP affects OTUs association to freezing behavior via modulating the bioavailability of memory-promoting phenolic acids. Investigating OTUs association with freezing behavior revealed that the ABX-induced decrease in freezing behavior (Fig. 1b. ABX vs. Ctrl) was associated with alternation in 17 OTUs. The most significant finding included a positive association with freezing percent of the genus Lactobacillus and negative association of genera Nevskia, Halomonas, veronii and Pseudomonas (Supplementary Table S5, comparing ABX vs. Ctrl). However, no OTUs were associated with the BDPP-memory promoting effect, shown as BDPP-induced increase of freezing behavior in sleep deprived mice (Fig. 1b, Supplementary Table S5, comparing SD + BDPP vs. SD) or with the ABX-induced masking of BDPP-memory promoting effect, shown as decrease of freezing behavior in ABX and BDPP-treated compared to BDPP-treated, sleep deprived mice (Fig. 1b, Supplementary Table S5, comparing ABX + SD + BDPP vs. SD + BDPP). Together these results suggest that while ABX-induced modulation of gut microbiota might directly interfere with normal memory consolidation, it does not directly interfere with the beneficial effect of BDPP on memory consolidation, raising the possibility for an indirect effect of ABX on extending BDPP-induced memory promotion. We therefore asked whether antibiotic treatment in the context of BDPP would modulate the association between OTUs and the bioavailability of BDPP-derived phenolic acids that were found to be specifically associated with memory promotion. Focusing on groups ABX + SD + BDPP vs. SD + BDPP, we investigated the association between OTUs and the 9 phenolic acids that exhibited plasma concentrations in association with freezing behavior in these two groups (Fig. 2). The overall association (Wald's test) between OTUs and phenolic acids bioavailability were detected and visualized in a quantile-quantile plot (Fig. 4a). Integration of the association between phenolic acid and freezing percent (Fig. 2), and the association between phenolic acid and OTUs (Supplementary Table S6) indicates the influence of OTUs on behavior traits through modification of the concentration of phenolic acids in the plasma (Fig. 4b). These results demonstrate the predicted effect of a given OTU on freezing percent, suggesting that their ability to modulate the bioavailability of BDPP-derived phenolic acids in plasma may be a possible mechanism. Our complete analyses identified 110 significant phenolic acids -OTU pair-wise significant correlations (Supplementary Table S6), where the most significant included the negative correlation between HA and the veronii species (Supplemental Table 6, ABX + SD + BDPP vs. SD + BDPP, FDR = 2.2e-9) and the positive correlation between 4-HPVA and the Dorea genus (Supplementary  Table S6, ABX + SD + BDPP vs. SD + BDPP, FDR = 3.6e-9). We found 52 OTUs associated with at least one or more phenolic acids, including the Adlercreutzia genus that was positively associated with three phenolic acids; HA, 4-HPVA, and GA (Supplementary Table S6, ABX + SD + BDPP vs. SD + BDPP). Interestingly when an OTU is associated with changes multiple phenolic acids, the direction of their association is mostly consistent, and it is relatively rare that an OTU modulates multiple phenolic acids in different directions.

Discussion
The present study shows that ABX treatment attenuates dietary polyphenols' memory promoting effect in a model of SD. This behavioral change is associated with distinct alterations in the profile of circulating dietary polyphenol-driven phenolic acids, mediated by ABX-induced dysbiosis of the gut microbial community. Our observations suggest a role for gut microbiota in dietary polyphenols' efficacy in promoting cognitive resilience.
In line with previously reported data 17 , we show that treatment with dietary polyphenols is effective in protecting against impaired memory under conditions of SD. Memory promotion was accompanied by dietary polyphenol-induced modification of 74.5% of the gut microbiota composition. Such an effect on gut microbiota is supported by studies of dietary polyphenol-rich extracts administration in both animals [26][27][28][29] and humans 30,31 . Interestingly, ABX-mediated gut microbiota dysbiosis both attenuated the dietary polyphenols' memory promoting effect and eliminated their effect on gut microbial composition. These findings suggest causal association of the bioavailability of dietary polyphenols that mediate cognitive resilience with gut microbiota composition following the completion of both the treatment regimen and the behavioral tasking. To further investigate such a causal association, future studies will analyze the change in phenolic acids plasma concentration and OTUs composition and their association to the behavioral outcome before and after the administration of BDPP and ABX.
Supporting other studies exploring the effect of antibiotic drugs on memory performance 32-35 , we found that ABX treatment itself resulted in gut microbiota dysbiosis and induced memory impairment in non sleep-deprived (2019) 9:3546 | https://doi.org/10.1038/s41598-019-39994-6 www.nature.com/scientificreports www.nature.com/scientificreports/ mice. We found 27.4% of ABX-induced gut microbiota depletion to be significantly associated with memory impairment. However, the change in microbiota composition seen following dietary polyphenols treatment in the absence or presence of ABX was not associated with cognitive performance. While this could be a result of low sample size of dietary polyphenol-treated mice, another explanation may stem from the evidence showing that specific phenolic metabolites derived from BDPP exert the beneficial biological activities on cognitive performance [17][18][19] . We recently reported that a BDPP-driven phenolic acid, 3-HPPA, which was found to accumulate in the brain, potentially interferes with the assembly of neurotoxic β-amyloid aggregates, which play key roles in neurodegeneration 36 . Here, we report an association between ABX-treatment induced alternation in plasma concentration of BDPP-driven phenolic acids, including 3-HPPA, and cognitive performance. We found that specific microbiota species are associated with BDPP-driven phenolic acid plasma concentrations. The most significant findings pointed out that ABX induced lower proportions of the veronii genus and higher proportions of the phenolic acid HA. In metabolomics analyses, plasma concentration of HA was found to be higher in humans with mild cognitive impairment when compared to control subjects 37 . ABX-induced lowered proportions of the Adlercreutzia genus was associated with lower proportions of the phenolic acid GA. GA was found to improve memory deficit and hippocampal long-term potentiation following oral administration in rats 38 . Together, these findings suggest that the association of dietary polyphenols-mediated gut microbiota composition with cognitive performance is mediated by its significant effect on phenolic acid concentration, which drives the cognitive response. The observed extensive association between microbiota species and phenolic acids suggest a functional mechanism by which the gut microbiome modifies the bioavailability of dietary polyphenols-driven phenolic acid plasma concentration, which in turn influences cognitive resilience. In order to become bioactive in the human body, polyphenols must undergo diverse intestinal transformations, modifications that are the result of the action of microbiota metabolism on backbone reorganizations. Our findings shed light on the importance of specific microbial taxa in the recognition of polyphenol structure and subsequent generation of compounds that are absorbed and that confer cognitive resilience. Data on the impact of gut microbiota on polyphenol metabolites and their mechanisms of action to promote cognitive function are scarce. Studies exploring the gut microbiota-brain axis indicated dysbiosis mediated cognitive deficits were accompanied by reduced expression of hippocampal c-Fos 39,40 , an immediate early gene that is a target of the cAMP response elements-binding protein, www.nature.com/scientificreports www.nature.com/scientificreports/ CREB, which is required for hippocampus-dependent long-term memory formation 41 . We previously showed that dietary polyphenols promote memory under stressful conditions and models of neurodegeneration through activation of CREB 17,18,21,42 . We recently identified specific BDPP-driven phenolic metabolites that were able to restore sleep deprivation-induced impaired memory through upregulation of CREB signaling 17 . Modulation of CREB signaling implies that c-Fos signaling may be an important event through which dietary polyphenols promote memory function 43 , suggesting a mechanism of action for gut microbiota-derived polyphenol metabolites in inducing cognitive resilience.
Our findings suggest that the efficacy of dietary polyphenols in promoting resilience against sleep deprivation-mediated cognitive impairment may be causally attributable to the composition of the gut microbiota. Given the safety and tolerability of BDPP, our studies provide preclinical evidence to support the hypothesis that not only the food polyphenols, but also gut microbiota composition must be taken into account for the potential implementation of polyphenols for the prevention/treatment of cognitive dysfunction. The results also support the potential translational application of prebiotics to enhance dietary polyphenol compounds -mediated resilience to cognitive deficits.

Materials and Methods
Chemicals. Polyphenol-free diet (AIN-93G) was purchased from Research Diets, Inc. (New Brunswick, NJ). Food-grade resveratrol was purchased from ChromaDex (Irvine, CA, USA). GSPE was purchased from Supplement Warehouse (UPC 603573579173, Bolingbrook, IL, USA). One lot of the resveratrol and one lot of the GSPE were used for this particular study and were stored at 4 °C in the dark. Concord purple grape juice was purchased from Welch Foods Inc. Animals, treatments and sampling. C57BL6/J male mice (n = 122) were purchased from Jackson Laboratory at the age of 8-10 weeks and group housed (4/5-mice per cage) in the centralized animal care facility of the Center for Comparative Medicine and Surgery at the Icahn School of Medicine at Mount Sinai. All mice were maintained on a 12:12-h light/dark cycle with lights on at 07:00 h in a temperature-controlled (20 ± 2 °C) vivarium. Mice were allowed to adapt to the new environment for at least 2 weeks and were tested at 3-4 months of age. Mice were fed a polyphenol free diet. For assessing the gut microbiota's role in dietary polyphenols' effect in a model of SD-induced memory impairment, following 2-weeks of polyphenol-free diet, mice were randomly subdivided for six treatment groups including: non sleep-deprived mice receiving regular drinking water (n = 20, Ctrl), sleep-deprived mice treated with regular drinking water (n = 20, SD), sleep-deprived mice treated with BDPP for 13 days (n = 21, SD + BDPP), non sleep-deprived mice treated with ABX mix for 23 days (n = 19, ABX), sleep-deprived mice treated with ABX mix for 23 days (n = 20, SD + ABX), and sleep-deprived mice treated with ABX for 10 days, and then BDPP supplemented with ABX for another 13 days (n = 22, SD + ABX + BDPP). ABX mix, composed of Ampicillin (1 mg/ml), Vancomycin (0.5 mg/ml), Neomycin (1 mg/ml) and Metronidazole (1 mg/ml) with the addition of Splenda (artificial sweetener, 1 mg/ml) to improve palatability of water, was delivered through the drinking water for 23 days prior and throughout behavioral testing. BDPP, composed of GSPE, RESV and CGJ, was delivered through the drinking water for 13 days prior to and throughout behavioral testing. The calculated daily intake of GSE was 200 mg/kg body weight (BW) 1,16 , resveratrol was 400 mg/kg BW 44,45 and the total polyphenols from juice extract was 183 mg/kg BW 46 . These doses were chosen based on the equivalent doses used in a previous study that showed efficacy in animal models for each component 17 . ABX mix was changed once a week. BDPP was changed once every two days. To study treatment effect on memory consolidation, n = 7-10 mice per group were sacrificed by CO 2 euthanasia at the end of the 5 h sleep deprivation that followed behavioral training (Day 22) and blood samples were collected through cardiac puncture by inserting a 23 G needle into the ventricle. Plasma was recovered following centrifugation, acidified with formic acid (1:10 vol/vol, with the final concentration of formic acid 0.2%) and stored at −80 °C. To investigate treatment effect on gut bacterial population, feces were collected on day 21. Since fecal collection took place before the SD procedure, gut microbiota analyses was conducted in the following treatment groups: Ctrl; mice from Ctrl and SD groups, BDPP; mice from SD + BDPP group, ABX; mice from ABX and SD + ABX groups; ABX + BDPP; mice from SD + ABX + BDPP group. For all experiments, mice body weight and food consumption was assessed once a week. Liquid consumption was assessed every 2 days. Data is summarized in Supplemental Fig. 1. All procedures were approved by the Mount Sinai IACUC and conducted in accordance with the guidelines and regulations.
Sleep deprivation. Mice were sleep deprived by using the Automated Sleep Deprivation System for Mice, Pinnacle technologies Inc., KS, USA 47,48 . The system uses timed movements of the randomly rotating bar to prevent mice from sleeping. Mice were acclimated to the Automated Sleep Deprivation System for 30 min/day for 3 consecutive days prior to the sleep deprivation procedure. For memory consolidation examination, mice were maintained in the Automated Sleep Deprivation System for 5 h starting immediately after a 10 min training trial. For each system, n = 5 mice from the same home cage were housed together with ad lib access to food and treatments containing drinking bottles. At the end of the 5 h SD mice were either sacrificed for tissue collection or placed back in the home cage for the next day behavioral testing.
Cognitive assessment of SD-induced memory impairment. For characterizations of the gut microbiota's role in polyphenols' treatment effect on SD-induced memory impairment, we conducted a contextual fear conditioning test as previously described 49 . On Day 1, mice were placed into a fear conditioning chamber (Med Associates, 30.5 × 24.1 × 21 cm) containing gray walls, grid floor, houselights at 50% with lemon scent, and allowed to explore for 120 s prior to 2 s foot shock (0.7 mA) from the metal grid and then for additional 30 s. www.nature.com/scientificreports www.nature.com/scientificreports/ After the training session on Day 1, non-sleep deprived mice were left undisturbed in their home cages, and sleep-deprived mice were kept awake by the Automated Sleep Deprivation System. On Day 2, 24-h after the training session, mice were placed into the same context and allowed to explore for 120 s without the shock. Memory for the context (contextual memory) for each animal was obtained by measuring freezing behavior, defined by body immobility except for respiration. Freezing behavior was scored over the 2 min testing period and data were presented as a percentage of time in freezing over total sampling period. Freezing behavior was recorded with a side-mounted NIR camera and measured with ANY-maze ™ tracking software (Stoelting Co., IL, USA. Version 5.1 Beta). All animals were handled for 5 min/day for 5 days prior to behavioral testing. Mice were habituated to the testing room for 30 min at the beginning of training and test day. All tests were conducted by the same experimenter.
16S rRNA gene amplicon sequencing. Stool samples of mice in treatment groups (Ctrl, BDPP, ABX, ABX + BDPP) were collected under the same condition on Day 21 (prior to SD or behavioral testing). Each mouse fecal pellet was weighed. Fecal DNA was extracted by bead-beating and quantified with Qubit assays. The 16S rDNA amplicon sequencing of the V4 hypervariable region was performed with an Illumina MiSeq (paired-end 250 bp). For each PCR reaction, 4 ng of purified template DNA was amplified in a reaction volume of 20 μl. Primers were designed to target the V4 region of the 16S rDNA (position 515 of the bacterial 16S rRNA gene to position 806). Each reaction was denatured at 98 °C for 30 sec followed by 26 cycles of (98 °C × 10 sec, 50 °C × 30 sec, 72 °C × 30 sec), followed by a final extension at 72 °C for 2 min. Reactions each contained a unique sequence index to enable pooling. Pools were purified with 60 μl AmpureXP beads added to 100 μl of sample (i.e., a beads to sample ratio of 0.6) and sequenced on an Illumina MiSeq instrument at a loading concentration of 10 pM. The 16S rDNA data were analyzed with QIIME 1.9.1 for taxonomic classification analysis of OTUs 52 .
A training dataset for taxonomic assignments was created using a modified NCBI taxonomy from the 'isolated named strains 16S' in the Greengenes database 53 . This dataset was manually curated by (i) removing strains in 'isolated named strains 16S' that had non-standard taxonomy or that were not members of the domain bacteria, and (ii) grouping strain level taxonomy from Greengenes assignments under a single NCBI species assignment. The trained dataset is available at http://gordonlab.wustl.edu/SuppData.html and was used to train the Ribosomal Database Project (RDP) version 2.4 (56) classifier and to assign taxonomy to picked OTUs.
Shannon indices, indicative of α-diversity, were then calculated from refaction plots. Weighted UniFrac distances were then computed to produce β-diversity dissimilarity matrices 54 .

Statistical analyses.
For statistical analysis of fear conditioning freezing behavior and Shannon index for gut microbial α-diversity comparing treatments groups, one-way ANOVA followed by Tukey multiple comparisons was performed. Differences in phenolic acid plasma concentration collected on Day 22 in SD + BDPP vs. ABX + SD + BDPP group were analyzed by Wilcoxon rank test. Significantly different was defined as Benjamin-Hochberg false discovery rate (FDR) < 10.7%. Principal component analyses (PCA) was conducted for analyses of differences in the composition of OTUs among treatment groups. Each mouse data was projected to the space of the first 2 principle components. Association of OTUs composition with freezing behavior or with phenolic acids bioavailability was conducted with Univariate linear regression. For OTUs association with phenolic acids bioavailability, a quantile-quantile (QQ) plot was generated to visualize the distribution of the p-values, comparing the observed p-values against expected p-values under the null hypothesis (ie, no association). All values are expressed as mean and standard error of the mean (s.e.m). Statistical analysis was performed using GraphPad Prism 5 software (GraphPad Software, Inc.) and R statistical software R 3.0.2.

Data Availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.