Adult male C57BL/6NCrl strains of Mus musculus (Charles Rivers Laboratories, Wilmington, MA) between 6 and 8 weeks of age were used in the study. Animals were given ad libitum access to water and the experimental diets for 2 weeks prior to, as well as throughout, stressor exposure. Mice were kept on a 12 h light/dark cycle (lights on at 06:00). All animal experimental procedures were conducted in the Animal Resources Core (ARC) facility with the approval of the Research Institute at Nationwide Children's Hospital Institutional Animal Care and Use Committee (Protocol #AR15-00075). After being received from the vendor, mice were placed on one of three experimental diets. Researchers were blinded to treatment groups throughout the study and sample analyses. The data reported are from three separate experiments. Experiment 1 tested the hypothesis that prebiotic diets could attenuate stressor-induced increases in immunoreactivity as measured by splenic proliferation/viability and cytokine production. Experiment 2 assessed whether the prebiotics also affected the gut microbiome/metabolome. Because these first two experiments identified links between stress, prebiotics, the gut microbiome, and B vitamins, Experiment 3 tested whether stressor-induced immunopotentiation could be directly attenuated by B vitamins.

Upon arrival, animals were randomly assigned to cages that received AIN-93G control diet supplemented with 15 g/kg galactooligosaccharide (GOS; FrieslandCampina, Zwolle, Netherlands), 15 g/kg polydextrose (PDX; Danisco, Terre Haute, IN), and 2.2 g/kg Lacprodan SL-10 sialyllactose (SL; Arla Foods Ingredients Group P/S, Aarhus, Denmark; designated “GOS+PDX+SL”); or AIN-93G supplemented with 2.2 g/kg SL only (designated “SL”). The diets were formulated to be isocaloric and contained similar carbohydrate, protein and fat levels based on AIN-93G specifications. Each diet contained identical amounts of minerals (AIN-93G-MX-94046, 35 g/kg) and vitamins (AIN-93-VX- 94047, 10 g/kg). Mice were randomly assigned to the non-stress control condition or to exposure to the social disruption stressor. Social disruption occurred on six consecutive days, with fecal samples collected the morning following the 5th day of stressor exposure, and metabolites assessed the morning following the 6th day of stressor exposure.

For all experiments, the social disruption (SDR) stressor was used, which entails placing an aggressor mouse (retired breeder C57BL/6NCrl) into the cage with 3 resident mice of the same strain for 2 h between 16:30 and 18:30. During SDR, resident mice were observed for subordinate behaviors such as upright defeat posture, exposed under flank, and fleeing and cowering in the corners of the cage. Aggressors that did not initiate an attack within 5 min were switched with a new aggressor. At the end of each SDR cycle, intruder mice were removed and resident mice left undisturbed until the following day when SDR was repeated with a new aggressor. The health status of the mice was carefully examined throughout the experiment; severely injured mice were removed from the study. Consistent with previous studies using SDR (4), <5% of mice met the early removal criteria. Mice in the control condition were not handled throughout the experiment and were fed either the control or an experimental diet.

We assessed the effects of prebiotic diets (Experiment 1) and ex vivo B vitamin supplementation (Experiment 3) on stress- and immune- stimulated splenocytes. C57BL/6NCrl were sacrificed the morning following the 6th exposure to the SDR stressor and spleens were removed and processed to create a single cell suspension as previously reported (33, 34). Spleens were gently mashed through a 100 μm cell strainer, suspended in RPMI 1640 medium (Corning, Manassas, VA) supplemented with 10% heat-inactivated fetal bovine serum (Gibco laboratories, Gaithersbrug, MD), and centrifuged at 1,000 rpm for 10 min at 4°C. After decanting the supernatant the pellet was incubated for 2 min at room temperature in Red Blood Cell (RBC) lysis buffer. Following the incubation, the pellet was re-suspended in CTLL-20 RPMI 1640 (1 M HEPES, 200 mM L-glutamine, 7.5% sodium bicarbonate, 100x Antibiotic Antimycotic solution (Sigma Aldrich), β-Mercaptoethanol) and centrifuged. The total numbers of splenocytes were counted using the Countess II FL Automated Cell Counter (Life Technologies). Cell suspensions were diluted to a concentration of 1 x 10⁶ cells/mL of CTLL-20 RPMI. For Experiment 1, mice were fed Control, GOS+PDX+SL, or SL diets for 2 weeks prior to and 1 week during stress (n = 9–12/diet/stress group). For Experiment 3, mice were exposed to SDR or remained in home cage but were only fed the control diet (n = 6/stress group). In Experiments 1 and 3, splenocytes were stimulated with LPS (E. coli O55:B55 (1 μg/mL); Sigma-Aldrich) in replicate 96-well cultures. In Experiment 3, B vitamins (nicotinic acid, nicotinamide, nicotinamide riboside, pyridoxal, pyridoxamine, and pyridoxine) were added at 1,000 nM to separate wells of splenocytes (evenly distributed by individual animal) and stimulated with LPS (1 μg/mL). B vitamer concentrations (1,000 nM) were chosen based off of a previously published report which displayed that 1,000 nM of B3 vitamers added to mammalian cell lines mimicked the physiological response to supplementation in vivo (as measured by increase in NAD+ levels) (35). After 18 h at 37°C-in 5% CO₂, supernatants were collected from half of the wells with a volume of 200 μL per well and cytokine levels (IL-6, IL-1β, TNF-α, IFN-γ) were assessed using the Mesoscale Discovery System V-PLEX Plus Proinflammatory Panel 1 Assay (MSD, Rockville, MD). The remaining wells were incubated for an additional 30 h at 37°C-in 5% CO₂ at which time the CellTiter 96® AQueous One Solution Cell Proliferation Assay (MTS; Promega, Madison, WI) was administered per manufacturer directions and reduced formazan measured at 490 nm. Serum IL-6 was also assessed via V-Plex Plus Proinflammatory Panel 1 Assay (n = 20–22/group), measured across Experiment 1 and 2.

Colonic contents of stressor-exposed and non-stressed control mice were collected the morning following the last day of stressor exposure and snap frozen. Frozen samples were shipped to Metabolon (Durham, NC) for processing using the automated MicroLab STAR system (Hamilton Company). Recovery standards were added prior to the extraction process. Proteins were precipitated with methanol under vigorous shaking for 2 min, followed by centrifugation. The resulting extract was divided into five fractions: two for analysis by two separate reverse phase (RP)/UPLC-MS/MS methods with positive ion mode electrospray ionization (ESI), one for analysis by RP/UPLC-MS/MS with negative ion mode ESI, one for analysis by HILIC/UPLC-MS/MS with negative ion mode ESI, and one reserved for backup. Organic solvent was removed with a TurboVap (Zymark) and extracts were stored overnight under nitrogen prior to analysis. Recovery standards, internal standards, and quality control samples (including a pool of human plasma characterized by Metabolon to have known compounds), a pool of aliquots from every sample in this study, pure water, and solvent extract were used. Instrument variability was determined by calculating the median relative standard deviation for the recovery and internal standards. Process variability was determined by calculating the median relative standard deviation for all endogenous metabolites present in 100% of the pooled aliquots from every sample. Experimental samples were randomized across the platform run with QC samples spaced between every third sample injection.

Raw data was extracted, peak-identified and QC processed using Metabolon's hardware and software. Compounds were identified by comparison to a library of over 3,300 purified standards to recurrent unknown entities. Biochemical identifications were based on retention index within a narrow RI window of the proposed identification, accurate mass match to the library (±10 ppm), and the MS/MS forward and reverse scores between the experimental data and authentic standards. Peaks were quantified using area-under-the-curve. Values were normalized to account for variability across run days, and the intensity values were rescaled to set the median equal to 1.

Fecal samples were collected by placing the mice into a sterile cage and collecting fecal pellets. DNA was extracted from the fecal pellets (~10 mg) using a QIAamp Fast DNA Stool Mini Kit following manufacturer's instructions, with slight modifications. Briefly, stool was incubated for 45 min at 37°C in lysozyme buffer (22 mg/ml lysozyme, 20 mM TrisHCl, 2 mM EDTA, 1.2% Triton-x, pH 8.0), then bead-beat for 150 s with 0.1 mm zirconia beads. Samples were incubated at 95°C for 5 min with InhibitEX Buffer, then incubated at 70°C for 10 min with Proteinase K and Buffer AL. Following this step, the QIAamp Fast DNA Stool Mini Kit isolation protocol was followed, beginning with the ethanol step. DNA was quantified with the Qubit 2.0 Fluorometer (Life Technologies, Carlsbad, CA) using the dsDNA Broad Range Assay Kit. Samples were standardized to at least 5 ng/μl before being sent to the Molecular and Cellular Imaging Center (MCIC) in Wooster, OH for library preparation. Amplified PCR libraries were sequenced from both ends of the 250 nt region of the V4-V5 16S rRNA hypervariable region using an Illumina MiSeq. DADA2 and Quantitative Insights into Microbial Ecology [QIIME] 2.0 (36) were utilized for amplicon processing, quality control and downstream taxonomic assignment using the ribosomal RNA database SILVA (37).

DNA extracted for 16S rRNA gene analyses (above) was submitted for metagenomic sequencing for one sample in the GOS+PDX+SL-stressed group at the Genomics Shared Research facility at The Ohio State University. Prior to library generation, microbial DNA enrichment was performed using the NEBNext Microbiome DNA Enrichment kit (New England BioLabs, Ipswich, MA) per the manufacturers' protocols. Libraries were generated using the NExtera XT Library System in accordance with the manufacturer's instructions. Genomic DNA was sheared by sonication and fragments were end-repaired. Sequencing adapters were ligated, and library fragments were amplified with five cycles of PCR before solid-phase reversible immobilization size selection, library quantification, and validation. Libraries were sequenced on the Illumina HiSeq 4,000 platform. All raw reads were trimmed from both the 5′ and 3′ ends with Sickle, and then each sample was assembled individually using IDBA-ud using default parameters (38). All scaffolds ≥ 2 kb were binned using MetaBat2 (39). Genes were called using MetaProdigal (40) and annotated using Diamond (41) against the UniRef database (42).

Prior to statistical analyses, bacterial proportions were assigned to genus level taxonomy and then were normalized by log transformation (13, 43). Normalized bacterial proportions were analyzed using a two factor ANOVA, with stress (Unstressed vs. Stressed) x diet (Control vs. GOS+PDX+SL vs. SL) as between-subjects factors. The Benjamini–Hochberg false discovery rate (FDR) method was implemented to avoid type-1 error.

To determine the extent to which the colonic metabolome differed between stress conditions and diets, Random forest (RF) classification was completed using R statistical package. RF was performed on all metabolites using at least 1,000 trees with Matthews Correlation Coefficient (MCC) used as an assessment for classification efficacy (44, 45). Next, to understand which metabolites best discriminate between stress and diet status, Boruta feature selection algorithm was implemented (46). Boruta feature selection uses RF and iteratively compares importance of variables with importance of pseudo-random variables. Variables that have significantly worse importance than pseudo-random variables were consecutively dropped, thereby limiting chance of random significance and type-I error (44). Variables “confirmed” by Boruta selection were deemed available for further analysis. Metabolic pathway analysis was performed on Boruta-confirmed features using KEGG Metabolic reference pathway (47).

Splenocyte reactivity was analyzed using a 2-factor ANOVA with stress and diet (Experiment 1)/B vitamin (Experiment 3) as a between subjects factor among LPS treated cells only. For Experiment 3, splenocytes from each animal were evenly distributed among all B vitamin treatments and thus treated as a within factor. Dunnet's post-hoc analysis were utilized to compare each treatment (diet or B vitamin) to control group only.

To examine the potential for prebiotics to attenuate stress-induced immune activation, we fed mice (n = 56) a control diet, a diet supplemented with GOS, PDX, and SL, or a diet supplemented with SL alone for 2 weeks. Half of the mice in each diet treatment (n = 9–10/diet group), were exposed to the social disruption (SDR) stressor for 2 h a day for 6 days. Twenty-four hours post SDR, mice were sacrificed for tissue collection (Methods Summary; Supplementary Figure 1). Stressor-exposed mice fed the control diet exhibited elevated serum IL-6 and spleen mass compared to unstressed mice, together indicating stress-induced immune activation, consistent with previous reports (34, 48) (Figures 1A,B, p < 0.05). However, stress-induced increases in spleen mass and circulating IL-6 were offset in mice fed GOS+PDX+SL and SL diets (Figures 1A,B, p > 0.05, Tukey HSD).

In order to further investigate the role of prebiotics in mediating stress-induced immunoreactivity, spleen cells harvested from control- and prebiotic- fed mice were stimulated ex vivo with LPS. Splenocytes from stressor-exposed mice fed the control diet exhibited enhanced proliferation/viability in response to an LPS challenge compared to unstressed mice (p < 0.05, Figure 2A). However, this phenomenon was attenuated in mice fed GOS+PDX+SL or SL diets (p > 0.05, Figure 2A). Splenocytes from stressor-exposed mice fed the control diet also produced significantly higher levels of IFN-γ, IL-1β, IL-6, and TNF-α upon LPS stimulation (p < 0.05; Figures 2B–E), while diets containing GOS+PDX+SL or SL limited the effects of the stressor on LPS-induced cytokine production (Figures 2B–E). Together, these data highlight the potential of dietary oligosaccharides to mitigate the exacerbated inflammatory response due to exposure to stressful stimuli.

To determine whether the gastrointestinal microbiota underlies the protective effects of dietary oligosaccharides against stress-induced immunoreactivity, we set up a similar experiment to Experiment 1, with mice randomly distributed across stress and diet conditions (n = 9–10/stress/diet group; Supplementary Figure 1). At sacrifice, we collected total colon contents for downstream metabolomics (LC/MS/MS), microbial community analysis (16S rRNA gene sequencing), and metagenomics. This allowed for the coordinated measurement of microbial metabolites alongside microbial taxa and genes present in the colon at sacrifice.

Collapsed across diet conditions, random forest alongside Boruta feature selection (46) revealed 64 metabolites that significantly differed between stressed and non-stressed mice (Figure 3). Using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (47), the metabolites delineated into key metabolic groups and pathways that were affected by stress. In particular, amino acids, dipeptides and B vitamins were significantly lower in stressed animals (Figure 3). Meanwhile, lipids and nucleotides were higher in stressed animals (Figure 3). In particular, eight nucleotide and nucleoside derivatives were found to be significantly higher in stressed animals (pseudouridine, thymidine, cytidine, guanosine, deoxyguanosine, methylguansoine, inosine, and deoxyinosine; Figure 3). This was observed in parallel with a significant reduction of three amino acids (asparagine, tryptophan and leucine; Figure 3). Of these amino acids, tryptophan and downstream metabolites were most extensively affected by stress. This included significant reductions in 5-hydroxyindole acetate, quinolinate and indole acetate, together indicating stress-induced disruptions in gastrointestinal tryptophan metabolism. Lastly, random forest analysis revealed a robust effect of stress on the concentration of B vitamins and closely associated metabolites (vitamers). Notably, stress led to lower levels of thiamin (B1), nicotinamide and nicotinamide riboside (B3), pantothiene and pantothenate (B5) and pyridoxamine (B6) (Figure 3).

After determining the effects of stress on the colonic metabolome, machine learning analysis (RF+Boruta) revealed 51 metabolites that discriminated between mice fed the control diet and mice fed GOS+PDX+SL (Figure 4), while 15 metabolites differentiated mice fed control diet and mice fed SL (Supplementary Figure 2). Random forest with Boruta selection also identified 22 metabolites that differentiated mice fed GOS+PDX+SL from mice fed SL alone (data not shown). Forty-two of the 51 metabolites were higher in the GOS+PDX+SL diet compared to the control diet (Figure 4). Within amino acid metabolism, only 3 amino acids were higher in the control diet (xanthurenate, N-acetylmethionine sulfoxide and 4-guanidinbutanoate). Of particular significance, B vitamins and associated metabolites were increased in the colonic contents from prebiotic fed mice (Figure 4). Specifically, GOS+PDX+SL feeding augmented concentrations of B3 (nicotinamide riboside and nicotinate), B5 (pantothenate) and B6 (pyridoxal and pyridoxamine) vitamins while SL feeding only led to increased levels of B6 (pyridoxal) (RF Boruta confirmed; Figure 4, Supplementary Figures 2, 3A–D).

In continuing to identify potential mechanisms underlying diet-induced attenuation in immunoreactivity, we next investigated the gut microbiome, which is a primary mediator of fiber degradation and stress-induced immune responses. Moreover, because select bacteria are capable of B vitamin synthesis, we sought to understand whether the microbiome was a potential mediator through which dietary oligosaccharides mediate B vitamin metabolism. First, we sequenced the V4-V5 hypervariable region of the 16S rRNA gene, isolated and amplified from colon contents of mice exposed to stress and prebiotic diets (Experiment 2, Supplementary Figure 1). Non-metric Multidimensional Scaling (NMDS) based on Bray-Curtis dissimilarity metric revealed microbial communities significantly clustered by stress treatment (mrpp, p < 0.001; anosim, p < 0.004) and dietary conditions (mrpp, p < 0.001; anosim, p < 0.001) (Figure 5A). Notably, GOS+PDX+SL diet led to robust effects on the microbiome community composition that were largely orthogonal to the effects of stress (Figure 5, orange), (mrpp, p < 0.001). The SL diet alone had a minor effect on overall microbial community composition (Figure 5A).

Several differences in bacterial genera were evident in stressor-exposed mice, including a lower abundance of Lactobacillus spp., alongside higher abundances of Bacteroides spp. and other taxa including Butyricimonas spp. and Parabacteroides spp. (FDR p < 0.05; Figure 5B). Regardless of stress, animals fed GOS+PDX+SL displayed a robust elevation in Bifidobacterium (FDR p < 0.05; Figure 5C) concomitantly with lower abundance of Ruminococcus, Oscillospira, and Lactococcus (FDR p < 0.05, data not shown). Because the GOS+PDX+SL diet resulted in higher amounts of B vitamins and prevented stressor-induced increases in IL-6, we probed the relationships between colonic B vitamins and serum IL-6. In doing so, we found that relative levels of nicotinamide riboside, nicotinamide, pantothiene and pyridoxamine in colonic contents were inversely associated with circulating IL-6 (r = −0.49, −0.47, −0.41, −0.47, respectively; all p's < 0.05). Moreover, B vitamins were associated with microbiome composition, including a strong relationship between B vitamins (nicotinamide riboside, pyridoxamine, pyridoxal) and Bifidobacterium spp. (r = 0.58, 0.61, 0.59, respectively; all p's < 0.05).

In light of the robust and specific effects of GOS+PDX+SL diet on Bifidobacterium spp, we next sought to understand the genetic potential of this genus to synthesize B vitamers. To accomplish this, we performed metagenomic sequencing (9.69 Gbps) on colonic contents from one animal exposed to stress and provided the GOS+PDX+SL diet, with a particular focus on Bifidobacterium assembly. Reads were quality trimmed and assembled into contigs. Contigs were then binned using MetaBat2 (39) and annotated and taxonomically assigned using Diamond (41) against the UniRef database. We recovered one metagenome-assembled genome (MAG) 100% identical by amino acid identity of the ribosomal S3 protein to Bifidobacterium pseudolongum, the taxa most significantly upregulated by GOS+PDX+SL diet. This genome was 98.89% complete, with <1% contamination, and assembled into only five scaffolds indicating that this genome is high quality by the Genome Standards Consortium (49) (Supplementary Table 1). Metabolic reconstruction of the Bifidobacterium MAG highlighted genes for the de novo synthesis of Vitamin B6, including pyridoxal 5′ phosphate synthase. Like many other Bifidobacterium, the sequenced Bifidobacterium MAG from this study lacks 6-phosphofructokinase, which is one rate-limiting step in glycolysis, but alternatively has fructose-6-phosphoketolase to bypass this step (50) (Supplementary Figure 4). Notably, we detected increased metabolites for this pathway in mice fed the GOS+PDX+SL diet (e.g., erythrose 4-phosphate, sedoheptulose 7-phosphate, and ribulose-5-phosphate), which are precursors to pyridoxal production (Supplementary Figure 4). The GOS+PDX+SL diet also resulted in increases in sugar monomers, glycolysis intermediates, and TCA metabolites, together indicating increased metabolic potential for de novo B vitamin synthesis by commensal microbes.

To further unravel the inhibitory effects of the GOS+PDX+SL and SL diets on stress-induced immunomodulation, we tested whether B vitamers would have direct, inhibitory effects on splenocyte proliferation and immune reactivity to LPS challenge. To test this hypothesis, C57BL/6NCrl mice (stressed n = 6; unstressed n = 6) fed a normal chow (control) diet were exposed to the SDR stressor for 6 days or left undisturbed as a control as described above (Experiment 3, Supplementary Figure 1). Twenty-four hours after the final stressor exposure, spleens were removed from the mice and single cell suspensions cultured with 1,000 nM of B vitamers for 48 h during concurrent stimulation with LPS (1 μg/mL). We replicated our previous findings showing that stress led to enhanced splenocyte proliferation in response to LPS (stress main effect p < 0.05; Figures 6A,C). In a more novel finding, we observed that supplementation of splenocytes with 1,000 nM of the B3 vitamer nicotinamide riboside (B vitamer main effect p < 0.05; Figure 6A) as well as the B6 vitamers, pyridoxal and pyridoxamine (Figure 6C) reduced cell proliferation in response to an ex vivo LPS challenge (^*B vitamer main effect p < 0.05; Figure 6). In addition, pyridoxal (^#B vitamer x stress; p < 0.05) and nicotinamide riboside (Trend, B vitamer × stress; p = 0.08) attenuated stress-induced enhancement of splenocyte proliferation in response to the LPS challenge (Figures 6A,C). Next, we analyzed cytokine production from ex vivo stimulated splenocytes 18 h after B vitamin supplementation and LPS treatment. Supplementation of splenocytes with 1,000 nM of nicotinamide riboside, pyridoxal and pyridoxamine attenuated LPS-induced IL-6, IL-1β, and IFN-γ production compared to cytokine production from splenocytes not treated with B vitamers (B vitamer main effect p < 0.05; Figures 6B,D). Pyridoxamine was also effective at limiting stress-induced enhancement of IL-6 and IL-1β production from splenocytes after LPS stimulation (B vitamer x stress p < 0.05; Figure 6D). Pyridoxal was most effective of all B vitamins in suppressing cytokine production from LPS stimulated splenocytes. This was evidenced by a robust attenuation in inflammatory cytokine production (IL-6, TNF-α, IL-1β, IFNγ), in LPS-stimulated splenocytes supplemented with pyridoxal (B vitamer main effect p < 0.05; Figure 6D).

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.