Physical activity shapes the intestinal microbiome and immunity of healthy mice but has no protective effects against colitis in MUC2-/- mice

The interactions among humans, their environment, and the trillions of microbes residing within the human intestinal tract form a tripartite relationship that is fundamental to the overall health of the host. Disruptions in the delicate balance between the intestinal microbiota and their host immunity are implicated in various chronic diseases including inflammatory bowel disease (IBD). There is no known cure for IBD, therefore, novel therapeutics targeting prevention and symptoms management are of great interest. Recently, physical activity in healthy mice was shown to be protective against chemically-induced colitis, however the benefits of physical activity during or following disease onset is not known. In this study, we examine whether voluntary wheel running is protective against primary disease symptoms in a mucin 2 deficient (Muc2-/-) life-long model of murine colitis. We show that 6 weeks of wheel running in healthy C57BL/6 mice leads to distinct changes in fecal bacteriome, increased butyrate production, and modulation in colonic gene expression of various cytokines, suggesting an overall primed anti-inflammatory state. However, these physical activity-derived benefits are not present in Muc2-/- mice harboring a dysfunctional mucosal layer from birth, ultimately showing no improvements in clinical signs. We extrapolate from our findings that while physical activity in healthy individuals may be an important preventative measure against IBD, for those with a compromised intestinal mucosa, a commonality in IBD patients, these benefits are lost.


Introduction
Inflammatory bowel diseases (IBD) encompassing Crohn's disease (CD) and Ulcerative 39 colitis (UC) are idiopathic, relapsing chronic diseases characterized by chronic inflammation of 40 the gastrointestinal tract. While pathology varies between UC and CD, both burden patients with 41 common debilitating clinical symptoms such as diarrhea, rectal bleeding, abdominal pain, and 42 weight loss. The etiology of IBD is not known, however a combination of genetic, immunological, 43 and environmental factors is implicated in its development. Most recently, the contribution of the 44 intestinal microbiota in IBD pathogenesis has risen as an active area of research (1). For 45 example, IBD patients have reduced gut microbial diversity (2) and are more likely to have been 46 exposed to antibiotics in 2-5 years preceding their diagnosis (3). In animal models, mice 47 genetically predisposed to colitis (IL-10 -/-) are resistant to disease onset while kept under germ-48 free conditions, however clinical signs instigate immediately following exposure to microbes (4). 49 With incidence of IBD and its burdens rising globally (5), there is an increasing demand 50 for novel therapeutics. Physical activity (PA) has been proposed as both a primary and an 51 adjunct therapy for prevention and treatment of various chronic diseases due to its well 52 documented ability to ameliorate low-grade systemic inflammation (6). Most recently, IBD has 53 been marked as a potential new candidate (7) to yield benefits from regular PA. Studies of PA in 54 rodents have shown attenuated clinical signs of chemically-induced colitis (8)(9)(10)(11)) that appear to 55 be dependent on the colitis model and type of PA. These studies, however, only assess the role 56 of PA as a preventive measure leading up to induction of acute colitis via a chemical toxin. As 57 so, the potential benefits of PA succeeding or during disease onset is not known. In this study 58 we aimed to address this knowledge gap by utilizing the mucin 2 knock-out (Muc2 -/-) mouse 59 model of chronic colitis. 60 The human intestinal tract is continuously exposed to the trillions of microbes residing 61 within the mucosal layer of the lumen. Under homeostatic conditions, these microbes are 62 tolerated by the host as they provide essential functions such as digestion of complex 63 carbohydrates, protection against enteric pathogens, and production of beneficial short-chain 64 fatty acids (SCFA), to name a few. Separating the luminal microbes from intestinal epithelial 65 cells (IEC) is a mucus bilayer largely composed of the highly glycosylated protein MUC2. In the 66 colon, the loosely structured outer mucus layer allows for colonization of microbes in a nutrient-67 rich environment, while the dense inner layer segregates them from the IEC (12). Inflamed 68 intestinal tissues of UC patients commonly display structural defects or thinning of this mucus 69 layer (13), leading to excessive exposure of microbial antigens to the host cells, prompting a 70 chronic state of inflammation and apoptosis leading to further loss of IEC integrity and thus 71 further exposure and injury. Muc2 -/mice or those with missense mutations impairing the release 72 of MUC2, are born with an underlying predisposition to intestinal inflammation that show rapid 73 progression of colitis (14). Muc2 -/mice generally display early clinical signs of colitis following 74 weaning (~ 1 month) and microscopic tissue pathology as early as two months of age, indicating 75 a moderate-level colitis, reaching high-severity by 4 months (15). 76 We hypothesized that introduction of Muc2 -/mice to voluntary wheel running (VWR) 77 immediately following weaning would reduce the severity and delay the onset of clinical signs of 78 were programmed to record the total number of revolutions at 1 hr intervals for the duration of 118 the experiment. Body weights, food consumption, and water intake were measured weekly at 119 approximately the same time during the light cycle. Food weight measurements consisted of 120 subtracting the week's remaining pellets on the cage lids and bottoms from that week's starting 121 weight. 122 Tissue collection -For fecal sample collection, mice were kept briefly in isolation in sterile and 123 DNAzap-treated containers until defecation. Collected fecal pellets, which were used for 124 microbiome surveying, were immediately snap-frozen in liquid nitrogen then stored in - 80 °C 125 until further analyses. Fecal samples were collected on day 1 immediately following assignment 126 to individual cages, and again on the final experiment day immediately preceding tissue 127 collection. Animals were euthanized by cervical dislocation while under deep isoflurane 128 anesthesia. The cecum was isolated, its content removed, and tissue frozen in liquid nitrogen 129 for further analyses of SCFA composition. Colon tissues were collected as follows: starting from 130 distal end, two consecutive ~1.5 cm sections were collected with the most distal section being 131 fixed in 10% neutral buffered formalin for histopathology and the proximal section was stored in 132 RNAlater (Thermo Fisher Scientific) for use in cytokine gene expression assays. All frozen 133 samples were then stored at -80 °C until further use. 134 Muc2 -/animals  135 was assessed based on an in-house clinical signs scoring system, and represented by a 136 variable we henceforth call "disease score". Briefly, each animal was graded weekly based on 137 the following: observed behavior from a distance, stool/rectal bleeding, stool consistency, weight 138 loss, and hydration, with each variable being assigned a score of 0-4. Humane endpoint was set 139 as a total cumulative score of ≥12, rectal prolapse, or a weight loss of >20% Stitched images were imported into ImageJ-version 1.51r (21) for scoring. Disease severity in 149 colonic cross sections from the Muc2 -/animals were assessed using a previously described 150

Clinical and Histopathological Scoring -Disease progression and severity in
scoring system (22). In brief, a total score was calculated for each mouse using the following 151 criteria 152 1. Edema, as compared to a healthy WT control: 0 = no change; 1 = mild (<10%); 153 2 = moderate (10-40%); 3 = profound (>40%) 154 2. Epithelial hyperplasia, average height of crypts as a percentage above the height of a 155 healthy control where 0 = no change; 1 = 1-50%; 2 = 51-100%; 3 = >100% 156 3. Epithelial integrity, shedding and shape of the epithelial layer as compared to healthy  157  control where: 0 = no change; 1 = <10 epithelial cells shedding per lesion; 2 = 11-20 epithelial  158  cells shedding per lesion; 3 = epithelial ulceration; 4 = epithelial ulceration with severe crypt  159  destruction  160 4. Cell infiltration, presence of immune cells in submucosa: 0 = none; 1= mild (2-43); 161 2 = moderate ; 3= severe . 162 The resulting variable, henceforth called "histopathological score", has a maximum value of 13. 163 Reverse Transcriptase-qPCR -To identify the potential immunological pathways involved in 164 PA-derived protection, we examined the gene expression of several key immune markers 165 commonly associated with colitis. The mRNA gene expression for tumor-necrosis factor alpha 166 (TNFα), interferon-gamma (IFNγ), resistin-like molecule beta (Relm-β), regenerating islet-167 derived protein 3 (RegIII-γ), transforming growth factor beta (TGF-β), chemokine C-X-C motif 168 ligand 9 (Cxcl9), and claudin 10 (Cldn10) were measured in colon tissues.  (23)). They not only serve as a primary 185 food source for the colonocytes, but have immunogenic properties that, in concert with the host 186 immunity, are integral in maintaining gut homeostasis. We previously showed that in healthy 187 humans, cardiorespiratory fitness was positively correlated with fecal butyrate (16), a SCFA with 188 known anti-inflammatory properties in the gut (24). We therefore hypothesized that SCFA 189 profiles of VWR mice would differ from SED, favoring the production of beneficial butyrate that 190 may be involved in protection against colitis. We therefore analyzed SCFA (acetic, propionic, 191 heptanoic, valeric, caproic, and butyric acids) in cecal tissues by gas chromatography (GC) as 192 described previously in (25 Amplicons were purified using AMPure XP beads and adapters and dual-index barcodes 213 (Nextera XT) were attached to the amplicons to facilitate multiplex sequencing. Following a 214 secondary clean-up step, libraries were quality controlled on an Experion automated 215 electrophoresis system (Bio-Rad), and sent to The Applied Genomic Core (TAGC) facility at the 216 University of Alberta (Edmonton, Canada) where they were normalized using fluorometric 217 method (Qubit, Thermo Fisher Scientific) and sequenced using the Illumina MiSeq platform with 218 a V3 reagent kits allowing for 2 x 300 bp cycles. 219 Bioinformatics -All bioinformatics processes were performed using a combination of R 220 statistical software (31) and the QIIME 2 platform (32) using the various build-in plugins 221 described below. Demultiplexed sequences were obtained from the sequencing facility and 222 primers removed reads using cutadapt (33). Sequences then underwent quality-filtering, 223 dereplication, denoising, merging, and chimera removal using DADA2 (34). The output of this 224 process is a feature table of amplicon sequence variants (ASV) that is a higher resolution 225 analogue of traditional OTU tables. To aid in removal of non-specific host contaminants, a 226 positive filter was applied to all reads using the latest available Greengenes (13_8) (35) 227 database (clustered at 88% identity). All ASVs were searched against the reference reads 228 using VSEARCH (36) and any that did not match the reference sequences at a minimum of 229 70% identity similarity at 70% alignment were discarded. For analyses encompassing 230 phylogenetic information, a phylogenetic tree was constructed using a SATé-enabled 231 phylogenetic placement (SEPP) technique as implemented in the q2-fragment-insertion plugin 232 (37) using a backbone tree build based on the SILVA (128) database (38). Taxonomic  233 classification of the ASVs were carried using IDTAXA (39). It has been proposed that the 234 functional repertoire of the gut microbiota is more sensitive to perturbation than taxonomic 235 changes, and therefore may be crucial in identifying underlying physiological signals (40). To 236 predict the functional potential and phenotype of the microbiome, we used BugBase (41) which 237 utilizes PICRUSt's (42) extended ancestral-state reconstruction algorithm for metagenome 238 composition prediction. As these tools require sequences to be classified against the 239 Greengenes taxonomy assignments, we used VSEARCH to pick closed-reference OTUs from 240 our denoised feature table at 97% similarity threshold against the 99% identity clustered 241 Greengenes database. 242 Statistical Analyses -All statistical analyses were performed using R version 3.5.1 unless 243 stated otherwise. During the 3rd week of the experiment, the VWR animals were unintentionally 244 exposed to 3 days of irregular light-dark cycles as a result of an electrical malfunction with the 245 lighting in the animal room. While the exact nature of this disruption is not known, the wheel 246 running data during this period suggests a period of reduced activity. The issue was resolved by 247 the 3rd day and the animals did not display any signs of stress or irregular behavior; we 248 therefore consider this to be of minimal impact to the experiment. However, as a precaution, we 249 chose to analyze the data as a 4 x 1 (groups) factorial design rather than 2 x 2 (activity x 250 genotype) as we could not definitively eliminate the possibility that wheel running in this group 251 was impacted by the brief interruption. 252 Wheel running -To determine whether WT and Muc2 -/ran similar distances throughout the 253 experiment, we first analyzed total weekly distances (km) run by each group across the 6 weeks 254 time using linear mixed-effects regression (LMER) using the lme4 package with individual 255 animals set as the random effect and groups as the fixed effects. Homoscedasticity and linearity 256 of the models were assessed using diagnostic plots of the residuals. 257 Body weights and food/water intake -To monitor overall behavioral changes of mice as a 258 result of PA between WT and Muc2 -/mice, we examined weekly body weights, food and water 259 intake across the 6 weeks. To account for natural differences in starting body weights, total 260 weight gained relative to starting body weights was calculated each week. Body weight, food 261 and water intake across the 6 weeks were each assessed separately using a repeated 262 measures LMER with time coded as a random effect and groups as a fixed effect. A Tukey HSD 263 post-hoc test with the Benjamini-Hochberg (BH) P adjustment method was used when an 264 overall significance (set as P<0.05) in the models were detected. 265 Clinical and Histopathological Scoring -We used a cumulative link model (CLM) with a logit 266 link to evaluate whether the disease score, and separately, the histopathological score differed 267 among treatment groups. This proportional odds type test is more appropriate for ordinal data 268 than classic linear regressions. For clinical scores, the model included time and groups as the 269 fixed effects, and individual animal ID as the random effect. For histopathological score, the total 270 average score of the MSED and MVWR groups were separately analyzed using the same 271 method, but without the time random effect. We implemented the analyses using the ordinal R 272 package. 273 Colon mRNA gene expression -To test whether the expression of colonic mRNA genes 274 differed across groups, we first explored the overall abundance of all surveyed genes 275 simultaneously using an ordination method. then be projected onto a PCA biplot. We visualized this using the Emperor interactive graphic 299 tool (45). To reveal possible group differences, a PERMANOVA (46) test was conducted on all 300 groups across time. Pairwise testing was then followed using a Kruskal-Wallis test with a BH 301 adjustment to control for false discovery rate (FDR). 302 The overall within-sample diversity (α diversity) for each sample was estimated based on the 303 species richness, Simpsons, and Shannon indices using the DivNet package (47). For each 304 group, the difference between a sample's week 6 and week 0 diversity score was calculated 305 and used to determine whether those changes differed from zero (Wilcoxon test) as well as 306 other groups (ANOVA). 307 Differential abundance testing of individual taxa was performed using the CornCob package 308 (48). 309 BugBase was used to determine high-level phenotypes of bacterial communities based on the 310 following default traits: Gram negative vs. Gram positive, biofilm forming, mobile element 311 containing, oxidative stress tolerance, pathogenic potential, and oxygen utilizing. Pre-and post-312 treatment differences in relative abundances of these elements were tested in each group using 313 a Kruskal-Wallis test with Benjamini-Hochberg adjustment of P values to control FDR. 314 315

316
Wheel running -For unknown reasons, one animal from each group did not run on the wheels 317 and so were excluded from further analyses. The WT group ran an average (SD) of 46.6 (18.4) 318 km in total throughout the 6 weeks, while the Muc2 -/animals ran slightly less at 40.7 km (21.5) 319 which correspond to ~ 1.3 and 1.1 km/day, respectively. While the WT showed a general trend 320 towards more wheel running, the differences were not statistically significant ( Figure 1A) likely 321 due to the highly variable nature of running data. 322 Body Weights, food and water intake -Weight gain was not significantly different across 323 activity levels, however as expected Muc2 -/mice gained less weight throughout the 6 weeks 324 ( Figure 1B). The mean (± SE) total weight gain of each group was: SED 33.   (Dev:  363 14.83, P<0.01) and the univariate tests showed significant differences in acetate, propionate, 364 butyrate, and valerate across groups. The results of the post-hoc analyses on these SCFAs and 365 total SCFA are shown in Figure 3. Total SCFA concentration was significantly higher in VWR 366 mice than all other groups, while SED mice had similar total SCFA to both Muc2 -/groups. VWR 367 mice also had significantly higher total acetate and butyrate than all the other groups and higher 368 propionate than SED. Overall, the major difference between Muc2 -/and WT animals was the 369 significantly reduced levels of butyrate in Muc2 -/mice and inversely, higher levels of propionate. 370 Valerate, caproate, and heptanoate were similar across all groups. In terms of relative 371 abundances, the main differences between Muc2 -/and WT were the higher propionate and 372 lower butyrate proportions in Muc2 -/animals. Importantly, the proportion of butyrate in VWR 373 mice (~12 %) was significantly higher than those in SED (~7.9 %). 374 Bacterial community analysis. Beta diversity: The PERMANOVA test showed a significant 408 difference between Muc2 -/and WT animals corresponding to clear clustering observed between 409 these groups on the PCA plots ( Figure 5A). Importantly however, there was a significant 410 distance between SED and VWR animal clusters prior to treatment assignment. This fact 411 strongly suggests the presence of a batch effect in our experiment which is likely explained by 412 the fact that the VWR animals were purchased at different times compared to the other groups 413 and their microbiome sequenced separately. As batch-effects are a well-known issue in short-414 read sequencing experiments (49), differences across groups are then likely confounded by 415 this. Therefore, to mitigate this effect, in all subsequent analyses, changes in microbiome are 416 either only compared within the same group across time, or the change within each group is 417 compared to changes in other groups. Pairwise analysis of each group comparing their week 0 418 to week 6 profiles showed changes in overall microbiome variation in all animals across time. In 419 Muc2 -/animals, these changes were non-uniform and did not follow a predictable pattern 420 ( Figure 5C)

Discussion
Muc2 -/mice vastly differ from WT in their colonic cytokine, SCFA, and microbial profiles Muc2 -/mice displayed clinical and histological symptoms of moderate colitis corresponding to the expected severity of this model at 11 weeks of age in our facilities. The colonic gene expression of inflammatory cytokine TNF-α, and the mucosal defense factor RELM-β, as well as antimicrobial peptide RegIII-γ were upregulated in Muc2 -/animals, as observed previously (15). Notably, the anti-inflammatory cytokine IL-10 was overexpressed in Muc2 -/compared to C57BL/6 WT. While in a healthy state, the expression of IL-10 may be associated with increased tolerance to inflammatory events, in Muc2 -/animals, this upregulation is essential in the host's efforts at suppressing the excessive inflammation resulting from continuous exposure to bacterial ligands. Indeed, Muc2 -/-+ IL-10 -/double knock-out mice show highly exacerbated colitis clinical signs (50) compared to deletion of either genes separately. The increase in IL-10 also has been previously observed in chemical models of colitis (9,11). We further detected significant overexpression of CXCL9 in Muc2 -/animals. CXCL9 is a chemokine involved in regulating leukocyte trafficking, likely in response to exposure of bacterial ligands to host cells. CXCL9 overexpression also has been reported in IBD patients (51). Overall, the cytokine profile of Muc2 -/animals reflect those expected in human IBD.
Muc2 -/mice born without a normal mucus layer house drastically less diverse and different bacterial communities than WT mice, as evident by the clear clustering of this group from WT animals in our PCA plots. Similar to patterns seen in IBD patients (52), or chemicallyinduced murine colitis (53,54), Muc2 -/animals had an overall reduced α-diversity compared to their WT counterparts. The dominant taxa in WT mice were generally of the Bacteroides genus, Clostridiales order, and Lachnospiraceae family, while Muc2 -/animals were dominated by members of the Muribaculaceae family (formerly known as S24-7 (55)) and Akkermansia genus of the Verrucomicrobia phyla. While our taxonomic classifier was unable to confidently differentiate between the 2 sole species (Akkermansia muciniphila and Akkermansia glycaniphila) within this genus, it is reasonable to assume that the observed taxa were A. muciniphila as A. glycanphilia has, to date, only been isolated from python feces (56). A. muciniphila is perhaps the most surprising finding in this group as this species is known -and named-for its ability to degrade mucin, and is broadly considered as a beneficial bacterium in a variety of chronic diseases including IBD (57)(58)(59). The broader implications of this finding are beyond the scope of the present study. However, it does warrant the reassessment of the characterization of A. muciniphila as a mucin loving species to one that thrives in the absence of mucin. The bacterial phenotypic traits of Muc2 -/animals were predicted to be higher in abundances of Gram-positive, aerobic, and biofilm forming groups compared to WT mice. Lastly, the cecal SCFA of Muc2 -/mice were composed of significantly less butyrate and higher propionate concentrations compared to SED animals. The increased propionate levels in these animals is likely associated with the high abundances of A. muciniphila, a prominent propionate producer (60,61). Overall, we found the Muc2 -/model of colitis to capture many components of human IBD, especially those with impaired mucosal integrity.

Wheel running in Muc2 -/mice does not reduce the severity of chronic colitis
Contrary to our primary hypothesis, we found that 6 weeks of wheel running in Muc2 -/mice did not improve the severity of clinical signs, histopathological scores, colonic expression of inflammatory cytokines, or abundances of cecal SCFAs, and did not alter the gut microbial composition in a consistent manner. These findings contrast others that show protective effects of VWR or forced treadmill running in chemically-induced models of colitis (8,10,11,62). The fundamental difference between those studies and ours is in the model of colitis used. Previously, VWR was initiated in healthy animals prior to disease induction with chemical toxins, whereas in our study, wheel running is imposed over an existing disease state as a therapeutic intervention. This would suggest that PA prior to disease onset primes various components of intestinal health, enhancing its tolerance to injury. The effects of PA following disease-onset on the other hand, are either abolished or are overwhelmed by stronger disease signaling. It is also possible that the physiological benefits of PA depend on the presence of a healthy mucosal layer. This is supported by our findings that wheel running in WT but not Muc2 -/animals leads to significantly lower levels of pro-inflammatory colonic cytokines, increased anti-inflammatory IL-10, and increased levels of beneficial SCFAs. Given that UC patients typically have defective and thinning colonic mucosal layers, this would suggest that exercise prescription in these populations may have limited direct benefits on their intestinal health. However, the welldocumented benefits of exercise are instituted across various other sites and systems of the body, which may indirectly result in improving primary and secondary disease symptoms through other pathways not accounted for in this experiment. For example, PA as a primary intervention has been associated with improved quality of life in IBD patients (63) and inversely correlated with loss of bone mass density, a common risk factor in this population (64,65).

VWR significantly attenuates pro-inflammatory, and upregulates anti-inflammatory cytokines in WT mice
Compared to SED animals, the wheel running mice showed lower levels of inflammatory cytokines TNF-α, IFN-γ, and TGF-β, all of which have been implicated in IBD (66). TNF-α is perhaps the most studied cytokine in relation to IBD as it plays a crucial role in innate and adaptive immunity and is directly involved in apoptotic processes in the intestines (67). It is found in significantly higher abundances in IBD patients (68) as well as murine colitis (69), making its regulation an obvious target for disease management. In fact, TNF-α inhibition using monoclonal antibodies is the most common target of biological therapies for moderate to severe IBD. The role of IFN-γ in colitis pathogenesis is less consistent across the literature, however its overproduction has been shown in CD (70,71) and UC patients (72). In DSS-induced colitis models, neutralization antibodies against IFN-γ significantly reduced disease severity (73), while IFN-γ -/mice were completely protected from disease clinical signs (74). Anti-IFN-γ antibody treatments in human IBD are less effective however, with their efficacy dependent on baseline C-reactive protein levels (75), highlighting the need for treatment personalization. TGF-β is a pleiotropic cytokine that is ubiquitously produced by many cells and is involved in various immune functions including both anti-and pro-inflammatory actions. These include suppression of immune responses through recruitment of Tregs which in turn produce IL-10, but TGF-β can also elicit potent Th17 responses to combat extracellular bacteria (76). TGF-β is found in higher concentrations in intestines of IBD patients (77,78), due to increased exposure of microbial ligands to host epithelial cells. Inversely, the attenuated levels of this cytokine in our VWR animals then may reflect a decrease in bacterial antigen exposure to the IEC suggesting reduced levels of host-microbe interactions in the mucosa. Alternatively, reduced TGF-β could also indicate reduced Treg activity in VWR mice, however, the increase in Treg derived IL-10 in these animals does not support this notion. IL-10 is an anti-inflammatory cytokine ubiquitously secreted by Tregs and is the primary driver of immunosuppressant actions in the intestines. Polymorphism in IL-10 promoters have been linked to IBD, making IL-10 supplementation a potential target for IBD therapy, however, clinical studies of IL-10 therapy to date have not been significantly effective (79). The significant increase in IL-10 in VWR mice suggests higher Treg activity which is associated with reduced inflammation. This is in agreement with others who showed a significant increase in murine intestinal IL-10 following treadmill running or swimming (80,81). However, it is unclear whether this reflects a beneficial increase in anti-inflammatory events, or simply an adaptive response to changes in the microbial composition. Gram-negative bacteria preferentially stimulate IL-10 production and are associated with higher virulence due to increases in abundance of lipopolysaccharides bound to their cell walls (82). The higher expression of IL-10 in VWR animals then is likely correlated with increased abundance of Gramnegative bacteria observed in these mice. Further investigations are needed to determine the consequence of these changes. Taken together, the reduction of these pro-inflammatory cytokines and increase in anti-inflammatory IL-10 in VWR animals suggests a primed antiinflammatory state in healthy WT but not diseased intestines, marking them as potentially important targets for prevention and reemission maintenance therapy.

VWR significantly augments SCFAs content in WT but not Muc2 -/mice
SCFAs are metabolic by-products of bacterial fermentation of dietary fibers in the colon and are involved in various physiological processes of the host. Aberrant intestinal SCFAs content has been implicated in various diseases such as irritable bowel syndrome, cardiovascular disease, certain cancer types, and IBD (83)(84)(85). The most abundant of these, acetate, propionate, and butyrate, which make up over >95% of SCFA in humans (23), are markedly decreased in IBD patients (86), while their exogenous delivery can reduce inflammation via inhibition of TNF-α release from neutrophils (84,87). Overall, increases in these SCFAs, especially butyrate, appear to positively influence IBD (88). We found an overall higher abundance of total cecal SCFAs, acetate, butyrate, and propionate in response to wheel running in WT but not Muc2 -/animals. This is in accordance with others showing higher butyrate concentrations in wheel running in rats (89), following exercise training in lean humans (30), and elite athletes (90). We've also previously observed a positive association between higher butyrate levels and VO2peak in healthy humans (16). The increase in these SCFAs may simply reflect higher energy demands of colonocytes which utilize SCFAs as their primary energy substrate. Interestingly, when we analyzed SCFAs content in relative abundances, we saw a significant increase in relative abundance of butyrate, but not acetate, or propionate. This suggests a preference in VWR animals for production of butyrate and its accompanying antiinflammatory properties. These findings further support the patterns of anti-inflammatory priming we observe in these animals, contributing to an overall healthier intestinal environment following physical activity. The mechanisms behind PA-induced changes in SCFA are not known, however, given that SCFA are primarily produced by the intestinal microbiota, it is highly likely that changes in SCFA are linked to the observed changes in the microbiome. SCFA affects microbiota dynamics as they are directly involved in chemical balance and pH regulation of the intestines (91) and in turn the microbiota can also affect SCFA production and use, establishing a bidirectional affiliation.
Wheel running has limited but significant effects on the intestinal bacterial composition of WT but not Muc2 -/mice In this study, neither time nor wheel running had any effect on any alpha diversity metrics measured. The effects of PA on alpha diversity is not consistent within the literature. For example, the findings here are in contrast to our own previous observations in healthy humans that showed a significant correlation between alpha diversity and cardiorespiratory fitness (16). Others have also reported that elite athletes have higher alpha diversity than sedentary controls (92), or that exercise training in mice leads to increased Shannon diversity (93,94). However, in agreement with the current experiment, PA has been shown to have no effect on alpha diversity in mice (29,95,96), rats (97), and humans (30,98). The reasons for these discrepancies are not clear, though multiple factors such as differences in animal vendors and facilities, DNA extraction methods and sequencing, bioinformatics analysis, and statistical testing methods, are likely involved. Additionally, one particularly important consideration in comparing animal models of PA is the total volume of activity performed. For example, across the aforementioned studies utilizing wheel running in mice, we noticed a wide range (~ 2.5-10 km) of average daily running distances reported. Different volumes of PA are likely to elicit different physiological responses which can extend to the microbiome.
Comparisons of the across-samples diversity (beta diversity) in Muc2 -/animals showed no patterns of change as a function of time or wheel running. In WT animals, however, the change in Aitchison distances between week 6 and 0 were significantly different between VWR and SED groups, indicating both time and wheel running as important factors in the observed shift in community composition. The magnitude of this change across time between the groups however was not different.
Univariate analyses of individual taxa in Muc2 -/animals also showed no significant changes in any microbial clades across either groups. A lack of significant change in these animals suggests that the presence of a healthy mucosal layer is required to mediate PA-induced changes in community composition in the colon. In WT animals, the relative abundances of over 20 taxa in each group were significantly different by week 6 (Figure 7), with only four of those taxa common to both VWR and SED groups. The changing taxa in either group belonged primarily to the Ruminococcaceae and Lachnospiraceae families, with some species increasing while others decreased. Notably, In SED animals, 5 species from the Bacteroides genus were decreased by week 6, while no changes in this genus were detected in VWR animals. Species within the Bacteroides genus are Gram-negative, obligate anaerobes that are among the most abundant found within the mammalian intestine and carry important functions such breaking down of complex glycans, refining the gut environment by reducing intracellular oxygen levels, and preventing the colonization of opportunistic pathogens (99). Bacteroides spp. are also one of the primary propionate producers in the mammalian gut (100), therefore, the observed reduction in members of this group in SED animals may in part explain their lower propionate levels compared to VWR mice. Given that we were unable to classify these ASVs to the species level, it is difficult to speculate further on the biological implications of these observations. Perhaps the most studied species in this group, Bacteroides fragilis, has been shown to be protective against DSS colitis in mice by stimulating IL-10 expression (101,102). Given that we also observed significantly lower IL-10 expression in SED mice, it is reasonable to speculate that at least one of the unclassified Bacteroides ASVs in this group is B. fragilis and linked to the relatively lower expression of this protective cytokine in those animals.
Analyzing the bacterial consortia based on their predicted phenotypic traits revealed additional information regarding the effect of wheel running on the overall community. Following wheel running, WT but not Muc2 -/mice, had significant reduction (~ 14%) in total abundance of Grampositive bacteria. This is supported by the observed decreases in several members of the Gram-positive Ruminococcaceae in these mice, as well as the attenuated expression of RegIIIγ, an antimicrobial peptide that specifically targets the surface peptidoglycan layer of Grampositive bacteria. The implications of this phenotypic shift in microbiota of healthy individuals is not known, but may provide a clue for understanding the adaptations of the intestinal environment to the physiological stresses of PA. Furthermore, mirroring the shift in Grampositive phenotype was the decreased relative abundances of bacteria containing mobile elements. These refer to microevolutionary processes such as transposons i.e. segments of DNA with the ability to move locations within the genome, and bacterial plasmids which are involved in horizontal gene transfer. These events are typically associated with sharing of virulence factors between bacterial cells and increased resistance to antibiotics. The higher abundances of mobile elements in bacteria from these mice is likely not indicative of antibioticresistance but rather associated with higher abundances of Gram-negative bacteria representing more mobile-elements. The results of these predictions should be interpreted with caution however, as these mobile elements can rapidly become population specific within an individual thus precluding inference across similar experimental groups (103).

Summary
In contrast to our hypothesis we found that 6 weeks of wheel running did not ameliorate any clinical signs of colitis in Muc2 -/animals, nor did it influence any components of the intestinal environment such as expression of various cytokines and production of SCFA. Wheel running in healthy WT C57BL/6 mice on the other hand, imposed various physiological effects on the gut, including downregulation of pro-inflammatory and upregulation of anti-inflammatory cytokine gene expression, and increased concentration of total SCFAs including butyrate, acetate, and propionate. Wheel running further led to a shift in bacterial community structure corresponding to higher abundances of Gram-negative bacteria. As these physiological changes have been associated with protection against chronic inflammatory diseases in humans such as IBD, we conclude that PA prior to disease onset can prime the intestines, enhancing their tolerance to inflammation. These benefits however are lost when PA is imposed in the absence of a healthy mucosal layer. Overall, the findings here suggest that PA in healthy individuals may be an important preventative medicine against intestinal diseases such as IBD. RegIII-γ-R GGCATCTTTCTTGGCAACTTC 50. 5 21 Supplementary Material 6. List of RT-qPCR primers used. Tm refers to the primer melting temperature. Length (bp) refers to primer length in base pairs. Primer names ending with '-F' or '-R' refer to forward and reverse primers, respectively.