Opposing diet, microbiome and metabolite mechanisms regulate inflammatory bowel disease in a genetically susceptible host

Inflammatory bowel diseases (IBDs) are chronic conditions characterized by periods of spontaneous intestinal inflammation and are increasing in industrialized populations. Combined with host genetics, diet and gut bacteria are thought to contribute prominently to IBDs, but mechanisms are still emerging. In mice lacking the IBD-associated cytokine, Interleukin-10, we show that low dietary fiber promotes bacterial erosion of colonic mucus, leading to lethal colitis. A fiber-free exclusive enteral nutrition diet also induces mucus erosion but inhibits inflammation by simultaneously increasing an anti-inflammatory bacterial metabolite, isobutyrate. Diet-induced inflammation is driven by Th1 immune responses, which increase in the presence of mucin-degrading bacteria and are preceded by expansion of natural killer cells and altered immunoglobulin-A coating of some bacteria. Inflammation occurs first in intestinal regions with thinner mucus. Our work underscores the importance of focusing on microbial functions—not taxa—contributing to IBDs and some diet-mediated functions block those that promote disease.

100% survival when maintained for 129 dpw (150 days total) as did wild-type mice fed either diet 161 (Fig. 1I). Weight loss in pups weaned to the FF diet corresponded with increasing fecal Lcn2 that 162 occurred around the same time (~40 dpw) of declining weight, further suggesting that weight loss 163 is associated with increasing intestinal inflammation (Fig. 1J). Interestingly, Il10 -/mice with a 164 specific pathogen free (SPF) microbiota and fed the FF diet did not lose weight as severely as those 165 colonized with SM14 (Fig. 1H). Compared to SM14-colonized mice fed the FF diet, the SPF mice 166 showed lower inflammation, measured by Lcn2, at 79 dpw (Fig. 1K), and less histological damage 167 (Extended Data Fig. 1P), revealing that a more complex, murine microbiota does not promote 168 the same level of disease in this model, a point that is addressed in more detail below. Interestingly, 169 SPF mice fed the FF diet did exhibit reduced mucus thickness (Fig. 1L), although not as severe as 170 SM14-colonized mice fed the FF diet, implying that it contains mucin-degrading bacteria capable 171 of eroding this layer in a diet-dependent fashion but other factors may offset inflammation 172 development. 173 The FR and FF diets differ substantially in several aspects of their macronutrient 174 composition beyond fiber (Supplemental Table 1). To more directly test the contribution of fiber, 175 we created modified versions of the FF diet in which 7.5% of the carbohydrate it contains (glucose) 176 was removed and replaced with an equivalent amount of different, food grade purified fibers from 177 oat, wheat or apple. A high sugar diet has been shown to promote inflammation, including in Il10 -7 /mice [24]. As a control for reducing the sugar concentration in our fiber supplemented diets, we 179 created a diet that contained the same amount of highly digestible starch, effectively exchanging 180 7.5% of free glucose for a polymeric form of the same sugar that should still be available to the 181 host via upper GI digestion. Our data reveal that the presence of 7.5% fiber from either of the three 182 sources, but not digestible starch, reduces inflammation measured by Lcn2 (Fig. 1M) as well as 183 weight loss and histopathology (Extended Data Fig. 2A, B). Even when all of the glucose (44%) 184 was replaced with digestible starch, adult mice colonized with the SM14 developed disease (Fig. 185 1M, Extended Data Fig. 2A, B). Bacteroides ovatus, a proficient degrader of both the pectic and 186 hemicellulosic polysaccharides expected to be present in the supplemental fibers used [25], was 187 one of the major responders to the fibers used in these experiments, increasing between 2-3 fold 188 in relative abundance in the community and this increased proliferation generally occurred at the 189 expense of mucin-degrading species like Akkermansia muciniphila or Bacteroides caccae 190 (Extended Data Fig. 2C-G). 191 To determine if restoring dietary fiber to mice that had already been fed the disease-192 promoting FF diet is capable of blocking inflammation, we returned colonized adult mice, which 193 ordinarily experience ~89% lethality by 60d, to the FR diet at either 30d or 40d (16 or 26 days 194 after being switched to FF). Both groups of mice that were returned to a high-fiber diet exhibited 195 no lethal weight loss (Extended Data Fig. 2H) and Lcn2 levels and histology at 60d were 196 significantly lower than mice maintained on the FF diet (Fig. 1N, O). Interestingly, fecal Lcn2 197 measurements over time showed that both groups of mice experienced a peak of inflammation 198 after fiber had been restored, which then began declining, indicating that low fiber induces 199 disruption to the host-microbe homeostasis in which inflammation is a lagging effect that can 200 eventually be reset (Fig. 1N).

202
The status of the mucus layer is a critical determinant of inflammation development 203 To more directly evaluate the disease-promoting role of mucin-degrading bacteria, we 204 colonized adult Il10 -/mice with a more simple synthetic microbiota containing only the 10 species 205 ("SM10") that have previously been shown to be unable to grow on mucin oligosaccharides [12]. 206 Mice colonized with a microbiota that lacked the four mucin degraders exhibited 100% survival 207 on the FF diet until 60d post colonization (n=7), lower cecal Lcn2 at 60d ( Fig. 2A) and reduced 208 histological damage (Fig. 2B). Removal of the 4 known mucin-degrading bacteria also 209 8 corresponded with decreased mucus erosion (Fig. 2C, Extended Data Fig. 3A), suggesting that 210 the damage they inflict on mucus integrity may be causal to the increased disease activity when 211 they are present. Consistent with this idea, SM10-colonized mice that were switched to the FF diet 212 for a total of 150d after colonization (136d after switch to FF) exhibited significantly better 213 survival (50%), revealing that the presence of mucin-degrading bacteria accelerates disease 214 development but that the presence of non-mucin-degrading bacteria will eventually elicit disease 215 in the context of the dysregulated Il10 -/immune system and fiber deprivation (Fig. 2D). 216 Interestingly, adding back single mucin-degrading bacteria to the SM10 community did not yield 217 the same level of inflammation observed with the full 14-member community, indicating that 218 multiple species may act synergistically to promote more severe disease ( Fig. 2A). Consistent with 219 the idea that low dietary fiber intake promotes the general proliferation of mucin-degrading 220 bacteria, the mice colonized with SM10 plus a single mucin degrader all exhibited expansion of 221 that mucin-degrading bacterium after the diet switch compared to SM14-colonized mice 222 (Extended Data Fig. 3B-G). This is in contrast to mice colonized with the full SM14, which 223 typically show expansion of Akkermansia muciniphila and B. caccae in response to low fiber 224 (Extended Data Fig. 3B). Despite a lack of weight loss (not shown) and lower Lcn2, the presence 225 of either B. thetaiotaomicron or A. muciniphila as the sole mucin-degrader produced histological 226 inflammation in the cecum that was statistically identical to the full SM14 (Fig. 2B,E). 227 As a separate test of the protective role for mucus in this colitis model, which could 228 potentially be influenced by the presence of dietary fiber independently of bacteria, we bred Il10 -229 /-Muc2 -/double knockout (DKO) germfree mice, colonized them with the SM14 and fed either the 230 FR or FF diets. As expected, these mice lost weight quickly (100% needed to be sacrificed before 231 30d on FF and before 34d on FR; n=5 and 7, respectively) and showed severe inflammation and 232 mortality regardless of diet ( Fig. 2F-J). These results suggest that the high sugar content of the 233 FF diet is not the main driver of the acute disease since the low sugar FR diet also allows 234 inflammation when the mucus layer is genetically eliminated. Interestingly, inflammation in DKO 235 mice developed throughout the lower intestine and was more severe in the colon than in the cecum 236 ( Fig. 2I-K). This is likely due to the mucus system being thinner and patchier in the cecum than 237 in the colon [26]. Thus, when Il10 -/mice experience low fiber-induced mucus erosion, the secreted 238 mucin barrier is likely to fail first in the cecum. In contrast, in DKO mice the uniform elimination 239 of Muc2 appears to promote more widespread inflammation that worsens more quickly in the 240 colon. 243 and adaptive immune responses, and the critical role of Th1/Th17 cells has been described in the 244 Il10 -/colitis model in conventional/SPF settings (Keubler et al., 2015). Nevertheless, how specific 245 microbial triggers influence the underlying immune pathways and how these immune responses 246 develop temporally are less clear. We repeated the same experiments described above with a 247 different line of Il10 -/mice in the same C57BL/6J background used above (mouse facility: In SM14-colonized mice that were weaned to the FF diet, inflammation could be detected 263 in the cecum via increased Lcn2 as early as 35 dpw and this effect was absent in mice colonized 264 from birth with SM10 (Fig. 3A). SM14 mice weaned onto the FF diet also exhibited low fecal 265 Lcn2 at 35 dpw, consistent with data shown above and supporting the conclusion that disease 266 develops earlier in the cecum in mice with intact mucus production (Fig. 3A). An expansion of 267 natural killer (NK) cells was detectable in the cecum of FF-fed mice as soon as 35 dpw (Fig. 3B, 268 left panel), and this diet-dependent expansion was similar in SM10-and SM14-colonized ceca. 269 Interestingly, NK cells expanded in the colons of both FR and FF-fed SM14-colonized mice as 270 soon as 35 dpw, but not in the colons of SM10-colonized mice (Fig. 3B, right panel). This 271 observation further suggests that the presence of mucin-degrading bacteria is more important to 272 elicit responses in the colon where the mucus system is thicker and mucus erosion is required to 273 10 increase contact with bacterial antigens. Further supporting region-specific development of host 274 responses, FF-fed germfree Il10 -/mice, but not WT germfree mice, showed increased NK cells in 275 the cecum, but not in the colon, at 79 dpw compared to FR-fed controls, implying that long-term 276 fiber deprivation can increase NK cell infiltration in the cecum independently of the microbiota 277 (Extended Data Fig. 4A, B). 278 While NK cells were more abundant at 35 days and later decreased, T cell recruitment 279 (especially CD4 + T cells) increased over time in both the cecum and colon of colonized Il10 -/mice 280 ( Fig. 3C and Extended Data Fig. 5A, B). In the ceca of colonized Il10 -/mice, fiber deprivation 281 increased both CD4 + and CD8 + T cell populations by 35 dpw, with CD4 + T cells increasing further 282 by 79 dpw independently of the diet (Fig. 3C, D). In the ceca of GF Il10 -/mice and SM14-283 colonized WT mice, CD8 + T cells and CD4 + T cells were also more abundant in FF-fed than in

287
Consistent with previous descriptions of Il10 -/colitis in conventional mice (Keubler etal.,288 2015), both SM10-and SM14-colonized mice generally exhibited infiltration of Th1 and Th17 289 cells over time. Especially for Th17 cells, this infiltration was higher in FF-fed mice in both the 290 cecum (Fig. 3E) and the colon (Fig. 3F). While Th17 cells reached nearly equivalent levels at 79 291 dpw in SM10 and SM14-colonized ceca (Fig. 3E), overall levels were lower in the SM10-292 colonized colons compared to SM14. Furthermore, fiber deprivation increased Th1 levels in the 293 SM14-colonized colons, but not in the SM10-colonized ones (Fig. 3F), consistent with the weight 294 loss observed only in FF-fed SM14 colonized animals. In line with the readouts mentioned above, 295 this observation supports the idea that the role of mucin-degrading bacteria is more important in 296 the colon where erosion of mucus facilitates the induction of the anti-microbial Th1 responses. To 297 provide functional support to these immune cell profiles, we compared levels of Th1-and Th17-298 type cytokine transcripts in ceca and mesenteric lymph nodes using qPCR for 35 and 79 dpw. Fiber 299 deprivation led to increased expression of Th1 (IFN-γ, TNF-α, IL-6, IL-12) and Th17 cytokines 300 (IL-17F, IL-22, IL-23 and TGF-β), as well as the mucin-inducing cytokine IL-13 in SM14-  SM10-and SM14-colonized mice, while the Th1-polarizing cytokines IFN-γ, IL-6 and IL-306 12 were only increased in SM14-colonized mice, consistent with a regional dependence on mucin-307 degrading bacteria to develop Th1 responses ( Fig. 3G and Extended Data Fig. 5C). 308 Despite the IL-10 deficiency, Foxp3 + regulatory T cell (Treg) recruitment was also 309 detectable and higher during FF feeding of SM14 Il10 -/mice (Fig. 3H, Extended Data Fig. 5D). 310 This expansion being independent of IL-10 suggests a mechanism driven by other regulatory 311 mediators such as TGF-β, whose transcript levels increased in SM14-colonized cecal tissues as 312 soon as 35 dpw (Extended Data Fig. 3Y). Regulatory T cells expressing Tbet or RORγt have been 313 proposed as counter-regulators of inflammatory Th1 and Th17 cells, respectively [27,28]. Alterations to IgA-microbiota interactions precede inflammation 329 Since coating with immunoglobulin A (IgA) has been proposed to identify bacteria that are 330 potentially more colitogenic [31], we next focused on how low fiber-induced inflammation 331 development alters bacterial IgA coating in our Il10 -/model. The IgA response is a common anti-332 microbial response in the gut and is usually upregulated in colitis in human and mouse models 333 [32]. Mirroring the trend observed for early inflammation (Fig. 3A), soluble IgA titers were 334 increased in the cecum but not feces of FF-fed SM14-colonized Il10 -/mice at 35 dpw compared 335 to FR (Fig. 4A). However, prolonged FF-feeding (79 dpw) resulted in depletion of IgA-producing 336 plasma cells in both the cecum and colon (Fig. 4B). This loss was also observed in FF-fed WT 337 mice that were colonized with SM14 ( Fig. 4B), suggesting a diet-driven effect on IgA-producing 338 cell loss rather than a depletion caused by the absence of IL-10. This is consistent with a published 339 report showing reduced titers of serum IgA and less IgA + B cells in the small intestine of wild-340 type mice fed a zero-fiber diet compared to a high-fiber diet [33]. In parallel with reduced IgA-341 producing cells after FF-feeding, the proportion of IgG-producing cells was increased at 35 and 342 79 dpw in both the cecum and colon of SM14-colonized Il10 -/mice (Extended Data Fig. 6A,B). 343 Consistent with the high production of Th1 cytokines in the presence of mucolytic bacteria, FF-344 feeding increased the proportion of IgG-producing cells only in SM14-but not in SM10-colonized 345 colons (Extended Data Fig. 6B) [34]. 346 We next sought to determine how changes in IgA-producing cells are reflected in bacterial 347 IgA coating. Intriguingly, analysis of IgA-coated bacteria revealed the presence of 2 differentially 348 coated populations in FR-fed mice: a large population with low-coating and a smaller population 349 with high coating and we observed that the highly coated population was nearly absent in SM14-350 colonized mice fed the FF diet (Fig. 4C). Consistent with previous studies [32], the amount of total 351 IgA-coated bacteria (both populations combined) was higher in FF-fed mice than in FR-fed mice 352 by 79 dpw (Fig. 4D, left). However, only the proportion of low-coated bacteria increased in FF-353 fed mice (Fig. 4D, right), while the highly-coated bacteria were diminished in fiber-deprived mice, 354 as early as 21 days of feeding (Fig. 4D, middle). Interestingly, the FF diet-induced increase in 355 total IgA-coating by 79 dpw was not observed in SM10-colonized mice, but was observed in SPF 356 Il10 -/mice and SM14-colonized WT mice, supporting a mechanism driven by the microbiota 357 rather than the IL-10 deficiency (Fig. 4E, left). However, highly coated bacteria were diminished 358 in SM10-colonized mice as in SM14-colonized mice as soon as 21 dpw (Fig. 4E, middle and 359 Extended Data Fig. 6C) and low-coated bacteria only slightly increased by 79 dpw (Fig 4E,   360 right), suggesting a mechanism that is partly driven by the presence of mucolytic bacteria.

361
Given the altered pattern of IgA coating and the fact that IgA coating has been proposed as 362 a marker of bacteria with more colitogenic potential [31], we examined the IgA coating of 363 individual SM14 bacterial members. Consistent with the overall early loss of highly coated 364 bacteria, changes in IgA-coating indexes in the SM14 community appeared as soon as 21 days of 365 feeding (Fig. 4F). Among the SM14 members, A. muciniphila, D. piger, E. coli and C. aerofaciens 366 showed reduced IgA coating index (ICI) values at one or both timepoints in FF-fed Il10 -/mice. In 367 wild-type mice, this FF diet-induced reduction in IgA coating was less severe for A. muciniphila, 368 but more pronounced for D. piger and C. aerofaciens, revealing that IL-10 deficiency affects the 369 13 IgA-coating profile of commensals in a species-dependent manner. In contrast, E. rectale showed 370 a low ICI in FR-fed mice and this increased with FF-feeding, a condition that corresponds to it 371 being present at very low levels (and therefore not contributing much to overall coating 372 measurements), likely due to its inability to compete for mucin-derived nutrients. The causes and 373 effects of variations in IgA coating are still being determined. However, our results demonstrate 374 that different combinations of microbial colonization, diet and host immune status can alter both 375 the amount and affinity of intestinal IgA, as well as the bacteria that it targets. Along with the data 376 shown above, our results suggest the possibility that fiber deprivation initiates early disruptions in 377 IgA-microbiota interactions along with a loss of IgA-producing plasma cells allowing A. 378 muciniphila and E. coli to expand in the absence of proper IgA coating and contribute to colitis. Because we observed that SPF mice fed the FF diet did not develop inflammation with the 382 same severity as those colonized with SM14 (Fig. 1H,K), despite having thinner mucus (Fig. 1L), 383 we performed co-housing experiments in which pups born to SPF-and SM14-colonized mothers 384 were mixed at weaning and exposed to each other's microbiomes. Co-housing SM14-colonized 385 Il10 -/pups with SPF mice prevented the weight loss phenotype observed in response to feeding 386 the FF diet (Extended Data Fig. 7A). However, it did not reduce cecal inflammation as measured 387 by both cecal Lcn2 (Fig. 5A) or histology (Fig. 5B), indicating that the weight loss and 388 inflammation phenotypes can be uncoupled in the presence of different microbes perhaps due to 389 the timing of barrier disruption or other changes in microbiota metabolism. Mice that were born 390 to SPF mothers showed a more variable response to co-housing with SM14-colonized pups, with 391 only some of these mice developing disease that was often similar in severity to FF-fed, SM14-392 colonized mice (Fig. 5A,B). Time course analysis of fecal Lcn2 revealed a larger difference 393 between co-housed and non-co-housed SPF mice than cecal measurements and this marker 394 increased in co-housed SPF mice after 65 days, a trend that was nearly identical to SM14 mice 395 (Fig. 5C). Although inflammation was variable in co-housed SPF mice, there was a positive 396 correlation in the two inflammatory measurements for this group (Fig. 5D). 397 The SPF pups are naturally colonized with a natively complex microbiota beginning at 398 birth, whereas the SM14 pups receive a simpler microbiota of known composition. The fact that 399 co-housed SPF pups develop worse inflammation compared to their littermates that are not co-400 14 housed suggested that some of the SM14 bacteria can invade the murine SPF microbiota and 401 worsening disease, perhaps synergizing with native murine microbes that perform some of the 402 same metabolic or immunostimulatory functions of other SM14 members. To determine which 403 SM14 species invaded the SPF microbiota and when, we performed 16S rRNA gene sequencing 404 on fecal samples from co-housed mice from both groups over the 100d time course. Our analysis 405 revealed that only some of the SM14 bacteria are transferred to SPF mice, a finding that might be 406 expected with human gut bacteria invading a more complex, mouse-adapted microbiota. However, 407 at least 6 members of the SM14 could be detected in co-housed SPF mice, often transiently or at 408 the very end of the time course when inflammation had begun to develop ( Fig. 5E-G, Extended 409 Data Fig. 7B-J). The most prominent of these invading bacteria was the human commensal E. coli 410 strain HS, which appeared in co-housed SPF mice around 22 dpw, prior to the onset of 411 inflammation, and gradually increased, eventually reaching levels >10% in most mice (Fig. 5E).

412
Two of the mucin-degrading SM14 bacteria (A. muciniphila and Ba. intestinihominis) showed a 413 similar trend, albeit reaching lower levels and with variability among individual mice (Fig. 5F,G). 414 Four other SM14 bacteria (C. aerofaciens, B. uniformis, B. thetaiotaomicron and B. caccae) 415 showed transient, small increases around the time that inflammation was increasing (~44 dpw) and 416 these organisms decreased thereafter (Extended Data Fig. 7B-J). A test of whether or not weekly 417 gavages of E. coli HS into SPF mice that were otherwise not exposed to additional SM14 bacteria 418 did not support that conclusion that this species is the sole cause of increased inflammation  A notable characteristic of the FF diet that we used is that its macronutrient composition 424 (Supplemental Table 1) resembles some exclusive enteral nutrition (EEN) diets that are used 425 clinically to treat some IBD presentations. EEN diets often contain low or no fiber [35] and have 426 proven to be effective at inducing IBD remission in humans, although the precise mechanism(s) 427 of action is still unknown [36]. Since dysfunction in IL-10 signaling is just one of several host 428 pathways implicated in IBD, it is possible that this particular immune signaling axis is sensitive to 429 lack of fiber, whereas others are not and those are the ones that benefit from EEN. To determine 430 if an EEN diet that lacks fiber promotes inflammation in our gnotobiotic Il10 -/model, we weaned 431 15 SM14-colonized pups onto a commercial EEN diet, which is normally taken as a liquid but in this 432 case was freeze-dried, pelleted and sterilized by gamma irradiation (water was provided ad 433 libitum). The average weight trajectories of 15 mice (3 separate experiments) supported the 434 conclusion that the low fiber EEN diet promotes some disease development, although not as severe 435 as the FF diet (Fig. 6A). Cecal Lcn2 measurements and histology further revealed a wide amount 436 of variation in the disease present in individual animals, with some mice resembling healthy FR-437 fed mice, some resembling diseased FF-fed mice and some intermediate (Fig. 6B,C). Interestingly, 438 the EEN diet increased the proportion of IgA-coated bacteria as soon as 21dpw and conserved the 439 high-coated population compared to mice fed FF (Extended Data Fig. 8), suggesting immuno-440 regulatory properties distinct from both the FR and FF diets. Despite experiencing less 441 inflammation, the EEN mice still exhibited reduced mucus thickness, which we expected given 442 the lack of fiber in the formula used (Fig. 6D). Measurements of short-and branched chain fatty 443 acids revealed that mice fed the EEN diet had elevated amounts of the branch-chained fatty acid 444 (BCFA) isobutyrate (Fig. 6E). While isobutyrate varied between 7-281 µMol/g cecal contents, 445 high levels were only weakly correlated with low inflammation (Fig. 6F). Isobutyrate is an isomer 446 of the more widely studied, anti-inflammatory short-chain fatty acid butyrate, which did not 447 increase, and isobutyrate is derived from L-valine fermentation by certain gut bacteria [37].

448
Interestingly, the other two BCFAs (2-methyl butyrate and isovalerate) did not increase despite 449 being present at similar abundance in soy and milk protein [38], the two dietary proteins used in 450 the EEN formulation (Fig. 6E). This suggests that the increase in isobutyrate may not be 451 attributable to increased bulk dietary protein fermentation, which would be expected to increase 452 all three BCFAs. Providing either isobutyrate or butyrate (both 35 mM) in the drinking water of 453 pups weaned onto the disease-promoting FF diet decreased weight loss (Fig. 6G) as well as the 454 amount of inflammation observed (Fig. 6H,I), revealing that either of these molecules can offset 455 the diet-and microbiome-induced damage observed in this model.

456
Feeding the EEN diet promoted changes in the composition of the SM14, which most 457 notably included a ~250-fold increase in the levels of E. rectale (Fig. 6J). While this Firmicute is 458 known to produce butyrate, measurements of culture supernatants from 13 of the SM species in 459 medium supplemented with L-valine plus other peptides did not reveal that E. rectale produces 460 isobutyurate under the medium conditions tested. Rather, this metabolite was produced by all 4 of 461 the Bacteroides, plus M. formatexigens (Fig. 6K). Nevertheless, to determine if the large increase 462 16 in E. rectale abundance was functionally connected with production of isobutyrate, we bred 463 neonatal mice born to parents harboring the SM14 community but lacking E. rectale ("SM13 464 minus E. rectale") and fed them the EEN diet. Consistent with a causal role (indirect or direct) for 465 E. rectale in isobutyrate production, mice lacking E. rectale failed to produce isobutyrate, as well 466 as butyrate as expected (Fig. 6L). These mice also exhibited increased lethality (62.5% survival at 467 100d) relative to EEN fed mice harboring the full SM14 (Fig. 6M), although there was not a 468 significant increase in cecal Lcn2 levels relative to the EEN/SM14 group (Fig. 6N).  The link between low dietary fiber and compromised thickness and/or permeability of the 483 mucus layer has been emerging through several studies [12][13][14]. Here, we make a direct connection 484 between the dietary fiber-gut microbiome axis in the context of IBD genetics, albeit in a way that 485 depends on the specific bacteria present (SM14 vs. SPF). The discovery that the EEN diet tested 486 exerts partial protection against inflammation was unexpected but may be unsurprising given the 487 complexity of diet-microbiome-host interactions in the gut. The EEN diet clearly holds the 488 potential to elicit mucus layer erosion, as well as inflammation, in this murine model, which we 489 attribute to its lack of fiber along with other potential contributing factors. However, the finding 490 that isobutyrate production may partially offset the development of inflammation, in a manner that 491 is dependent on the presence of E. rectale, provides exciting opportunities to explore new 492 microbiome pathways that are both influenced by diet and can exert beneficial effects. Identifying 493 the critical bacteria in our SM14 and their corresponding pathways for isobutyrate production will 494 be essential and could illuminate why ~15% of patients given EEN do not respond [42], perhaps 495 because they are missing these bacteria/pathways. beneficial processes (e.g., butyrate, isobutyrate) while reducing detrimental events like mucus 516 erosion. While some of these levers can be manipulated through diet, a promising path for 517 intervention may eventually include adding or replacing specific bacterial taxa within a person's 518 individual microbiota with more optimal strains or perhaps even those that are engineered to 519 increase beneficial metabolites. proceeded to diet change as previously described. The synthetic microbiota bacteria were grown 543 in their respective medium (Desai, 2016) or a modified YCFA medium [46] prior to community 544 assembly for gavages. The bacteria were cultivated under anaerobic atmosphere maintained with 545 a gas mixture (85% N2, 10% H2, 5% CO2) to an optical density (absorbance 600nm) ranging from 546 0.5 to 1.0. The communities were assembled by mixing equal volumes of each specific bacterium 547 and aliquoted into sealed screw cap tubes with its own headspace. Each mouse was gavaged with 548 0.2 mL of its specific community, depending on the experiment, with freshly prepared inocula for 549 two-three consecutive days. A humane endpoint was used for mice that lost >20% of their starting 550 weight and these mice were counted as lethalities in the weight loss and survival curves shown.

551
Animals were euthanized using CO2 asphyxiation for 5 minutes followed by cervical dislocation.

552
The gastrointestinal tracts were retrieved quickly, to prevent autolysis, and the sections separated.

553
Cecal and colon contents of each animal were flash frozen in liquid nitrogen and kept at -80°C 554 until further use.

559
Rettenmaier (Schoolcraft, MI, USA) and added to the FF diet at 7.5% w/w with a corresponding 560 decrease in the amount of glucose in the FF diet. Soluble, highly digestible starch was provided 561 by Cargill (Gel Instant 12412).

563
DNA from fecal and cecal samples were isolated using a bead-beating phenol:chloroform

579
Freshly retrieved ileal, cecal and distal colon tissues were transferred into RNAlater TM 580 (QIAGEN) and kept at 4°C up to a week. Then, RNAlater TM was removed and tissues were stored 581 at -80°C until further use. Frozen tissues were transferred into 1 mL of TRIzol reagent 582 (Invitrogen TM ), homogeneized with a 5 mm metal bead on a bead beater for 8 min at 30Hz and 583 centrifuged for 3 min at 13000 rpm, 4°C. The supernatant was recovered, mixed thoroughly with 584 200 µl of chloroform and incubated at room temperature for 2-3 min before a centrifigation for 15 585 min at 13000 rpm, 4°C. The aqueous phase was recovered, mixed again with an equal amount of 586 chloroform and centrifuged for 15 min at 13000 rpm, 4°C. The aqueous phase was recovered, 587 mixed by inversion with 500 µl of isopropanol, incubated for 10 min at room temperature, and 588 centrifuged for 10 min at 13000 rpm, 4°C. The pellet was washed with 1 ml Ethanol 70% and 589 centrifuged for 5 min at 10000 rpm. The supernatant was discarded and the pellet dried for 5-10 590 min at 37°C, resuspended with 50 µl nuclease-free water and incubated for 15 min at 56°C. Finally, 591 samples were treated with DNase following the Thermo Scientific DNase1, RNase-free Protocol,  A Waters Acquity UPLC BEH TSS C18 column (2.1 x 100mm, 1.7µm) column was used 649 with mobile phase A) consisting of 0.1% formic acid in water; mobile phase (B) consisting of 650 0.1% formic acid in acetonitrile. Gradient program: mobile phase (B) was held at 15% for 1 min, 651 increased to 55% in 19 min, then to 99% in 20 min and held for 2 min before going to initial 652 condition and held for 4 min. The column was at 40 C and 2 µl of sample was injected into the            Weight trajectories of adult C57BL/6J Il10 -/mice colonized with the human synthetic microbiota containing 14 species (SM14) and maintained on a fiber rich (FR) diet or switched to a fiber free (FF) diet after 14 d of colonization. Curves represent polynomial (quadratic) equations fit to all of the weights gathered at various days for the two treatments. Weights were measured more frequently after day 40 due to the declining health of the FF group and animals that were euthanized were excluded from the curve at later points. Two FF-fed animals from an early experiment were found dead on arrival and could not be included in subsequent analyses. All others were euthanized if they reached less than 80% starting weight and counted as lethality. The right axis shows survival in each group over time. The number of mice in each group, along with sexes are indicated in the figure legend. B. Representative cecal histology of FF (top) and FR (bottom) fed colonized Il10 -/mice. In the top panel, block arrow points to a particularly large ulcer and the line arrow to an area of edema. Scale bars, 250 μm. C. Quantitative, blinded histological scoring of ileal, cecal, and colonic tissue taken from colonized Il10 -/mice fed the FF diet. Bold horizontal bar represents the mean and lighter error bars the S.E.M. (n=15-20, one-way ANOVA and post hoc test with Holm-Šídák's multiple comparison test) D. Quantitative, blinded histological scoring of cecal tissue taken from colonized Il10 -/mice fed the FR and FF diets, along with additional treatments discussed in the text to manipulate individual diet (FR/FF), colonization (SM14 or germfree, latter indicated by dashed lines) and host genotype (wild-type, WT, or Il10 -/-) variables. Here, and in subsequent panels, the red box highlights the condition with most severe inflammation. Bold horizontal bar represents the mean and lighter error bars the S. K.           Figure 3. Mucolytic bacteria influence the inflammatory pathways of colitis. A. Lipocalin-2 levels in the cecal contents and feces of GF and colonized Il10 -/mice (n=4-7, two-way ANOVA and post hoc test with Original FDR method of Benjamini and Hochberg). B. Proportion of NK cells among CD3-CD45+ cells in the cecum and colon lamina propria of SM10-and SM14-colonized Il10 -/mice (n=4-11, two-way ANOVA and post hoc test with Original FDR method of Benjamini and Hochberg). . Data are represented as mean ± SD. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

C.
Days post-weaning         Fig. 1N, showing that no observable weight loss occurred prior to 30d or 40d fiber interventions.

Figure ED6
Il10  Diet  d35  d79  d35  d79  d35  d79  d79  d35  d79  d35  d79  d35  d79  d79  Timepoint  -- (CD138+), IgD-and IgG-producing cells among CD3-CD45+ cells in the cecum (A.) and colon (B.) of GF, SM10-and SM14-colonized Il10 -/and WT mice (n=4-11, two-way ANOVA and post hoc test with Original FDR method of Benjamini and Hochberg). C. Percentages of total (top), IgAhigh (middle) and IgAlow (bottom) -coated bacteria as shown in (C) in the feces of SM10-, SM14-or SPF-colonized Il10 -/mice and SM14-colonized WT mice fed the FR or the FF diet for 21 days (n=5-8, two-way ANOVA and post hoc test with the two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli). Data are represented as mean ± SD. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. L. Cecal lipocalin measurements in SPF mice that were gavaged weekly with E. coli HS and fed the FF diet. (n=6-7, one-sample Student's t-test and Wilcoxon test). K. PCR analysis using E. coli HS specific primers [12] for the presence of E. coli HS in mice that were mock gavaged weekly with PBS (left) or E. coli HS (right). Extended Data Figure 8. The EEN diet affects IgA coating differently than FR and FF feeding. Percentages of total (left), IgA high (middle) and IgA low (right) -coated bacteria as shown in Fig. 4C in the feces of SM14-colonized Il10 -/mice fed the FR, FF or EEN diet for 21 or 79 days (n=3-5, two-way ANOVA and post hoc test with the two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli). Data are represented as mean ± SD. *p < 0.05; **p < 0.01; ***p < 0.001. *Note the only fiber added to the FF diet is crystalline cellulose as a bulking agent and is not known be accessible by any SM14 bacteria.