Plant-based Enteral Nutrition Modifies the Gut Microbiota and Improves Outcomes in Murine Models of Colitis.

Exclusive enteral nutrition (EEN), ie, restricting patients to commercially available liquid diets with avoidance of solid food, is an effective treatment for Crohn’s disease (CD). However, commercial formulas used for EEN generally have low fiber and high sugar content and contain emulsifiers that worsen colitis in animal models. Despite its efficacy, EEN appears to accentuate microbial changes observed in CD, specifically a loss of protective anaerobes and the growth of pathogens. Plant-based enteral nutrition (PBEN) represents a high-fiber alternative to conventional enteral nutrition (CEN) without added sugar or artificial ingredients. The advent of commercially available PBEN is relevant to inflammatory bowel disease because diets low in fruits and vegetables represent a risk factor for CD, and because many plant-derived compounds are therapeutic in preclinical colitis models.4, 5, 6

We sought to determine the effects of PBEN on the microbiota and outcomes in murine colitis models. We initially observed that male C57BL/6 mice randomized to receive PBEN (Liquid Hope; Functional Formularies, Centerville OH) experienced better outcomes in a 4% dextran sulfate sodium (DSS) (MW 40,000-50,000, Thermo Fisher Scientific, Waltham, MA) colitis model than chow (IsoPro RMH 3000; LabDiet, St Louis, MO) or 2 conventional formulas (Vital or Pediasure; Abbott Laboratories, Lake Bluff, IL). We refer to these groups as PBEN, CHOW, CEN1, and CEN2. Mice fed PBEN experienced less weight loss than CHOW, CEN1, and CEN2 mice after exposure to DSS (Figure 1, P << .001). At death, plasma interleukin 6, fecal lipocalin, histologic scores of inflammation, and colon length suggested less inflammation in PBEN (Supplementary Figure 1). Similar results were obtained in numerous permutations of the original experiment: females instead of males, Taconic Biosciences rather than Jackson Laboratory, 2% DSS instead of 4%, and the rectally instilled trinitrobenzene sulfonic acid colitis model (Supplementary Figure 2A). We also modified the original experiment by alternating formula feeding and 4% DSS water every 12 hours to confirm that the benefit of PBEN was not an artifact from admixing DSS and liquid formulas (Supplementary Figure 2B). Finally, to determine whether these observations extend beyond chemically induced colitis, we randomized mice in a T-cell transfer colitis model to receive CHOW, PBEN, or CEN. Improved outcomes were observed with PBEN in all experiments (Supplementary Figure 2C).

Figure 1

Effect of diet on outcomes in 4% DSS colitis model. Weight curves in mice receiving experimental diets for 7 days, followed by DSS exposure (n = 8/group). CHOW mice received DSS in their drinking water. Asterisks denote significance in post hoc test between only PBEN and all other individual groups (ANOVA, Tukey HSD, P < .05). Error bars represent standard deviation.

Supplementary Figure 1

Effect of diet on gut inflammation in DSS colitis models. Plasma interleukin 6 measurements (n = 8/group) (A), normalized fecal lipocalin concentrations (n = 14-20/group) (B), histologic scores of mucosal injury (n = 8/group) (C), and colon length (n = 11-16/group) (D) in mice receiving CHOW, CEN1, CEN2, or PBEN for 7 days, followed by 4% DSS mixed into the diets. CHOW animals were exposed to the same concentration of DSS mixed in water. Asterisks denote significance (ANOVA, Tukey HSD, P < .05). Lipocalin concentration (pg/μL) was normalized to total protein concentration (μg/μL) in stool. (C) Representative slides of H&E staining of colonic mucosal tissue from each diet group.

Supplementary Figure 2

Effect of diet on outcomes in alternative colitis models. (A) Disease Activity Index (DAI) in mice fed experimental diets for 7 days, followed by a single dose of TNBS administered by enema on day 8. Asterisks denote where mean DAI was significantly lower in mice fed PBEN compared with CHOW, CEN1, and CEN2 in post hoc testing (n = 8/group; ANOVA, Tukey HSD, P < .05). (B) DAI in mice fed experimental diets for 7 days, followed by water only, alternating every 12 hours with the experimental diet mixed with 4% DSS (day 8 to day 19). Asterisks denote where mean DAI was significantly lower in mice fed PBEN compared with CHOW, CEN1, and CEN2 in post hoc testing (n = 8/group; ANOVA, Tukey HSD, P < .05). (C) DAI in mice undergoing T-cell transfer (day 8) fed experimental diets for 40 days. DAI measurements started on day 28. Asterisks denote where mean DAI was significantly lower in mice fed PBEN compared with CHOW, CEN1, and CEN2 in post hoc testing (n = 8/group; ANOVA, Tukey HSD, P < .05). (D) Weight curve in germ-free (GF) mice fed CEN1 or PBEN for 7 days, followed by 4% DSS exposure (n = 9/group), compared with specific pathogen-free mice (SPF, n = 28/group). Asterisks denote significance between the PBEN-GF group compared with the PBEN-SPF group (ANOVA, Tukey HSD, P < .05). (E) Weight curves in mice fed CEN1 and PBEN receiving enteral antibiotics throughout a 7-day feeding period, followed by 4% DSS exposure (n = 8/group). Asterisks denote significance between mice fed PBEN with antibiotics and mice fed PBEN without antibiotics (ANOVA, Tukey HSD, P < .05). Error bars represent standard deviation.

To assess the effect of each diet on the gut microbiota, we analyzed 16S rRNA gene sequences within fecal pellets collected before and after a 7-day feeding trial. At baseline, the number of observed taxa (alpha diversity) was equal across all groups. After the feeding trial, alpha diversity was significantly decreased in CEN1/CEN2 compared with PBEN/CHOW mice (Figure 2; analysis of variance [ANOVA], Tukey honestly significant difference [HSD], P << .001). Similar results were seen with community composition (beta diversity; Supplementary Figure 3). Baseline samples from all groups initially clustered together in principal coordinates analysis space. After 7 days of feeding, PBEN samples clustered separately from the other groups (Supplementary Figure 3A and B; permutational multivariate analysis of variance [PERMANOVA], P < .05). PBEN and CHOW samples were marked by increased commensal anaerobes (Clostridiales, Lachnospiraceae, and Ruminococcus) (Supplementary Figure 3C; LEfSe, P < .05). CEN1/CEN2 samples were remarkably similar to each other and were enriched with gram-negative pathogens from the family Enterobacteriaceae, a finding previously described in CD patients receiving EEN.3, 7

Figure 2

Effect of diet on gut microbial diversity. Alpha diversity of fecal microbial communities (observed OTUs) before and after a 7-day trial of the experimental diets. Asterisks denote significance (ANOVA, Tukey HSD, P < .05).

Supplementary Figure 3

Effect of diets on composition of gut microbial communities. Beta diversity comparisons of microbial communities in the stool between mice before and 7 days after starting the diets (n = 8/group). Shown is the principal coordinates analysis of weighted Unifrac (A) and abundance Jaccard distances (B). Axis labels indicate the proportion of variance explained by each principal coordinate. PBEN samples after 7 days of feeding clustered separately from CHOW and CEN1/CEN2 samples (PERMANOVA, P < .05). (C) Relative abundance of bacteria taxa that were differentially abundant in cecal contents of PBEN mice after a 7-day feeding trial. Relative abundances of S24-7, Clostridiales, Lachnospiraceae, and Ruminococcus were statistically significantly higher in mice fed PBEN and CHOW compared with CEN1 and CEN2 (LEfSe, P < .05). Relative abundance of Enterobacteriaceae was elevated in mice fed CEN1 and CEN2 compared with CHOW and PBEN (LEfSe, P < .05).

To ascertain whether the gut microbiota mediates the protective effect of PBEN, we repeated the 4% DSS colitis model in germ-free animals receiving CEN1 or PBEN for 7 days before and during DSS exposure. As shown in Supplementary Figure 2D, weight loss in the absence of a gut microbiota was severe in both the PBEN and CEN1 groups. In addition, we administered antibiotics to mice fed CEN1 or PBEN for 7 days before and during DSS exposure. Antibiotics significantly worsened outcomes in PBEN mice (Supplementary Figure 2E; P < .05) but did not affect outcomes in CEN1 mice. These results support the conclusion that gut microbes are not required for the onset of colitis but are indeed necessary for the protective benefit of the PBEN diet.

Because of the impact of the microbiota on the protective effect of PBEN, we measured the concentrations of microbial metabolites in cecal contents from mice receiving PBEN, CEN1, or CEN2 for 21 days (Supplementary Figure 4). Samples from CHOW mice were not analyzed. PBEN samples contained increased amounts (P < .05) of the bile acids lithocholate and taurolithocholate and the plant-derived hydroxycinnamic acid, which have been shown to exert anti-inflammatory activity or confer protection in colitis models.8, 9 Also, CEN1/CEN2 mice possessed higher concentrations of amino acids than PBEN (P < .05), consistent with a described association between fecal amino acids, gut dysbiosis, and disease activity in CD. Concentrations of short-chain fatty acids were not different among diets except for propionate, which was significantly increased in CEN1 compared with CEN2 and PBEN (P < .05).

Supplementary Figure 4

Concentration of cecal metabolites in mice fed PBEN, CEN1, and CEN2 for 21 days. Concentrations of secondary bile acids (nmol/L per mg of cecal contents) 7-ketolithocholic acid (7-ketoLCA), γ-muricholic acid (HCA), lithocholic acid (LCA), taurochenodeoxycholic acid (TCDCA), taurodeoxyhyocholic acid (TDHCA), taurohyocholic acid (THCA), taurolithocholic acid (TLCA), and tauroursodeoxycholic acid (TUDCA) were elevated in mice fed PBEN compared with CEN1 and CEN2 (n = 4/group; ANOVA, Student t test with Bonferroni correction, P < .05). Concentration of amino acids (ng/mL per mg of cecal contents) isoleucine, leucine, methionine, and phenylalanine were increased in mice fed CEN1 and CEN2 compared with PBEN (ANOVA, Student t test with Bonferroni correction, P < .05). Concentration of short-chain fatty acids (normalized concentration to an internal standard) acetate and butyrate were not significantly different between groups. Propionate was significantly increased in CEN1 compared with PBEN and CEN2 (ANOVA, Student t test with Bonferroni correction, P < .05).

In summary, the benefits of PBEN result in part from diet-driven changes in the gut microbiota, which in turn impact bile and amino acid metabolism. We acknowledge that many components of plant-based diets likely impart health benefits independently of the microbiota. We also acknowledge that currently available formulations of CEN have yielded positive outcomes as a clinical treatment for patients with CD. We therefore conclude that further studies are indicated to determine whether PBEN can improve outcomes even further for patients requiring supplemental enteral nutrition.