Address for Correspondence Allan M. Goldstein, MD, Massachusetts General Hospital, 55 Fruit St., Warren 1153, Boston, MA 02114. Tel: 617-726-0270; fax: 617-726-2167; e-mail: email@example.com
Background Congenital aganglionosis (Hirschsprung’s disease) results in colonic dysmotility and a risk for Hirschsprung’s-associated enterocolitis (HAEC), whose cause is unknown. We hypothesized that aganglionosis leads to microbiome changes that may contribute to HAEC risk.
Methods Colon and fecal samples were collected from endothelin receptor B-null (Ednrb−/−) mice, an established model of colorectal aganglionosis, at postnatal day 7 (P7), P20, and P24. We determined microbiome composition by 16S ribosomal RNA gene pyrosequencing and fecal metabolite profile by nuclear magnetic resonance spectroscopy.
Key Results Wild-type (WT) mice exhibited increasing species diversity with age, with mutant mice possessing even greater diversity. WT and mutant microbiomes, both fecal and colonic, significantly segregated by principal coordinates analysis based on species composition at all ages examined. Importantly, mutant mice contained more Bacteroidetes and less Firmicutes than WT, with additional genus- and species-level differences observed. Notably, mutant P7 colon was dominated by coagulase-negative Staphylococcus species, which were rare in WT. Mutant fecal metabolite profiles also differed, particularly in the abundance of formate, a short-chain fatty acid product of microbial fermentation.
Conclusions & Inferences Colorectal aganglionosis is associated with early and sustained disruption of the normal colonic and fecal microbiome, supporting the enteric nervous system as a determinant of microbiome composition. Furthermore, the differences observed suggest a potential contributory role for the microbiome in the etiology of HAEC. These findings provide a basis for further studies to determine the causative role of specific bacterial communities in HAEC and the potential to restore the normal microbiome in Hirschsprung’s disease.
The intestinal microbiome comprises a vast population of microbes essential for the maintenance of intestinal homeostasis, including regulation of gut development, mucosal inflammation, intestinal immunity, epithelial cell turnover, and barrier integrity.1–3 Given the importance of the microbiome in human health, understanding the factors determining its composition and diversity is essential. Many of the factors identified to date are external, including environment, diet, antibiotic usage, and hospitalization.4 In human infants, gestational age and mode of delivery have a major effect,5 which is particularly important as early microbial exposure has a lasting effect into adulthood.6 Microbiome composition is also regulated by intrinsic host factors, such as paneth cell defensins,7 mucosal IgA,8 and host genotype.9 This host–microbial interdependence, whereby the environment regulates microbial composition that in turn impacts the intestinal milieu, has been implicated in the etiopathogenesis of inflammatory bowel disease, irritable bowel syndrome, and necrotizing enterocolitis.10,11
Hirschsprung’s disease is a congenital megacolon due to terminal aganglionosis resulting from the failure of neural crest-derived cells to form the distal enteric nervous system (ENS).12 Although its most obvious manifestation is the absence of motility in the aganglionic segment, the most serious complication is the development of Hirschsprung’s-associated enterocolitis (HAEC), an inflammatory colitis that causes distension, diarrhea, and fever and can lead to bacterial translocation, sepsis, and death.13 Its etiology is unknown, but abnormalities in epithelial barrier function,14 mucosal immunity,15,16 and the microbiome have been proposed. Previous work has shown that bifidobacteria and lactobacilli are significantly decreased in Hirschsprung’s disease, and are even lower in HAEC.17 Further evidence implicating bacteria is provided by overgrowth of the toxigenic bacterium, Clostridium difficile, in association with HAEC.18 Analysis of fecal samples collected during HAEC episodes and remission phases suggest that particular microbiota are associated with HAEC, although contributing species could not be identified with the methodology employed.19
We used 16S ribosomal RNA (16S rRNA) gene pyrosequencing to characterize the fecal and mucosa-associated colonic microbiota at multiple ages in the endothelin receptor B (EdnrB)-null mouse, which develops colorectal aganglionosis and HAEC similar to human Hirschsprung’s disease.20,21 We also analyzed fecal metabolite data in 21-day-old wild-type (WT) and mutant mice. Our results show significant early and sustained disruption of normal microbiome composition, identifying an important role for the ENS in establishment and maturation of the microbiome and supporting our hypothesis that the colonic microbiome may be a key player in the etiopathogenesis of HAEC.
Materials and methods
Experiments were approved by the Institutional Animal Care and Use Committee. Ednrbtm1Ywa mice on a hybrid C57BL/6J-129Sv background (JAX #003295) were housed in identical conditions, maintained on 12-h light–dark cycle at 25 °C and fed standard rodent chow (Prolab Isopro RMH 3000 Irradiated; PMI Nutrition International, St. Louis, MO, USA). Homozygous mutant mice are identified by their white coat color and genotype confirmed by polymerase chain reaction. Mice were weaned at postnatal day 21 (P21).
Colon and stool isolation
Ednrb+/+ (WT) and Ednrb−/− (mutant) mice were euthanized by CO2 asphyxiation on P7, P20, and P24. Colon was removed and feces collected. After rinsing the colon with sterile phosphate-buffered saline, the distal two-thirds was frozen at −80 °C. The following numbers of mice were used: P7, n = 5 mutant and WT; P20, n = 5 mutant and WT; P24, n = 3 mutant and WT.
16S rRNA gene pyrosequencing
Tag-encoded FLX amplicon pyrosequencing was performed as described22–25 using Gray28F 5′TTTGATCNTGGCTCAG and Gray519r 5′ GTNTTACNGCGGCKGCTG, with primers numbered in relation to the primary sequence of E. coli 16S rRNA.26 Initial generation of the sequencing library utilized one-step PCR with 30 cycles, generating amplicons extending from the 28F primer with average read length of 400 bp. Tag-encoded FLX amplicon pyrosequencing analyses utilized a Roche 454 FLX instrument with Titanium reagents, and Titanium procedures performed at the Research and Testing Laboratory (Lubbock, TX, USA).
Initial sequence processing and classification All failed sequence reads, low-quality sequence ends, tags, and primers were removed, and sequence collections depleted of non-bacterial ribosome sequences and those with degenerate base calls, homopolymers >5 bp in length, reads <200 bp and chimeras,27 as described.22–25 For classification, sequences were queried against a custom database of high-quality bacterial 16S rRNA gene sequences curated monthly from NCBI (05-01-11). Database sequences were characterized as ‘high quality’ based on criteria similar to those utilized by RDP ver 9.28 Using a .NET and C# analysis pipeline, the resulting BLASTn outputs were compiled.22–25
Based on BLASTn-derived sequence identity and validated using taxonomic distance methods, bacteria were classified at the appropriate taxonomic levels. Sequences with identity scores to well-characterized 16S sequences >97% identity were resolved at species level, 95–97% at genus level, 90–95% at family level, 85–90% at order level, 80–85% at class level, and 77–80% at phylum level. The percentage of sequences (based on all analyzed sequence reads for each sample) assigned to each taxon was determined for each sample, providing relative abundance information among the individual samples. Evaluations presented at each taxonomic level, including percentage compilations, represent all sequences resolved to their primary identification or their closest relative.22,25,29 The taxonomic classification was used to generate hierarchical dendrograms of community structure (NCSS 2007, Kaysville, UT, USA). Dendrograms for genera were based upon highest average percentage relative abundance.
Diversity estimates Alpha diversity (examination of diversity within samples): A nominal 4871 sequences from each sample were selected based on highest average quality score, sequences trimmed to 350 bp and aligned with MUSCLE.30 A distance matrix was calculated from the alignment with PHYLIP.31 Diversity estimates were calculated using MOTHUR.32
Beta diversity (comparison of diversity between samples): Microbiome differences among samples were illustrated using principal coordinates analysis (PCoA), with distances among samples calculated using weighted and unweighted UniFrac distance measures.33 UniFrac measures the phylogenetic distances among samples. Unweighted UniFrac considers only differences in composition, whereas the weighted UniFrac considers relative abundances in addition to presence/absence. UniFrac distances were calculated using qiime,34 and all other analyses were conducted in r,35 using the vegan36 package.
Fecal metabolite analysis
Fecal samples were collected from WT (n = 3) and mutant (n = 3) P21 mice. Metabolite extraction and 1H nuclear magnetic resonance (NMR) spectroscopy of fecal metabolites was performed by Chenomx (Edmonton, AB, Canada) as described.37,38 Individual metabolite concentrations were normalized against total metabolite concentration in each sample.
From a univariate perspective, differences in alpha diversity, as well as the relative abundances of genera and phyla, were examined using two-way analysis of variance (anova) followed by Tukey post hoc analyses. Prior to analysis, relative abundances were transformed using a logit transformation by applying the methods of Warton and Hui.39 From a multivariate perspective, differences in UniFrac distances among groups were evaluated using distance-based redundancy analysis (dbRDA).40 For the dbRDA, first, distances among samples were calculated using UniFrac distances, and then an anova-like simulation was conducted to test for group differences. Fecal metabolite data were analyzed by Student’s t-test, because of the relatively simple experimental design that lacked multiple time points.
Samples and datasets
DNA was isolated from colon and feces of P7, P20, and P24 mutant and WT mice. Colon samples were obtained for the analysis of surface-associated microbial communities. Pyrosequencing yielded a total dataset of 708 242 sequences >200 bp in length. After performing quality control depletions as above, 449 252 sequences generated taxonomic information. An average of 8639 ± 4072 sequences per sample were utilized for taxonomic classification.
Sequence diversity within samples (alpha diversity)
Sequence diversity within each specimen revealed a wide range of bacterial species richness based on multiple biodiversity indices (Tables S1 and S2). Based on the number of operational taxonomic units (OTUs), a greater than 10-fold difference in richness was observed between the least and most diverse samples (Fig. 1). The Shannon index, which assesses species richness and evenness (i.e., the extent to which sequence reads within a sample are distributed across species), also exhibited a wide range of values (Tables S1 and S2). In contrast, species richness within biological replicates (mice of the same genotype and age) were highly similar, based on the standard error depicted in Fig. 1. This reproducibility allowed us to make statistically meaningful comparisons between mice of different genotypes and ages.
Colon samples Major differences were observed in microbiome diversity between WT and mutant colon at P7, with the number of OTUs significantly greater in mutant mice (P = 0.047; Fig. 1). Corresponding differences in the Shannon index at 3%, 5%, and 10% divergence (Supplementary Table S1) also identified greater species evenness (more prevalence of certain species) in mutant P7 colon. The increased species richness and evenness in P7 mutant colon indicates an increased overall variety of species, but a smaller number of species representing a greater percentage of the microbiome at this stage. Mutant diversity at P20 and P24 was not significantly greater than WT. Advancing age was associated with significant increases in microbiome diversity in WT colon (Supplementary Tables S1 and S2). The number of OTUs was much higher at P20 than P7 (Fig. 1; P = 0.003), and also higher at P24 than P20 (P < 0.001). The latter difference was also observed in mutant mice (P < 0.001).
Fecal samples The number of OTUs in mutant fecal samples was higher than in WT samples at both P7 (P = 0.063) and P20 (P = 0.006) (Fig. 1). Fecal microbiomes also showed an age-associated increase in OTU’s, with a significant increase between P7 and P20 for both WT (P < 0.001) and mutant (P < 0.001). The significant increase in WT colon between P20 and P24 was not observed in fecal samples.
Sequence-based relationships between samples (beta diversity)
Colon samples Multivariate analysis of UniFrac distances among colon sample groups using dbRDA showed significant differences associated with age (P = 0.01 unweighted; P = 0.01 weighted) and genotype (P = 0.01 unweighted; P = 0.02 weighted). PCoA of weighted and unweighted UniFrac distances (Fig. 2) was used to further interpret the relationships between WT and mutant microbiomes at each age. Non-overlapping 95% confidence intervals indicate statistically significant differences between samples. PCoA of unweighted UniFrac distances (Fig. 2A) revealed significant differences between all age groups. Segregation by genotype (WT v. mutant) was observed at P20. Weighted distances (Fig. 2B) corroborated the age-based differences, but also revealed significant genotype-based differences in all three age groups. With the exception of P20 mutant, the weighted confidence ellipses were tighter than those obtained from unweighted distances, suggesting that taxon abundance is an important predictor for both age- and genotype-based microbiome differences in colon samples.
Fecal samples dbRDA of UniFrac distance for fecal samples also showed significant overall differences in microbiome composition due to age (P = 0.01 unweighted, P = 0.01 weighted) and genotype (P = 0.02 unweighted; P = 0.01 weighted). Fecal PCoA analysis showed similar segregations as seen in the colon, except that unweighted distances yielded stronger separation in all three age groups (Fig. 2A). In contrast, weighted analysis (Fig. 2B) showed only a significant genotype-based difference at P20.
To identify specific taxa responsible for the microbiome differences identified, sequences were classified into 18 phyla, 33 classes, 76 orders, 168 families, and 444 genera. Sequences fell into three dominant phyla – Firmicutes, Bacteroidetes, Proteobacteria – and major differences between WT and mutant were observed (Fig. 3).
Colon samples In colon samples, mutant mice exhibited lower abundance of Firmicutes and more Bacteroidetes than WT at P24; however these trends were not statistically significant (P = 0.190 for Firmicutes, P = 0.923 for Bacteroidetes). Proteobacteria were increased in mutant compared to WT at P20 and P24, although these differences only approached significance (P = 0.088). We observed major age-related differences between P7 and P20 in both WT and mutant mice, with mean abundance of Firmicutes decreasing and Bacteroidetes increasing with age.
To identify genus-level differences in the colon, the relative abundance of the 50 dominant genera was visualized by heat-map analysis and dendrograms generated. Marked differences between younger (P7) and older (P20, P24) mice are readily visualized (Fig. 4A). P7 mice are dominated by fewer genera at higher abundance, whereas P20 and P24 mice contain more genera at lower relative abundance for each genus, a pattern consistent with the increased OTU diversity seen in older mice (Fig. 1). When genotype differences are considered, significant clustering of mutant and WT samples is observed. In P7 samples (Fig. 4A), one cluster is composed entirely of WT mice, while another contains only mutants. Notably, mutant P7 microbiomes are both richer and more even in genus-level composition than the corresponding WT samples (Fig. 4A,B), as seen at the species/OTU level (Fig. 1). In samples from older (P20, P24) mice, the dendrogram shows three major branches, one composed exclusively of mutant samples.
Further analysis of colon sequence data revealed significant genus-level differences between WT and mutant for five of the genera that had at least 4% relative abundance (Fig. 5). These differences were found at both P7 and P24. At P7, significant differences (P < 0.005) in relative abundance were observed for Staphylococcus and Lactobacillus, which were significantly more abundant in P7 mutant and P7 WT, respectively. At P24, Staphylococcus and Lactobacillus sequences are virtually absent, and genotype-based differences are seen in other genera. For example, fewer Clostridium sequences were found in mutant colon than in WT colon, with the opposite relationship seen for Coprobacillus and Bacteroides. We also examined genus-level differences as a function of age in WT mice alone. This revealed significant differences between P7 and older mice, with Lactobacillus more abundant in younger mice, and Clostridium more abundant in older mice (P < 0.005; Fig. 5).
Most sequences could not be assigned with confidence below genus level due to the limits of resolution of 16S rRNA sequences within certain genera and the likelihood of novel taxa in our samples. However, sequences could be assigned to the coagulase-negative staphylococci, particularly Staphylococcus xylosus and Staphylococcus lentus. In older mice, S. xylosus is virtually absent in WT and mutant. In contrast, in WT P7 colon, its relative abundance ranged from 2.2% to 11.9% (mean = 4.7 ± 4.1%; Table 1). Much higher abundance was observed in two mutant P7 colons (Mut-80 = 91.6% and Mut-81 = 85.2%). Interestingly, two mutant P7 colon samples with low abundance of S. xylosus had very high abundance of S. lentus (Mut-92 = 21.4% and Mut-93 = 59.1%). In the remaining mutant colon (Mut-91), S. lentus occurred at 4.6%, much higher than WT P7, where S. lentus was absent. In summary, all but one mutant P7 colon was dominated by one of two coagulase-negative staphylococcal species.
Table 1. Relative abundance (%) of 16S rRNA gene sequences attributed to Staphylococcus xylosus and Staphylococcus lentus in individual WT and mutant P7 mouse samples
Staphylococcus xylosus Colon
Staphylococcus xylosus Feces
Staphylococcus lentus Colon
Staphylococcus lentus Feces
Fecal samples Mutant fecal samples had fewer Firmicutes (P = 0.092) and more Bacteroidetes (P = 0.023) at P20, relative to WT, but these trends were not maintained at P24. Similarly, more Proteobacteria were found in mutant P24 feces than in WT P24 feces (not significant), but this difference was not observed at P7 or P20. Age-related changes in WT fecal phylum-level composition also resembled those seen in the colon, with similar changes in the abundance of Firmicutes and Bacteroidetes (Fig. 3).
Heat-map analysis of the 50 most abundant genera in fecal samples showed similar age- and genotype-related patterns of genus-level richness and evenness as in colon samples. P20 and P24 samples were richer and more even than P7 samples (Figs 1 and 4B), as were mutant P7 relative to WT P7 samples. P20 and P24 fecal samples exhibited more clusters than the corresponding colon samples, making genotype-specific clustering harder to discern than for colon (Fig. 4B). Significant differences were observed between mutant and WT feces in the relative abundance of Tannerella, which was much more abundant in P20 mutants (Fig. 5). Of note, we used the genus label Tannerella to refer to a genus-level taxon within family Prevotellaceae that could not be unambiguously assigned to one genus, but was most closely related to Tannerella and Barnesiella. Non-significant trends were observed for fewer Lactobacillus in mutant P20 and P24 feces, and more Coprobacillus in mutant P24 feces. The latter supports the significant genotype-based differences in Coprobacillus abundance in P24 colon. Unlike colon samples, fecal samples showed no enrichment of S. xylosus in mutants, although one mutant fecal sample was enriched for S. lentus, which was absent from WT feces (Table 1).
Fecal metabolite analysis
To identify functional correlates of the microbial differences identified, fecal metabolome analysis was performed by NMR spectroscopy of P21 samples. Significant differences between WT and mutant mice were identified for several metabolites (Table 2). A high degree of variation in metabolite concentrations between individuals of the same genotype prevented many of these trends from reaching statistical significance, although significant (P ≤ 0.05) genotype-based differences were observed for seven metabolites, including formate, a short-chain fatty acid exclusively produced by microbial fermentation, which was significantly lower in mutant samples.
Table 2. Mean concentrations (μmol L−1) of fecal metabolites found to differ significantly (P ≤ 0.05; Student’s two-tailed t-test) between wild-type (WT) and mutant (Mut) mice
WT (mean ± SD)
Mut (mean ± SD)
0.91 ± 0.49
0.16 ± 0.28
8.53 ± 3.24
3.50 ± 1.48
1.17 ± 0.46
2.02 ± 0.46
0.68 ± 0.32
2.44 ± 1.08
1.48 ± 0.18
1.05 ± 0.18
0.13 ± 0.03
0.03 ± 0.02
1.32 ± 0.11
1.85 ± 0.30
An important feature of HAEC is that it can occur before or after removal of the aganglionic segment and, when it occurs prior to surgery, it can involve the aganglionic and normoganglionic colon,41 suggesting a broader defect not limited to the aganglionic segment. We therefore hypothesized that aganglionosis leads, either directly or indirectly, to early disruptions in colonic microbiome composition with long-term effects that persist after removal of the aganglionic bowel. Although others have identified potential microbial changes associated with Hirschsprung’s disease,17–19 a comprehensive microbiome comparison has not previously been undertaken.
The EdnrB mutant mouse was used as it is phenotypically similar to human Hirschsprung’s disease, with homozygous mutants exhibiting distal colorectal aganglionosis. Mice die at 4–5 weeks of age from enterocolitis, manifesting as lethargy, anorexia, shivering, and weight loss.42 The time points examined in this study were chosen to encompass disease progression and weaning, with P7 mice exclusively suckling and P20 mice suckling and having access to solid food. At P24, mice are completely weaned and close to the usual time of HAEC onset. Mice in this study showed no clinical sign of enterocolitis, as our aim was to identify microbiome changes predisposing to HAEC rather than those resulting from it. High-throughput 16S rRNA gene pyrosequencing allowed the characterization of alpha and beta diversity, and determination of the taxonomic affiliation of sequences contributing to differences between mouse groups based on age and genotype.
Maturation of the normal mouse colonic microbiome
Surprisingly, few methodologically comparable studies examine normal maturation of the colonic microbiome. Previous work in rodents utilized profiling methods such as gradient gel electrophoresis43 or real-time PCR and culturing.44 Pyrosequencing-based characterization of human infant microbiome development provides data for comparison. The mouse microbiomes in our study were dominated by Firmicutes and Bacteroidetes, consistent with previous studies in humans,45–48 mice,9 and other mammals.49 Dramatic differences in microbiome structure were observed between younger (P7) and older (P20 and P24) mice, with a significant increase in species richness and evenness at older ages, as described previously in mice43 and human infants.50 The highest numbers of OTUs (1200–1500) occurred in P24 colon, a degree of richness similar to the richest samples in fecal microbiomes from diverse mammals.49 We also observed large age-related differences in microbiome composition, with a significant Firmicute dominance at P7 followed by increasing Bacteroidetes abundance in older mice. Similar changes have been described in human infants, where the transition to an ‘adult-like’ microbiome occurs with addition of solid food to the nursing infant.50 Although P7 mice only suckle, by P20 they supplement with mouse chow, hence differences between these ages may be partly attributed to diet. Despite the temporal variation in colonization profile among human infants, Bacteroidetes were present to some extent in nearly all children by 1 year of age,6 suggesting that the acquisition of Bacteroidetes is a feature common to microbiome maturation in both mice and humans.
At the genus level, age-associated differences in microbiome composition were characterized by fewer Lactobacillus in older mice, which exhibited higher abundance of Clostridium. In our study, dietary changes may have favored reduction in Lactobacillus, which are specialized for lactose metabolism, whereas Clostridium may proliferate in older mice as sugars derived from breakdown of dietary plant polysaccharides become available for fermentation. Bifidobacteria, long considered the dominant flora of the breast-fed infant,51 were not detected in large numbers in our dataset, consistent with recent results in human infants,6 suggesting that the importance of early bifidobacteria colonization may be over-emphasized.
Differences between WT and mutant microbiomes
Our results support the hypothesis that aganglionosis disrupts the normal colonic microbiome, suggesting that the ENS is an important determinant of microbiome composition. We found that mutant colon samples exhibit greater species richness at P7, suggesting that aganglionosis creates an environment supporting microbiome diversification, perhaps through the provision of new chemical niches.52 Analysis of OTU-level composition showed segregation of WT and mutant microbiomes at all ages examined. Mutant mice thus exhibited not only a greater number of microbial species, but their composition was significantly different. These results contribute to the growing evidence on the influence of intrinsic host factors on microbiome composition,7–9 adding the ENS to the list of regulatory elements.
Some of the differences observed could be attributed to differences in particular taxonomic groups. At the phylum level, decreases in the abundance of Firmicutes and increases in Bacteroidetes were observed in mutant colon and feces. These differences at P20 and P24, ages close to the time of HAEC onset, offer possible clues to the etiopathogenesis of HAEC. Shifts in these phyla have been reported in inflammatory bowel disease, where depletion of Firmicutes47,53 and increases53 or decreases47 in Bacteroidetes have been described. Reduction in Firmicutes and proliferation of Bacteroidetes are also seen in diabetes54 and alcoholic liver disease,55 with the opposite trend observed in patients with a high-fat diet56 or obesity.46,48 At the genus level, significant increases in the abundance of Coprobacillus and Bacteroides were identified in mutant P24 colon, suggesting additional bacterial groups that may be associated with the development of HAEC.
Mutant P7 colon exhibited a marked reduction in Lactobacillus, consistent with findings in human Hirschsprung’s disease,17 and a significant increase in Staphylococcus. These changes may be due to local environmental differences in aganglionic colon. Lactobacillus species bind mucus57 and displace pathogens through this binding.58 Hirschsprung’s-associated decreases in mucin production14 may affect the ability of Lactobacillus to colonize, permitting Staphylococcus to proliferate in their place. Although few of the genus-level differences could be traced to species level, dramatic enrichment of coagulase-negative staphylococci was observed in mutant P7 mice. Staphylococcus xylosus is considered a non-pathogenic commensal of rodent skin,59 but can cause spontaneous infections in mice deficient in phagocyte superoxide production,60 and spontaneous keratitis in mice with abnormal corneal epithelial integrity.61Staphylococcus lentus, frequently isolated from animals, can be a pathogen in immunocompromised patients.62 The remarkable dominance of these species in our mutant samples suggests a potential contribution to HAEC onset. Although their absence at later ages precludes a direct role in promoting HAEC, they may contribute to a milieu favoring colonization of HAEC-promoting organisms.
Previous studies suggest the utility of fecal microbiomes as proxies for the colonic microbiome, although fecal–mucosal differences have been found to contribute substantial variation in microbiome data.45 Because the relationship between colonic and fecal microbiomes in Hirschsprung’s disease has not been reported, we chose to study both sample types. We found clear differences in colonic and fecal microbial composition at phylum and genus levels, and stark differences in the abundance of coagulase-negative staphylococci. Importantly, the most significant genus-based differences at P24, the age most relevant to HAEC, were observed in colon, not fecal, samples. It is not clear whether colonic or fecal microbiomes are most relevant to HAEC onset, but our results emphasize the importance of considering both when examining the contribution of the microbiome to intestinal pathology.
Fecal metabolite profiling revealed major differences in metabolic activity between WT and mutant mice. We found a significantly lower concentration of formate in P21 mutant samples compared with WT, suggesting differences in microbial fermentation pathways. Short-chain fatty acids, such as formate, have important roles in maintaining colonic mucosal health, and an association between reduced formate and human intestinal inflammation has been reported.63,64 The altered microbiome in aganglionic mice may alter microbial fermentation, predisposing mutants to enterocolitis, either directly through an effect of formate levels on inflammation or indirectly through the action of unidentified metabolic sequelae to formate reduction.
Identifying the microbiome changes associated with aganglionosis provides a framework for future studies to determine the role of specific microbial populations in inducing inflammation in HAEC. Such studies will provide a basis for early intervention to reduce HAEC risk in Hirschsprung’s disease, and have broader implications for understanding the relationship between the ENS and microbiome maturation.
This research was supported by the United States National Science Foundation (EPS-0447681).
No competing interests declared.
All of the sequence data obtained in this study will be made available to any investigator upon request.
NLW performed data analysis and interpretation, and contributed significantly to writing the manuscript; AP performed all of the mouse work, including breeding, genotyping, isolation of colon and feces; SED and SBC performed pyrosequencing and data analysis, and contributed to writing the manuscript; AMG oversaw all of the research, supervised the work of AP, and worked with NLW on interpreting the data and writing the manuscript.