Colonic mucosal microbiota is associated with bowel habit subtype and abdominal pain in patients with irritable bowel syndrome

Abstract

Mucosal microbiota differ significantly from fecal microbiota and may play a different role in the pathophysiology of irritable bowel syndrome (IBS). The aims of this study were to determine if the composition of mucosal microbiota differed between IBS, or IBS bowel habit (BH) subtypes, and healthy controls (HCs). Sigmoid colon mucosal biopsies were obtained from 97 Rome-positive patients with IBS (28% IBS-constipation, 38% IBS-diarrhea, 24% IBS-mixed, and 10% IBS-unsubtyped) and 54 HCs, from which DNA was extracted. 16S rRNA gene sequencing and microbial composition analysis were performed. Group differences in α and β diversity and taxonomic level differences were determined using linear regression while controlling for confounding variables. IBS BH subtype was associated with microbial α diversity (P = 0.0003) with significant differences seen in the mucosal microbiota of IBS-constipation versus IBS-diarrhea (P = 0.046). There were no significant differences in α or β diversity in the mucosal microbiota of IBS versus HCs (P = 0.29 and 0.93, respectively), but metagenomic profiling suggested functional differences. The relative abundance of Prevotella_9 copri within IBS was significantly correlated with increased abdominal pain (r = 0.36, P = 0.0003), which has not been previously reported in IBS. Significant differences in the mucosal microbiota were present within IBS BH subtypes but not between IBS and HCs, supporting the possibility of IBS BH subtype-specific pathogenesis. Increased Prevotella copri may contribute to symptoms in patients with IBS.

NEW & NOTEWORTHY Gut mucosal microbiota differs significantly from fecal microbiota in irritable bowel syndrome (IBS) and may play a different role in its pathophysiology. Investigation of colonic mucosal microbiota in the largest cohort of patients with IBS and healthy controls accounting for confounding variables, including diet demonstrated significant differences in mucosal microbiota between IBS bowel habit subtypes but not between IBS and healthy controls. In addition, the study reported gut microbiota is associated with abdominal pain in patients with IBS.

INTRODUCTION

Irritable bowel syndrome (IBS) is a multifactorial disorder of brain-gut interactions with a prevalence of up to 11% (1) that has been associated with altered gut microbiota (2, 3). IBS can be categorized into subgroups based on the predominant bowel habits (BHs): diarrhea-predominant [IBS with diarrhea (IBS-D)], constipation-predominant [IBS with constipation (IBS-C)], mixed [IBS with mixed bowel habit subtype (IBS-M)], and unsubtyped [IBS unsubtyped (IBS-U)]. Several studies have found differences in the fecal microbiota based on the IBS BH subtype, especially between IBS-C and IBS-D, but no consistent trends have been demonstrated (4). Bowel habit-associated dysbiosis may have pathophysiological functional consequences in IBS. For example, IBS-D has been associated with dysbiosis in some bacterial groups involved in bile acid transformation, and increased bile acids have been correlated with increased stool frequency in patients with IBS, suggesting a possible role for these bacteria in IBS-D pathophysiology (5). In addition, IBS-C is associated with higher methane production and an increased abundance of methanogenic archaea, suggesting a functional role for bacterial dysbiosis in patients with IBS-C (6).

A majority of studies have focused on fecal microbiota since it is easier to obtain samples in a noninvasive manner, but there is increasing evidence that mucosal and fecal microbiota differ significantly from one another (7, 8). Fecal microbiota represents the bacterial communities that exist throughout the gastrointestinal (GI) tract, whereas mucosal microbiota represents site-specific microbiota that likely plays an important role in IBS pathology due to its close contact with host epithelial and immune cells (9). Both fecal and mucosal microbiota studies are characterized by heterogeneity in methodology, resulting in mixed and often inconsistent results. No consistent trends have been found in mucosal microbiota studies so far, partly due to the limited number of studies.

Mucosal microbiota studies tend to have a small sample size, focusing on the IBS-D bowel habit subtype or female sex, and usually consist of a single ethnic/racial group, limiting the applicability to a wider population of patients with IBS (Supplemental Table S1; all Supplemental material is available at https://doi.org/10.6084/m9.figshare.19747606). In addition, many studies lack descriptions of probiotic use and dietary intake that can alter the gut microbiota. Mucosal biopsy samples were most often taken from the sigmoid colon but were also taken from other sites in the small and large intestine (8, 10).

The overall goal of our study was to investigate the mucosal microbiota in a larger cohort of patients with IBS and healthy controls (HCs), while adequately controlling for confounding factors to better understand the role of altered gut mucosal microbiota in the pathophysiology of IBS. The study aims were to 1) compare the mucosal microbiota between patients with IBS and HCs, 2) study the effect of IBS BH subtype on mucosal microbiota, 3) measure correlations between the mucosal microbiota and IBS symptoms, and 4) assess predicted functional differences in the mucosal microbiota in patients with IBS.

METHODS

Study Population

Patients with IBS and HCs, ages 18–55 yr, were recruited by community advertisement. The diagnosis of IBS and BH subtypes [IBS with diarrhea (IBS-D), IBS with constipation (IBS-C), IBS with mixed bowel habit subtype (IBS-M), and IBS unsubtyped (IBS-U)] were based on a medical history, physical examination, and symptom-based Rome criteria (11). HCs had no personal or family history of IBS or other chronic pain conditions. Additional exclusion criteria for all participants included infectious or inflammatory disorders, active psychiatric illness over the past 6 mo assessed by structured clinical interview for the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition, Text Revision (Mini International Neuropsychiatric Interview; 12). Excluded medications included no corticosteroids for 6 mo, no antibiotics or narcotics for 3 mo, and no probiotics for 1 mo before the procedure. Tobacco and alcohol use was prohibited for 12 h before the study. Participants were excluded if they smoked more than one-third pack per day or had met the criteria for current alcohol or substance abuse or dependence. We did not exclude patients who were on proton pump inhibitors (PPIs). Only two patients reported taking omeprazole or pantoprazole at the time of the procedure. The University of California, Los Angeles Institutional Review Board approved the study, and participants signed a written informed consent before the study. Participants were compensated.

Questionnaires

In addition to Rome diagnostic symptom criteria for IBS (13), overall severity, abdominal pain, and bloating over the prior week were assessed with numeric rating scales (range 0–20; 14). Current anxiety and depression symptoms were measured with the hospital anxiety and depression scale (HAD, anxiety, range 0–21, depression, range 0–21; 15). These questionnaires are validated, widely used instruments and include instructions to on how to complete them.

Data pertaining to diet was collected using a diet checklist developed by our institution with the intention to represent the diet or diets that best reflect what individuals consume on a regular basis, as described previously (16). The questionnaire includes choices for a standard or modified American, Mediterranean, vegan, vegetarian, gluten-free, dairy-free, and/or a low fermentable oligosaccharides, disaccharides, monosaccharides, and polyols (FODMAP) diet. The diets were classified as standard or restrictive diets depending on the elimination of a certain food group(s) or components by choice. Restrictive diets were likely initiated to reduce GI symptoms and included gluten-free, dairy-free, and low FODMAP diets. The diet survey was validated in a subset of patients using the food frequency questionnaire (FFQ) and diet diary, as described previously (16).

Statistical Analysis for Clinical and Diet Data

All the analyses were performed using R software (17). Group comparisons for clinical, demographic, and dietary factors were performed for IBS and HCs, within IBS BH subtypes, and IBS BH subtypes and HCs, using general linear models for continuous variables with group as an independent factor variable, and Fisher’s test for categorical variables.

Mucosal Microbiota Collection and Data Generation

Mucosal biopsy collection and storage.

Mucosal biopsies were collected at 30 cm from the anal verge during sigmoidoscopy after bowel preparation with tap water enemas. Biopsies were snap-frozen in liquid nitrogen and stored at −80°C. All participants were instructed to hold aspirin and nonsteroidal anti-inflammatory drugs (NSAIDs) for 72 h before the procedure.

DNA extraction and 16S rRNA gene sequencing.

Bacterial DNA from biopsies was extracted using PowerSoil DNA Isolation Kit (MO BIO Laboratories) with bead beating, following the manufacturer’s protocol. The 16S rRNA gene sequence libraries were generated using the V4 hypervariable region, amplified using the 515F and 806R primers (18), and sequenced on the Illumina HiSeq 2500 platform (Life Sciences, Branford, CT). The raw reads were processed using DADA2 pipeline in R (19). Briefly, the raw data were demultiplexed, and primers, adapters, and linker sequences were eliminated. Silva v132 reference database was used to eliminate chimeras and assign taxonomy.

Predicted Metagenomic Profiling

Metagenomic data was predicted from 16s rRNA sequencing using PICRUSt2 through the QIIME2 platform with default parameters (20). The differences in subsequent predicted bacterial gene abundances between IBS and HCs and between IBS BH subtypes were then analyzed using DESEq2 with P values adjusted for multiple hypothesis testing (21).

Statistical/Bioinformatics Analysis of Microbiota Data

General linear model was used to identify confounding variables that were significantly associated with microbial β diversities. Group differences in α diversity (richness or bacterial diversity within a sample) and β diversity (evenness or bacterial diversity between samples) were assessed using a general linear model while controlling for potentially confounding variables, including age, body mass index (BMI), HAD anxiety scores, and sex as continuous variables, and race, restrictive diet, and sequencing batch as factor variables. Total cohort of subjects (n = 151) was used to analyze microbiome data. Missing data on covariates were handled using frequent category imputation for categorical data and mean values for continuous data. We imputed the restrictive diet variable for 37 patients with IBS and 26 healthy controls that had no diet data when using it as a covariate, to maximize the power. Prevalence of consuming a standard diet was ∼90%. Since only a minority of participants were on a restrictive diet, we believe most of the participants were consuming a nonrestrictive diet, minimizing the effect of imputation. Significant findings were reanalyzed and confirmed using nonimputed data wherever possible. For testing the β diversity group differences between diets, a subset of samples with diet variables was used.

α Diversity metrics included Chao1 and Shannon index. β Diversity analysis was performed using DEICODE, a form of Aitchison distance that is robust to high levels of sparsity, implemented in Quantitative Insights Into Microbial Ecology (QIIME) environment (22). Adonis, a permutational analysis of variance, was used to test for associations with demographic characteristics and clinical factors. Hierarchical clustering on distance matrix was used to identify microbiome-based clusters within subjects. For α diversity calculations only, the samples were rarefied to even depth for calculation of α diversity measures using “rarefy_even_depth” function. The read depth [mean (standard deviation)] was 29,709 (33,526.25) and rarefied to 538 counts. α Diversity was also calculated using unrarefied data to avoid data loss (23), and relative abundance (RA) was used for β diversity and multivariate analysis and calculated using “transform_sample_counts” in “phyloseq” package in R. Counts were analyzed at various taxonomic levels, including phylum, family, genus, species, and amplicon sequence variants (ASVs), to comprehensively identify major and minor shifts in microbial abundance. ASVs are exact sequence variants (24) compared with operational taxonomic units (OTUs), which are clusters of reads that are 97% similar (25).

Data were prefiltered to exclude any samples with all 0’s and low abundance bacteria (counts < 10). PhyloSeq package was used for preprocessing and visualization. Differential abundance analysis was performed using DESeq2, which uses a Bayesian approach to fit nonrarified count data to a negative binomial model (21). P values were corrected for multiple testing using Benjamini-Hochberg (26) method implemented within DESeq2.

Correlations between IBS Symptoms and Bacteria

To reduce the dimensionality while analyzing correlations between clinical variables and microbiota data in a multivariate analysis, sparse partial least squares (sPLS) regressions were performed, which use variable selection achieved by introducing LASSO penalization pairs of loading vectors (27). Filtered ASV data were converted into relative abundances and the residuals obtained after covarying for age, sex, BMI, race, sequencing batch, restrictive diet, and HAD anxiety were used as input. Tuning parameters were selected using “tune.spls” function from mixOmics package. First three components were included in the analysis. In addition, the relationship of bacterial genera and species with clinical features selected by the model was tested separately using partial correlation analysis, controlling for covariates including, age, sex, BMI, race, sequencing batch, restrictive diet, and HAD anxiety to visualize and quantify the significance of associations. A P value < 0.05 was considered significant.

RESULTS

Clinical Characteristics

We analyzed colonic mucosal microbiota of 151 individuals, including 97 patients with IBS (28% IBS-C, 38% IBS-D, 24% IBS-M, and 10% IBS-U) and 54 HCs. Baseline characteristics are outlined in Table 1. Patients with IBS-M and IBS-U were combined and referred to as IBS-M in this study. Age, sex, BMI, race, and education were similar in patients with IBS and HCs (P > 0.05). Compared with HCs, IBS showed increased current anxiety and depression scores [t = 6.7 and 4.03, respectively, degrees of freedom (df) =147, P < 0.05], however, the mean depression scores were low (noncase/nonclinical depressive states) in both groups [mean (SD): IBS = 3.37 (3.51) and HCs = 1.26 (1.93)]. Overall symptom severity of IBS was moderate (Table 2). Medication use was reported by 42% of participants (IBS, 54%, HCs, 22%) and was significantly different between IBS and HCs. Most participants used laxatives or antidiarrheal drugs (IBS only), and/or nonsteroidal anti-inflammatory drugs only on an as-needed basis. None of the participants were taking probiotics or antibiotics. Diet data were available on 65 patients with IBS and 29 HCs. The subgroup with diet data had similar demographic characteristics as the overall group (P > 0.05; Supplemental Tables S2 and S3).

Table 1.

Demographic characteristics of the study population

	Total IBS (n = 97)	HCs (n = 54)	Statistic	P Value	IBS-C (n = 27)	IBS-D (n = 37)	IBS-M + U (n = 33)	Statistic	P Value
Age, mean (SD)	31.85 (11.04)	32.09 (11.02)	−0.13#	0.89	30.67 (11.22)	32.49 (11.6)	32.09 (10.52)	0.65#	0.52
Sex F, n (%)	63 (65%)	26 (48%)	3.38 (1)$	0.07	19 (70%)	20 (54%)	24 (73%)	3.15 (2)$	0.23
BMI, mean (SD)	25.37 (4.64)	25.36 (3.68)	0.02#	0.98	24.12 (2.87)	25.35 (5.78)	26.42 (4.21)	1.05#	0.30
Race, n (%)			0.5 (3)$	0.92				9.55 (6)$	0.15
White	51 (57%)	32 (51%)			13 (48%)	18 (50%)	17 (55%)
Asian	20 (20%)	10 (21%)			7 (26%)	8 (22%)	5 (15%)
Black	10 (10%)	5 (11%)			6 (22%)	2 (6%)	2 (6%)
Multiracial	16 (14%)	7 (17%)			1 (4%)	8 (22%)	7 (21%)
Education level, n (%)			5.2 (3)$	0.16				6.51 (6)$	0.37
High school	2 (2%)	4 (8%)			0 (0%)	1 (3%)	1 (3%)
Some college	47 (49%)	21 (41%)			13 (48%)	20 (54%)	14 (44%)
College graduate	19 (20%)	15 (29%)			9 (33%)	5 (13%)	5 (16%)
Post-graduate work	28 (29%)	11 (22%)			5 (19%)	11 (30%)	12 (37%)
HAD anxiety (0–21), mean (SD)	7.08 (4.53)	2.57 (2.51)	6.71#	<0.001	6.07 (4.63)	7.44 (5.19)	7.52 (3.57)	1.19#	0.24
HAD depression (0–21), mean (SD)	3.37 (3.51)	1.26 (1.93)	4.03#	<0.001	3.19 (3.45)	3.67 (3.31)	3.18 (3.83)	0.54#	0.59
Medications, n (%)
Antidepressants	12 (12%)	0		0.004&	3 (11%)	5 (14%)	4 (12%)		1&
Benzodiazepines	3 (3%)	0		0.55&	2 (7%)	1 (3%)	0 (0%)		0.28&
Laxatives	6 (6%)	0		0.09&	5 (18%)	0 (0%)	1 (3%)		0.006&
Anti-diarrheal agents	11 (11%)	0		0.008&	1 (4%)	5 (14%)	5 (15%)		0.34&
Fiber supplements	5 (5%)	1 (2%)		0.42&	3 (11%)	0 (0%)	2 (6%)		0.09&
NSAIDs as needed	33 (33%)	11 (20%)		0.09&	10 (37%)	10 (27%)	13 (40%)		0.51&
PPIs	2 (2%)	0		0.54&	0	1 (1%)	1 (1%)		1&

HAD, hospital anxiety and depression scale; IBS, irritable bowel syndrome; HCs, healthy controls; IBS-C, irritable bowel syndrome with constipation; IBS-D, irritable bowel syndrome with diarrhea; IBS-M, irritable bowel syndrome with mixed pattern; IBS-U, irritable bowel syndrome unsubtyped; NSAID, nonsteroidal anti-inflammatory drugs; PPIs, proton pump inhibitors; n, number of participants; #t statistic; $, χ² statistic (degrees of freedom); &P value from Fisher’s exact test.

Table 2.

IBS symptom severity in patients with IBS

IBS Symptoms,Mean (SD)	Total IBS (n = 97)	IBS-C (n = 27)	IBS-D (n = 37)	IBS-M + U (n = 33)	t-Statistic	P Value
Overall IBS severity (0–20)	9.09 (4.31)	7.42 (4.64)	10.65 (3.74)	8.67 (4.17)	3.04	0.003
Abdominal pain (0–20)	8.96 (4.11)	7.81 (4.19)	9.73 (4.03)	9 (4.05)	1.84	0.07
Bloating (0–20)	10.72 (4.73)	11.48 (4.24)	11.24 (4.57)	9.52 (5.19)	−0.2	0.84
Usual severity (1–5)	3.14 (0.63)	2.89 (0.59)	3.16 (0.62)	3.303 (0.64)	1.78	0.08
VSI score (0–75)	35.19 (16.29)	34.82 (14.91)	34.76 (16.97)	35.97 (17.04)	−0.014	0.99
IBS SSS (0–500)	212.90 (87.50)	215.75 (89.41)	219.04 (75.59)	204.76 (99.89)	0.117	0.91

IBS, irritable bowel syndrome; IBS-C, irritable bowel syndrome with constipation; IBS-D, irritable bowel syndrome with diarrhea; IBS-M, irritable bowel syndrome with mixed pattern; IBS SSS, irritable bowel syndrome severity scoring system; IBS-U, irritable bowel syndrome unsubtyped; VSI, visceral sensitivity index; n, number of participants.

Association of Mucosal Microbiota with Characteristics of the Participants

Microbial β diversity of the mucosal microbiota was significantly associated with demographic factors including age, sex, BMI, and race, and other factors including sequencing batch and restrictive diet (P < 0.05, Supplemental Tables S4 and S5). Within IBS, age, sex, sequencing batch, and restrictive diet were significantly associated with microbial β diversity (P < 0.05, Supplemental Tables S4 and S5). However, the consumption of a restrictive diet was consumed by a minority of study participants and was not significantly different between patients with IBS-C versus patients with IBS-D (21% vs. 17%, P = 0.76).

Mucosal Microbiota in Patients with IBS Compared with Healthy Controls

There was no association between IBS status and microbial α or β diversity after controlling for confounding variables (α diversity: t = −1.06, P = 0.29 and β diversity: F = 0.11, P = 0.93, respectively, Supplemental Fig. S1, A and B). There was no significant association between IBS and mucosal microbiota subgroups. Taxonomic analysis revealed small but statistically significant taxonomic shifts in the microbiota of IBS compared with HCs, including enrichment of Campylobacter ureolyticus (species) and Finegoldia magna (species) [false discovery rate (FDR) = 1.52·e⁻¹⁰, and 0.02, respectively (Supplemental Fig. S1C)]. Unsupervised clustering did not reveal an association between IBS status and mucosal microbiome-based subgroups [number of clusters (K) = 4, P = 0.48, Supplemental Fig. S2].

Association of Mucosal Microbiota with IBS Bowel Habit Subtypes

We compared microbial α and β diversities between the three IBS BH subtype groups and between IBS BH subtypes and HCs. The microbial α diversity was significantly different between BH subtypes (F = 6.72, P = 0.0003, Fig. 1A). This difference in the α diversities persisted when repeated on nonrarefied or nonimputed data (P = 2e-05 and P = 0.007, respectively). Pairwise analysis identified decreased α diversity in patients with IBS-C compared with patients with IBS-D (t = 2.02, df = 137, P = 0.046) and a trend for decreased α diversity in IBS-C compared with HCs (t = 1.78, df = 137, P = 0.077). However, there was no difference in the β diversity between BH subtypes (F = 0.3, P = 0.93, Fig. 1B).

Figure 1.

α And β diversity and taxonomic differences in mucosal microbiota of IBS bowel habit subtypes. A: microbial α diversity, measured by Chao1 estimator, was significantly associated with IBS bowel habit subtypes (P = 0.0003). The Y-axis shows residual Chao1 estimates after regressing out the effect of covariates. B: microbial β diversity was not significantly different within IBS bowel habit subtypes and compared with HCs (P = 0.93). At the genus level, seven microbes were significantly different between IBS-C and HCs (FDR < 0.05; C), three microbes were significantly different between IBS-D and HCs (FDR < 0.05; D), and seven microbes were significantly different between IBS-D vs. IBS-C in the mucosal microbiota (FDR <0.05; E). FDR, false discovery rate; HCs, healthy controls; IBS, irritable bowel syndrome; IBS-C, irritable bowel syndrome with constipation; IBS-D, irritable bowel syndrome with diarrhea. IBS-C (n = 27), IBS-D (n = 37), IBS-M and U (n = 33), HC (n = 54).

Taxonomic analysis identified significant shifts in RA of microbes within BH subtypes compared with HCs. Most differences associated between BH subtypes and HCs were observed in IBS-C, where seven taxa were identified at the genus level to be significantly different (Table 3). Compared with HCs, the microbiota of IBS-C showed depletion of high abundance bacterium, Erysipelotrichaceae_UCG-003 (RA = 0.2%, FDR = 3.78E-03), moderately abundant bacteria Acinetobacter and Fournierella (RA = 0.01 and 0.02%, FDR < 0.05), and low abundance bacteria Bradyrhizobium and Aneroplasma (RA < 0.005%, FDR < 0.05, Table 3 and Fig. 1C). IBS-C had increased levels of Fournierella and Anerococcus compared with HCs (RA < 0.005%, FDR < 0.05, Table 3 and Fig. 1C). Compared with HCs, the mucosal microbiota of patients with IBS-D was associated with a significant increase of Prevotellaceae_NK3B31_group (phylum Bacteroidetes; RA = 0.019%, FDR = 1E-08) and Fournierella (phylum Firmicutes; RA = 0.005%, FDR = 4.4E-05), and a decrease in genus Anaerofilum (phylum Firmicutes; RA = 0.016% and FDR = 0.025, Fig. 1D).

Table 3.

Differentially abundant bacteria in IBS bowel habit subtypes

Comparison Groups	Phylum	Genus	Fold Change	P Value	FDR	Relative Abundance %
IBS-C vs. HCs
	Firmicutes	Fournierella	29.97	2.02E-14	1.10E-12	0.020
	Proteobacteria	Acinetobacter	−11.69	2.82E-03	3.85E-02	0.011
	Proteobacteria	Bradyrhizobium	−21.50	4.10E-08	1.12E-06	0.002
	Firmicutes	Anaerococcus	15.70	6.12E-05	0.001	0.007
	Tenericutes	Anaeroplasma	−30.00	2.00E-14	1.10E-12	0.001
	Firmicutes	Allisonella	−15.05	1.22E-04	2.23E-03	0.009
	Firmicutes	Erysipelotrichaceae_UCG-003	−4.04	2.43E-04	3.78E-03	0.261
IBS-D vs. HCs
	Bacteroidetes	Prevotellaceae_NK3B31_group	30	<1E-08	<1E-08	0.019
	Firmicutes	Anaerofilum	−12.5	6.86E-04	0.025	0.016
	Firmicutes	Fournierella	18.2	8E-07	4.4E-05	0.005
IBS-M + U vs. HCs
	Bacteroidetes	Muribaculum	−21.21	5.93E-07	6.34E-05	0.019
IBS-C vs. IBS-D
	Elusimicrobia	Elusimicrobium	−16.49	2.69E-06	9.86E-05	0.050
	Proteobacteria	Acinetobacter	−10.55	2.67E-03	4.20E-02	0.035
	Proteobacteria	Bradyrhizobium	−30.00	1.59E-17	1.75E-15	0.010
	Bacteroidetes	Prevotella	−20.16	1.02E-08	5.63E-07	0.004
	Bacteroidetes	Muribaculum	−14.17	5.22E-05	1.15E-03	0.088
	Proteobacteria	Desulfovibrio	−13.01	1.14E-05	3.14E-04	0.074
	Firmicutes	UC5-1-2E3	−12.52	3.61E-04	6.62E-03	0.021

IBS-C, irritable bowel syndrome with constipation (n = 27); IBS-D, irritable bowel syndrome with diarrhea (n = 37); IBS-M, irritable bowel syndrome with mixed pattern, IBS-U, irritable bowel syndrome unsubtyped (IBS-M and IBS-U, n = 33), FDR, false discovery rate.

Within IBS BH subtypes, Acinetobacter and Bradyrhizobium were decreased in IBS-C compared with IBS-D, which were also found to be decreased in IBS-C compared with HCs. Other microbes that were increased in IBS-D compared with IBS-C include Prevotella, Desulphovibrio, and Mucribaculum (FDR < 0.05, Table 3 and Fig. 1E). The relative abundance of IBS-C versus IBS-D associated taxa was varied among patients and was represented in a minority of patients (Supplemental Fig. S3).

Mucosal Microbiota and IBS Symptoms

Sparse PLS regression of microbial genera and ASVs onto clinical characteristics identified bacteria associated with clinical symptoms. At the genus level, the bacteria selected on component 1 by the multivariate model included, Prevotella_9, Clostridium_sensu_sricto_1, Allisonella, and Mitsuokella that were positively correlated with IBS symptoms including overall severity, bloating, and abdominal pain, whereas Bacteroides, Parasutterella, and Parabacteroides were negatively correlated with IBS symptoms (Fig. 2). At the ASV level, sparse PLS analysis revealed a positive correlation between IBS symptoms and microbes, including Prevotella_9 copri (species), Lachnoclostridium (genus), and Clostridium_sensu_sricto_1 (genus), and a negative correlation between IBS symptoms and Bacteroides vulgatus (species) and Bacteroides uniformis (species) (Supplemental Fig. S4).

Figure 2.

Heatmap of correlations between mucosal microbiota and demographic and clinical characteristics of patients with IBS (n = 97). The correlation heatmaps show microbiota selected by sparse partial least squared analysis (sPLS) at genus level that were positively or negatively correlated with IBS symptoms, including abdominal pain, bloating, and overall symptom severity. X-axis represents IBS symptom severity measures, and Y-axis represents microbiota. Red color indicates a positive correlation and blue indicates a negative correlation. IBS, irritable bowel syndrome.

In addition, a canonical correlation analysis between Prevotella-9 genus and severity of abdominal pain and a correlation between P. copri and severity of abdominal pain revealed significant association between these bacteria and IBS symptoms (genus level: covariate-adjusted P = 0.03, r = 0.21, Supplemental Fig. S5A, species level: covariate-adjusted P = 0.003, r = 0.30, Supplemental Fig. S5B).

Predicted Metagenomic Functional Pathways in the Microbiota of IBS and HCs

Predictive metagenomic profiling of 16S rRNA data identified enrichment of genes associated with phenylalanine metabolism (fold change = 4.5, FDR < 0.05) and a downregulation of genes associated with pathways including “Transporters,” “Ether lipid metabolism,” and “Arginine biosynthesis,” in the mucosal microbiota of IBS compared with HCs (Fig. 3). Regarding IBS BH subtype differences, we found significant differences only between IBS-D and IBS-M, which included downregulation of pathways such as “Prokaryotic defense system” and “DNA repair and recombination proteins” in the microbiota of IBS-D compared with IBS-M (Supplemental Fig. S6).

Figure 3.

Metagenomic functional pathways associated with mucosal microbiota in IBS. Predictive metagenomic profiling of 16S rRNA sequencing of the mucosal microbiota identified enrichment of genes associated with “phenylalanine metabolism” in patients with IBS (n = 97) compared with HCs (n = 54). HCs, healthy controls; IBS, irritable bowel syndrome.

DISCUSSION

Mucosal microbiota in patients with IBS is relatively less studied compared with fecal microbiota. To our knowledge, this is the largest mucosal microbiota study in IBS that included all BH subtypes while controlling for the confounding effects of diet and several other relevant demographic factors, enabling the identification of microbes relevant to IBS pathophysiology. The important findings of the study are 1) IBS status alone is not associated with significant shifts in mucosal microbiota, supporting the findings from other studies, 2) IBS BH subtypes are significantly associated with mucosal microbial richness, 3) an increase in colonic mucosal Prevotella_9 genus and P. copri species is correlated with abdominal pain in patients with IBS, and 4) metagenomic profiling of microbial metabolic pathways showed an upregulation of phenylalanine metabolism in IBS.

Our study identified a novel association of mucosal microbiota, Prevotella_9, specifically, Prevotella_9 copri with abdominal pain in the patients with IBS. Increased abundance of P. copri in the fecal microbiota has been associated with altered immune function (28). In addition, P. copri has been associated with inflammatory conditions, including subclinical and new-onset rheumatoid arthritis (RA; 29–31). Although IBS is not a primarily inflammatory disorder, immune activation and microscopic inflammation have been reported in a subset of patients with IBS (32). In postinfection IBS (PI-IBS), changes to the innate and adaptive immune systems have been associated with abdominal pain that persisted after the enteric infection (33, 34).

At the genus level, our study also identified an association of increased abundance of Prevotella with IBS symptoms. Increased abundance of Prevotella genus has been previously associated with IBS-D (35). Bacteria belonging to genera including Prevotella produce mucin degrading enzymes, which leads to altered thickness in the mucus layer affecting epithelial barrier function (36). Increased intestinal permeability is an important pathophysiologic feature of IBS and has been found in a subset of patients with IBS-D and IBS-C (37). Our previous study suggested a role for altered expression of barrier function and inflammation-associated genes in IBS (38). These studies further support a role of Prevotella in immune function and intestinal barrier dysfunction contributing to visceral pain in IBS (36). Since our IBS cohort did not have medical histories suggestive of PI-IBS, it is likely that the association of P. copri and abdominal pain is not specific to PI-IBS.

Another important finding from this study was an association of IBS BH subtypes with α diversity of mucosal microbiota. Some studies recommend avoiding rarefication for data normalization. Our analysis on unrarefied data also revealed significant differences in the α diversity between bowel habit subtypes. Decreased microbial richness in IBS-C compared with IBS-D, and to a lesser extent compared with HCs, suggest a different microbial profile of IBS-C compared with IBS-D and HCs. This aligns with our previous studies that showed greater differences in mRNA and microRNA profiles in IBS-C than in IBS-D compared with HCs (38, 39). Of the bacteria decreased in IBS-C compared with HCs, lower abundance of butyrate-producing bacteria (40) Erysipelotrichaceae_UCG-003 has been associated with immune function (41) and Allisonella with intestinal barrier function (42), which are important factors in the pathogenesis of IBS. Studies comparing the gut microbiota between IBS BH subtypes also found a decrease in the abundance of Erysipelotrichaceae in patients with IBS-C (40, 43), which aligns with our finding that this bacterium is significantly decreased in the colonic mucosa and is associated with IBS-C. Although there were no β diversity differences, there was a difference in α diversity, and taxonomic differences even after correcting for multiple testing, suggesting that there are some differences in microbiota between bowel habit subtypes. Our findings suggest that there may be differences in microbial composition based on IBS BH subtype rather than IBS status and raises the possibility that the pathogenesis in IBS may be related to the BH subtype or clinical presentation.

Some bacteria that were significantly enriched in the microbiota of IBS-D compared with IBS-C, including Desulfovibrio, Prevotella, and Muribaculum, may have been due to differences in individual dietary components that were not quantified in our diet survey. In particular, the enrichment of Prevotella could be the result of higher fiber consumption in patients with IBS-D to relieve symptoms of diarrhea through bulking agents (44, 45). Significant enrichment of Bradyrhizobium and Acinetobacter was identified in the mucosal microbiota of patients with IBS-D in comparison with HCs and patients with IBS-C but are also likely related to differences in diet (46, 47).

Our findings indicate that IBS status is not associated with the mucosal microbiota, which is consistent with other mucosal microbiota studies in IBS (48, 49). We observed an increase in relative abundance of F. magna and C. ureolyticus species in the microbiota of IBS compared with HCs. C. ureolyticus is associated with gastroenteritis and traveler’s diarrhea (50, 51), and F. magna is an opportunistic pathogen that has been associated with inflammation and infection, but the significance of these microbes in IBS pathophysiology is yet to be determined (51, 52). The same microbes have not been reported in other mucosal microbiota studies, which may be a result of differences in study population and methodologies. A few mucosal microbiota studies reported an increase in Bacteroidaceae family and Bacteroides genus (53, 54), but these findings may be a result of fecal contamination since an increase in Bacteroides has been demonstrated in many fecal microbiota studies (4). In fact, one of the mucosal microbiota study collected biopsies from unprepped bowel, whereas the other did not include any descriptions of how the bowel preparation was done (53, 54).

Predicted metagenomic profile of mucosal microbiota showed an enrichment of genes associated with phenylalanine metabolism in IBS compared with HCs. Phenylalanine is an aromatic amino acid that has been identified as a major fecal metabolite in patients with IBS and IBS-D (55, 56). Studies on gnotobiotic mice models have demonstrated a role for phenylalanine in intestinal permeability and systemic immunity (57), supporting a functional role for mucosal microbiota in IBS. The predictive metagenomic profiling uses hidden-state prediction approaches that use a reference phylogeny and not solely relying on reference taxonomic units (20), which may explain in part a significant association of functional pathways but not many ASVs with IBS.

Our study has limitations. We only sampled the sigmoid colonic mucosa and studies have demonstrated that significant differences in mucosal microbiota exist along the intestinal tract, e.g., lower-α diversity in the small intestine compared with the large intestine (8, 10). Therefore, our findings may not represent changes throughout the entire GI tract that are relevant to IBS. We used a diet checklist to determine if participants consumed a restrictive diet and if there were differences in diet between IBS and HCs. We validated this diet checklist in our previous study evaluating the effect of diet on fecal microbiota in IBS, which included a larger number of participants with ∼50% of the participants that overlapped with our cohort (16). Although we controlled for restrictive diet in our analysis, the diet survey does not quantify or account for differences in individual food components, which may have confounded some of our findings at the taxonomic level. Our study does not involve analysis of the fecal microbiota, as we have already published a large study on fecal microbiota in relation to diet (16), with only ∼24% overlap between the previous cohort and the one used in the present study. In addition, although an objective measure of bowel habits such as transit time is not available on these patients, a careful investigation of bowel habit subtype based on Rome criteria was performed during the medical history. Although some members of Prevotella genus including P. copri have been associated with proinflammatory roles, other Prevotella species have been associated with beneficial roles, suggesting a role for multiple factors including bacterial community-mediated effects (58), which warrant further investigation. However, some of the reported species or ASV-level taxonomic differences warrant independent replication given the limitations of short read 16s rRNA sequencing in accurately identifying microbes at species/ASV-level. In addition, the role of predicted pathways in IBS pathogenesis warrants additional investigation.

In conclusion, our study showed that there are significant differences in the mucosal microbiota between IBS BH subtypes, especially between patients with IBS-C compared with HCs and patients with IBS-D. IBS is a heterogeneous disorder, and our findings further support the possibility of different microbial communities contributing to subtype-specific mechanisms of pathogenesis in IBS. Although IBS status alone was not associated with overall shifts in mucosal microbiota compared with HCs, metagenomic profiling results suggest functional differences in the mucosal microbiota between IBS and HCs that could lead to intestinal barrier dysfunction and symptoms of IBS. Furthermore, this is the first study to identify Prevotella in the mucosal microbiota of IBS in association with increased abdominal pain, which may help elucidate immunogenic mechanisms in a subset of patients with IBS. Finally, our study reaffirms the importance of controlling for diet in any study of the gut microbiota, as diet likely influences colonic mucosal microbiota.

SUPPLEMENTAL DATA

Supplemental Tables S1–S5 and Supplemental Figs. S1–S6: https://doi.org/10.6084/m9.figshare.19747606.

GRANTS

This study was supported by National Institutes of Health Grants P50 DK64539 (to E. A. Mayer), P30 DK 41301 (J. E. Rozengurt), 1R21 DK104078 (to L. Chang), and UL1TR0001881 (to S. M. Dubinett) and by Veterans Affairs Grant IK2CX001717 (to J. P. Jacobs).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

S.M.-J., J.P.J., and L.C. conceived and designed research; V.L. performed experiments; C.C., S.M.-J., T.S.D., A.L., and N.J. analyzed data; C.C., S.M.-J., T.S.D., A.L., J.S.L., N.J., E.A.M., and L.C. interpreted results of experiments; C.C., S.M.-J., and L.C. drafted manuscript; C.C., S.M.-J., T.S.D., A.L., J.P.J., J.S.L., N.J., E.A.M., and L.C. edited and revised manuscript; S.M.-J., and L.C. approved final version of manuscript.