Abstract
-
Background/Aims
- Inflammatory bowel disease (IBD) has become a global health concern. With the growing evidence of the gut microbiota’s role in IBD, studying microbial compositions across ethnic cohorts is essential to identify unique, population-specific microbial signatures.
-
Methods
- We analyzed stool samples and clinical data from 254 IBD patients (226 ulcerative colitis, 28 Crohn’s disease) and 66 controls in northern India using metagenomic shotgun sequencing to assess microbiota diversity, composition, and function. Results were replicated in 436 IBD patients and 903 controls from the Netherlands using identical workflows. Using machine learning, we evaluated the generalizability of Indian IBD signals to the Dutch cohort, and vice versa.
-
Results
- Indian IBD patients exhibited reduced bacterial diversity and an abundance of opportunistic pathogens, including Clostridium, Streptococcus, and oral bacteria like Streptococcus oralis and Bifidobacterium dentium. There was a significant loss of energy metabolic pathways and distinct co-occurrence patterns among microbial species. Notably, 39% of these signals replicated in the Dutch cohort. Unique to the Indian cohort were oral pathobionts such as Scardovia, Oribacterium, Actinomyces dentalis, and Klebsiella pneumoniae. Both Indian and Dutch IBD patients shared reduced butyrate producers. Machine-learning diagnostic models trained on the Indian cohort achieved high predictive accuracy (sensitivity 0.84, specificity 0.95) and moderately generalized to the Dutch cohort (sensitivity 0.77, specificity 0.69).
-
Conclusions
- IBD patients across populations exhibit shared and unique microbial signatures, suggesting a role for the oral-gut microbiome axis in IBD. Cross-ethnic diagnostic models show promise for broader applications in identifying IBD.
-
Keywords: Inflammatory bowel diseases; Gastrointestinal microbiome; Dysbiosis; Machine learning; Microbiota
INTRODUCTION
Inflammatory bowel disease (IBD), comprising Crohn’s disease (CD) and ulcerative colitis (UC), is a chronic inflammatory disorder of the gastrointestinal tract. The phenotype and clinical manifestations of IBD show considerable variations across the globe [1-3]. The origin of this heterogeneous clinical presentation remains poorly understood, though interactions between the host genetic architecture and environmental factors (such as diet, smoking, antibiotics, hygiene, infections, stress, breastfeeding, etc.), mediated via modulation of gut microbiota, have been proposed to shape the clinical expression of the disease [4,5].
Among these non-genetic components, the gut microbiota has emerged as a key determinant of susceptibility to IBD and its clinical phenotype. Previous studies, primarily conducted in European and North American populations, have highlighted certain microbial signatures associated with IBD, such as reduced alpha and beta diversity, reduction of the anti-inflammatory butyrate producers of the family Lachnospiraceae and Ruminococcaceae and an increase in the proinflammatory bacteria of the family Enterobacteriaceae [6-9]. Significant variations in gut microbial composition and metabolic pathways have also been noted among healthy individuals across the globe [10-12]. However, whether these differences are exclusive to healthy individuals or are translated to diseased states, remains elusive.
Given the increasing recognition of the role of gut microbiota in comprehending and predicting IBD phenotypes, it becomes imperative to explore the microbial compositions across diverse ethnic cohorts. We hypothesized that the gut microbiota may demonstrate variations among IBD cohorts with diverse ethnic, genetic, and geographic backgrounds. Assessing the magnitude of these differences is pivotal for evaluating their impact on the susceptibility, clinical phenotype, progression, and treatment responses between distinct ethnic groups. Here, we report the microbiome analyses based on metagenomic shotgun sequencing of patients with IBD from a population-based cohort in north India. This was then, using an identical workflow, compared with a Dutch IBD cohort for evaluation of distinctive and shared gut microbiota. Additionally, we report the development of a microbiota-based machine-learning diagnostic model to predict IBD.
METHODS
1. Study Population
Patients with IBD were enrolled from the Department of Gastroenterology at Dayanand Medical College and Hospital, Ludhiana, Punjab, India. The diagnosis of IBD was based on the accepted clinical, radiological, endoscopic, and histopathological assessments as per the European Crohn’s and Colitis Organisation and the European Society of Gastrointestinal and Abdominal Radiology guidelines [13,14]. Patients with any prior intestinal or IBD related surgery, current or past cancer, major cardiovascular, neurological, hepatic or renal illness, pregnancy/lactation, and use of antibiotics/probiotics within the previous 6 months were excluded.
The healthy controls were selected from a separate community-based project, comprising 66 healthy individuals, also of north Indian ancestry, from the general population in Punjab, India. A stringent screening questionnaire focused on the medical history, medication usage, and gastrointestinal complaints was used to screen the healthy controls. This was followed by a meticulous physical examination by a medical professional to ensure that the selected healthy controls were healthy and devoid of co-morbid diseases or concomitant medication use.
2. Data Collection
The demographics (age, sex) and baseline disease characteristics (including disease activity, duration, extent, and concomitant therapy) were recorded for all the patients at the time of stool sampling. Disease activity at the time of stool sampling was determined by standardized and accepted clinical activity scores: the Harvey Bradshaw Index and partial Mayo score for patients with CD and UC, respectively [15,16]. The Montreal classification was used to describe the disease location and behavior [17]. Disease duration was calculated by subtracting the date of diagnosis from the date of stool sampling. Concomitant treatment for IBD at the time of sampling, including 5-aminosalicylates, corticosteroids, thiopurines, biologics and small molecules, was recorded.
3. Fecal Sample Collection and DNA Isolation
The IBD patients and healthy controls were thoroughly briefed on stool collection procedures during face-to-face sessions, allowing ample opportunity for any queries or apprehensions to be addressed. To avert any alterations in the gut microbial profile induced by colon preparation, stool samples were collected either before or at least 7 days post colonoscopy/sigmoidoscopy. Patients were tasked with collecting stool samples in their own homes and immediately storing them in the provided icebox at a temperature of 2–8 °C. Upon arrival at the hospital, typically within 4–6 hours of collection, fecal samples were divided into aliquots to prevent future freeze-thaw cycles, labeled with a unique identifier, and preserved at −80 °C. All samples remained frozen until DNA-isolation. Fecal DNA was extracted using DNeasy PowerLyzer PowerSoil kit (Qiagen/MO BIO cat# 12855-50; Hilden, Germany) as per manufacturer’s protocol.
4. Metagenomic Data Generation and Processing
Fecal metagenome sequencing was performed with Novaseq S4 using a commercial facility. Quality control of raw sequencing data was performed using KneadData (v.0.12.0) pipeline: reads aligning (using Bowtie2 v.2.5.1) to human genome (reference GRCh37/Hg19) were removed, BBduk tool (v.38.93-0) was used to remove sequencing adapters, and Trimmomatic tool (v.0.39-2) was used to trim reads to PHRED quality ≥20 [18-21]. Trimmed reads shorter than 50 base pairs were discarded. Microbiome composition was profiled using MetaPhlAn tool (v.4.0.6) coupled to marker database vOct22_CHOCOPhlAnSGB_202212, while microbiome biochemical functionality was profiled using HUMAnN3 pipeline (v.3.7) integrated with the DIAMOND alignment tool (v2.1.4), UniRef90 protein database (v201901b) and ChocoPhlAn pan-genome database (v201901b) [18,22-24]. As a final quality control step, samples with >70% of unclassified reads were deemed low quality and discarded from further analysis. Unclassified reads were removed from MetaPhlAn results and relative abundances were recalculated prior to all statistical analyses. In total, we identified 16 phyla, 195 classes, 223 orders, 270 families, 800 genera, 1,573 species, and 554 pathways from bacteria or archaea. The analyses were performed on the Hábrók high-performance computing facility (University of Groningen and University Medical Center, Groningen, The Netherlands).
5. Calculation of Firmicutes-to-Bacteroidetes ratio, Alpha Diversity and Beta Diversity
Firmicutes to Bacteroidetes (F/B) ratio was calculated as
F/B ratio=lnRelative abundance (Firmicutes)Relative abundance(Bacteroidetes)
Microbiome alpha diversity (Shannon diversity, Simpson diversity and richness) were calculated at level of microbial species using diversity function (vegan package for R) [25]. Beta diversity was calculated using Aitchison distance (Euclidian distance of centered-log-ration [CLR] transformed relative abundances, with zeros replaced by minimum non-zero value divided by 2) at the level of microbial species and HUMAnN3-profiled MetaCyc pathways. For this analysis, we included species with relative abundance ≥ 1.0e-8 and prevalence ≥ 2%, and pathways with relative abundance ≥1.0e-5 and prevalence ≥ 5% to reduce sparsity of data.
6. Calculation of Microbiome Variation Explained by Phenotypes
The proportion of microbiome composition variation explained by individual phenotypes was calculated by permutational multivariate analysis of variance using distance matrices (ADONIS) implemented in the ADONIS function of R package vegan (v.2.4-6). Analysis was performed on the microbiome beta diversity for each phenotype using univariate ADONIS with 5,000 permutations. An equivalent analysis was performed on the Aitchison distance matrix calculated using relative abundances of MetaCyc microbial biochemical pathways. Results were deemed significant at false discovery rate (FDR) of <0.05.
7. Calculation of Differentially Abundant Taxa in IBD Patients
Microbial taxa and pathways present in at least 5% of samples and with relative abundance >1.0e-5 were used for differential abundance analysis (12 phyla, 94 classes, 107 orders, 133 families, 304 genera, 576 species, and 324 pathways). Multivariable linear regression was used to calculate differential abundance of CLR-transformed microbial taxa and pathways. Results were deemed significant at FDR of <0.05.
8. Training and Testing of Predictive Models for IBD
To assess the predictive power of microbiome for diagnosis of IBD, we trained gradient boosting trees (implemented in R caret package as “gbm” algorithm) [26]. The model was trained on a training set consisting of 233 samples randomly selected from the IBD patients and controls, while preserving original ratio between cases and controls (as implemented in R caret function createDataPartition). Microbial taxa in the training set were pre-processed by CLR transforming the relative abundances, scaling and centering the values (by subtracting mean and dividing the values by standard deviation), and discarding taxa with near-zero variance (frequencies of most common to second most common values >95:5 or less than 10 unique values) or high correlation (R2 >0.95) with other taxa. It was implemented using caret package preProcess function with parameters “scale”, “center”, “nzv”, and “corr.” Model was trained using caret package function train with 3 repeats of 10-fold cross-validation, optimization for Cohen’s Kappa value as metric of model accuracy, and tuning length parameter=10 (metric=“kappa”; tuneLength=10). Model performance was evaluated on test set consisting of 58 samples not used for the model training.
9. Microbiome Network Analysis
We constructed microbiome co-abundance networks for IBD patients and healthy controls at species and genus level to evaluate microbe-microbe interactions, and potential differences in these interactions in the patients and controls. The Species, Genus, and Pathway networks were constructed for IBD patients and healthy controls, using 200 highest-variance genera and 300 highest-variance species and pathways. The co-abundances were calculated as Pearson correlations of CLR relative abundances and significant edges were selected using simulation-based approach with 10,000 bootstraps and significance cutoff of FDR <0.001. Nodes with top quintile values for degree, betweenness, closeness and eigenvector were designated as Network Hub nodes. IBD and control networks were compared based on differences in degree, betweenness, closeness of nodes, using permutation-based test (100 permutations) to assign P-values to differences in these values. The analysis was performed using NetCoMi package for microbial network construction (netConstruct function), analysis (netAnalyze function), and comparison (netCompare function) [27].
10. Replication of Results in Dutch Cohorts of IBD Patients and Population Controls
We compared our results to microbiome of IBD patients in European population; we re-analyzed raw metagenomic data from 1000IBD and Lifelines-DEEP cohorts using identical bioinformatics (described above) and statistics [28,29]. Results of differential abundance analysis were considered replicated if the microbial feature was significantly differentially abundant in both cohorts at FDR <0.05 and the direction of the association was identical.
To test the generalizability of predictive models across the populations, we used the model trained on the Indian cohort to classify 1000IBD and Lifelines-DEEP participants as IBD or non-IBD samples and vice-verse. For these tests, the values of all features present in the training data (Indian IBD patients and controls), but not in the test set (1000IBD and Lifelines-DEEP cohorts), were set to 0.
11. Ethical Considerations
Ethical approval was obtained from the Institutional Ethics Committee of the Dayanand Medical College and Hospital, Ludhiana (IEC number: 2020557). Informed consent was obtained from all the participants. Ethical approval for the 1000IBD and Lifelines DEEP cohorts was obtained by the University Medical Center Groningen Institutional Review Board (IRB; #M12.113965, 2008.338).
RESULTS
1. Clinical Characteristics
A total of 268 patients (236 UC and 32 CD) and 66 healthy controls were initially enrolled from the north Indian cohort. However, 14 patients (10 with UC and 4 with CD) were excluded due to failed quality control tests on their stool samples. The analyzed cohort consisted of 226 patients with UC (mean age 39 ± 13 years, 47% females) and 28 patients with CD (mean age 38 ± 13 years, 36% females). The mean disease duration was 5.12 ± 5.66 years and 4.63 ± 5.55 years for patients with UC and CD, respectively. At the time of collection of stool samples, 107 patients (42.12%, 94 UC and 13 CD) were in clinical remission. A majority of the patients were ovo-lactovegetarians. The detailed information on baseline patient and disease characteristics, and concomitant medications in the north Indian cohort is summarized in Table 1 and microbiota measurements are summarized in Supplementary Table 1. The summary statistics for the Dutch cohort are provided in Supplementary Table 2 and microbiota measurements in Supplementary Table 3.
2. Composition of Gut Microbiota in Indian Cohort
1) Microbial Diversity
A statistically significant decrease in Shannon index was noted for patients with IBD compared to the healthy controls. However, no significant differences were observed between patients with UC and CD (Fig. 1A). The alpha diversity indices (Richness, Shannon index and inverse Simpson index) were also decreased in patients with IBD (Supplementary Fig. 1) The differences in gut microbial composition (beta diversity) between IBD patients and healthy controls were also observed in the principal coordinate-analysis of species composition. Significant differences were found in the first 2 principal coordinates (PCoA1 P<2.2e-16, PCoA2 P=2.4e-5) and on the level of whole-microbiome composition (ADONIS R2=0.07, P<0.0002). The gut microbiota of healthy controls formed a distinct cluster, separate from the gut microbiota of patients with IBD. While there was only partial overlap between the gut microbiota of healthy controls and patients with IBD, there was a notable overlap between patients with UC and CD (Fig. 1B and C).
2) Gut Microbiota Composition
Firmicutes, Bacteroidetes, Actinobacteria, and Proteobacteria were the predominant phyla in both patients with IBD and healthy controls. The Actinobacteria were significantly increased in patients with UC (24.5% vs. 11.2% in healthy controls, P=2.75e-11, FDR=6.38e-08). On the contrary, the Bacteroidetes were decreased in patients with both UC and CD when compared to healthy controls, though statistical significance was not reached (12.9% in UC patients, 31% in healthy controls, P=0.001, FDR=1 and 10.8% in CD patients, 31% in healthy controls, P=0.049, FDR=1) (Fig. 2). Additionally, the F/B ratio was 0.84 ± 1.57 in healthy controls, 2.85 ± 2.5 in patients with UC, and 2.12 ± 1.47 in patients with CD (Supplementary Fig. 2).
A total of 319 microbial taxa were altered in patients with IBD, with 82 and 32 taxa significantly affected in UC and CD, respectively (FDR<0.05) (Supplementary Table 4). At the genus level, in patients with both UC and CD, the abundances of Lactobacillus, Streptococcus, Solobacterium, Granulicatella, Limosilactobacillus, Lancefieldella, Rothia, and Faecalimonas were significantly increased (FDR<0.05). The abundances of Paraprevotella, Gemmiger, Clostridiaceae, and Ruminococcus were decreased. Interestingly, there was no significant difference between the relative abundance of Faecalibacterium prausnitzii or Faecalibacterium intestinalis (FDR >0.05) between the healthy controls and patients with IBD. However, a significant reduction of Faecalibacterium SGB15346 (currently uncharacterized species, which was previously bundled together with F. prausnitzii in MetaPhlAn2 [and 3] databases) was observed in UC (FDR<0.05) (Supplementary Tables 1, 4).
On multivariate linear regression analysis, the composition of the microbiota in patients with IBD remained notably distinct from that of healthy controls. The genera Mogibacterium, Bifidobacterium, Lactobacillus, Enterococcus, Streptococcus, Lancefieldella, and Actinomyces were increased in patients with IBD. On the other hand, the genera Clostridiaceae, Lachnospiraceae, Ruminococcus, Dorea, Catenibacterium, and Roseburia were decreased. Notably, the bacteria primarily inhabiting the oral cavity (Lancefieldella parvula, Streptococcus mitis, Bifidobacterium dentium, Streptococcus oralis, etc.) were significantly enriched in patients with UC. The patients with CD also exhibited an enrichment of oral bacteria, however, the observed differences did not reach statistical significance (Fig. 3).
Seventy-one metabolic pathways were differentially expressed between patients with IBD and healthy controls. Table 2 summarizes the significantly altered MetaCyc pathways (FDR<0.05). In patients with UC, there was an upregulation of the biosynthetic pathways of amino acids (lysine, threonine, methionine), pyrimidine deoxyribonucleotides and starch, while the biosynthesis of fatty acids (oleate, stearate, palmitoleate), and alcohol (butanediol, farnesol) were increased in both UC and CD. The biosynthesis of biotin and thiamine diphosphate, gluconeogenesis and folate transformation were decreased in patients with UC (Supplementary Fig. 3, Supplementary Table 5).
We also examined the association between IBD activity and the microbiome in the Indian cohort. While we did not observe a significant correlation between IBD activity and overall microbiome composition, we identified a link with microbial functional profiles. We also analyzed dietary habits and found no significant associations with microbiome composition or function. However, we identified significant associations between microbiome variance (both composition and function) and the use of specific medications, including 5-aminosalicylates, prednisolone, tofacitinib, proton pump inhibitors, corticosteroids, and azathioprine (Supplementary Fig. 4). However, when correlating these variables with the relative abundances of individual bacterial taxa or pathways, no statistically significant associations were identified after multiple-testing correction.
3) Co-abundance Network Analysis
Differences in microbial species and pathway co-abundances vary significantly between healthy controls and patients with IBD. Our group has previously reported IBD-specific co-abundance networks in a Dutch cohort [30]. A similar analysis on Indian cohort revealed distinct co-abundance patterns in healthy controls and patients with IBD (Pearson correlation >0.3). Fig. 4 demonstrates the contrasting network characteristics between the 2 cohorts. The clustering coefficient within these networks serves as a measure of the complexity of microbial interactions, indicating robust connections among microorganisms. Healthy controls showed a more connected network (edge and vertex connectivity=4, edge density 0.32), while the IBD network had a less connected network (edge and vertex connectivity=1, edge density 0.14), suggesting a more dynamic and active microbial community in healthy individuals and fewer microbiome-microbiome interactions in patients with IBD. Additionally, the average path length, representing the compactness of the network and strength of microbial interactions was 1.56 in healthy controls compared to 2.72 in patients with IBD. Network analysis identified 3 hubs in IBD patients’ network (unclassified genus GGB13404, family Clostridium and unclassified genera, GGB9699 and GGB9644 from family Oscillospiraceae), while the healthy microbiome network had 7 hubs (genera Bacteroides, Dialister, Erysipelatoclostridium, Faecalibacterium, Klebsiella, Lancefiedella, and candidate genus Nanosynsacchari).
3. Cross-Ethnic Comparison of Gut Microbiota Composition in Patients with IBD
We next aimed to explore the similarities in the gut microbiota composition between Indian and the Dutch IBD cohorts. We re-analyzed raw metagenomic data from 1000IBD and Lifelines-DEEP cohorts using identical bioinformatics and statistical analyses, including quality control (as described above). Through this comparison, we identified both unique and overlapping microbial signatures. In the Indian cohort, we observed a depletion of Prevotella and some yet-to-be-characterized species. Conversely, there was an enrichment of specific pathobionts commonly found in the oral cavity, including Scardovia, Oribacterium, Actinomyces, Megatherium and Klebsiella (Supplementary Fig. 5 and Supplementary Tables 4, 6).
Upon replication of the metabolic pathways in the Dutch IBD cohort, we observed enrichment of the mevalonate pathway, biosynthesis of peptidoglycans, and homocysteine and cysteine interconversion in both UC and CD patients. Additionally, both the Indian and the Dutch cohort exhibited an increased biosynthesis of butanediol and fructan in UC (Supplementary Fig. 3 and Supplementary Tables 5, 7, 8).
4. Development of Machine Learning Model to Predict IBD
To assess the applicability of microbiota-based predictive models across different populations, we employed the model trained on an Indian cohort to categorize 1000IBD and Lifelines-DEEP participants as either IBD or non-IBD samples, and vice versa, and a gradient boosting trees machine learning model. This model was trained on 75% of the available data and was subsequently tested on the remaining 25% data. Upon testing on the Indian IBD cohort, utilizing gut microbial signatures from the Indian population, the model demonstrated high accuracy (specificity 0.95, sensitivity 0.84) for identifying the characteristic microbial dysbiosis signal associated with IBD. Furthermore, when this same model was applied to the Dutch cohort, it showed good generalization (specificity 0.69, sensitivity 0.77). Likewise, the model constructed using Dutch IBD microbiome data achieved high accuracy in predicting IBD within the Dutch population (specificity 0.75, sensitivity 0.94) and good generalization to Indian cohort (specificity 0.65, sensitivity 0.77) when cross-applied to the Indian IBD cohort, suggesting a significant similarity in microbial dysbiosis signatures between the Indian and Dutch populations (Fig. 5).
DISCUSSION
Major shifts in the gut microbiota composition have previously been reported in patients with IBD, influencing immune response and inflammation. However, the majority of these individual studies focused on participants from a single geographic region, concentrating on evaluating the relationships between gut microbiota and various host covariates. The gut microbiota is influenced by, apart from the host-microbiota interactions, various environmental exposures. Therefore, conducting cross-ethnic comparative analyses is crucial for gaining a comprehensive understanding of the impact of microbiota on disease development, progression, and treatment responses. This study represents one of the first efforts to elucidate the ethnicity-specific IBD-associated changes in the gut microbiota across 2 diverse populations.
Confirming prior reports, patients with IBD exhibited a loss of microbial diversity, with no significant divergence in diversity between UC and CD [7]. The dysbiosis of the gut microbiota in IBD patients was profound: the abundances of 82 and 32 taxa were altered in patients with UC and CD, respectively compared to healthy individuals (FDR <0.05, replicated in Dutch population). Also, 71 biochemical pathways were altered in Indian patients (FDR <0.05). Of note, 39% of these signals were replicated in the Dutch cohort (at FDR <0.05). We identified strong dysbiosis shared across Indian and Dutch IBD patients, characterized by expansion of opportunistic pathogens (e.g., Streptococcus and Lactobacilli) and oral bacteria (e.g., S. oralis and B. dentium) as well as reduction in butyrate producers. In addition, we found novel pathobionts specific to the Indian cohort (including oral bacteria from genus Scardovia and Oribacterium; Actinomyces dentalis and Klebsiella pneumoniae). Machine-learning models trained on the Indian cohort were highly predictive in the Indian test set (sensitivity 0.84, specificity 0.95) and generalized to the Dutch cohort (sensitivity 0.77, specificity 0.69). Although there are differences in cohort sizes and the balance between cases and controls, the models trained on the Dutch cohort also generalized well to the Indian cohort (sensitivity 0.77, specificity 0.65). The area under the curve values for the 2 models were comparable: Indian model applied to Dutch data: 0.79 ± 0.03 and Dutch model applied to Indian data: 0.75 ± 0.06.
The gut microbial composition was dominated by the phyla Firmicutes and Bacteroidetes, but a high F/B ratio was seen in patients with IBD. This is in contrast to the previously reported observations where IBD was associated with a low F/B ratio [5,31-33]. The increased F/B ratio in the current study was primarily due to the decreased abundance of the phyla Bacteroidetes (genera Bacteroides, Alistipes, Parabacteroides, and Prevotella), rather than an absolute increase in the Firmicutes. Though most preclinical and clinical studies have demonstrated that Bacteroidetes exhibit proinflammatory properties, contributing to IBD, a meta-analysis reported lower levels of Bacteroides spp. to be associated with IBD [6]. The decrease in Bacteroidetes in the current study could probably be attributed to the geographic and ethnic variations in the relative abundances of these bacteria, probably influenced by the diet, host genetics, and the host microbe interactions [34,35]. Also, the ubiquitous use of 5-aminosalicylates, which is associated with decrease in Bacteroidetes, suggest a pharmacological influence contributing to the observed differences in Bacteroidetes [36]. Furthermore, lack of smokers in the current cohort (smoking is associated with increased Bacteroidetes), could potentially explain the decreased abundance of Bacteroidetes [37]. Disease activity is another modifier of the F/B ratio. While much attention is often given to the F/B ratio in the context of gut microbiota dysbiosis, it is crucial to recognize that this ratio can be affected by changes in other bacterial phyla as well. Dysbiotic increases in other phyla, such as Proteobacteria and Actinobacteria, also contribute to overall dysbiosis and may influence the F/B ratio in patients with IBD [38].
An important finding from our study was the enrichment of the gut microbiota by the bacteria from the oral cavity (Scardovia, Oribacterium, Streptococcus spp., A. dentalis and K. pneumoniae). There is growing evidence indicating that the migration of oral bacteria into the intestinal tract plays a significant role in the development of inflammatory diseases due to their immune-stimulatory properties. While our study did not include a direct analysis of oral microbiota, our conclusions about oral pathobiont contributions are based on the enrichment of taxa typically associated with the oral cavity (Rothia, Streptococcus, Neisseria, Prevotella, Klebsiella, and Gemella) within the gut microbiome of IBD patients. This phenomenon has been reported in prior studies and suggests potential translocation of oral bacteria to the gut in IBD [39-41]. Our study reiterates the connection between oral and gut dysbiosis [42,43]. Under normal physiological circumstances, the intestine resists the colonization of the non-native pathobionts. However, the inflammatory milieu of IBD may render the intestine more permissive to the colonization by the oral bacteria [44,45]. It is plausible that oral dysbiosis contributed to gut dysbiosis, and colonization of oral microbiota in the gut potentially perpetuated the gut dysbiosis, leading to sustained aberrant chronic inflammation [39,46-48]. However, we acknowledge that without direct sampling of the oral microbiota, we cannot definitively establish the oral origin of these taxa. While further evaluation of this relationship is essential, efforts aimed at controlling oral pathobionts could potentially alleviate intestinal inflammation.
Multiple metabolic pathways were differentially expressed in patients with IBD compared to healthy controls, some of these were unique to the Indian IBD cohort while others were shared between the Indian and the Dutch cohorts. Most of the altered metabolic pathways play an important role in oxidative stress, β-oxidation, glycolysis, and the tricarboxylic acid cycle that are essential for energy metabolism, maintenance of the integrity of the intestinal barrier and mediation of the host immune responses [49-51]. Our findings contribute to the proposition that disruptions in lipid metabolism, deficiencies in amino acids, and dysregulation of energy homeostasis likely contribute to both the onset and sustenance of IBD. Disruptions in β-oxidation and lipid metabolism can impair energy generation and contribute to oxidative stress, which has been implicated in intestinal barrier dysfunction and chronic inflammation, key features in the pathogenesis of IBD. Similarly, deficiencies in essential amino acids may compromise mucosal healing and immune regulation, exacerbating disease progression. Furthermore, the dysregulation of pathways involved in energy metabolism also suggests an impaired ability of intestinal epithelial cells to maintain homeostasis, potentially predisposing individuals to the onset of IBD. Once the disease is established, these metabolic imbalances may further sustain chronic inflammation by promoting immune dysregulation, oxidative stress, and barrier dysfunction.
Based on the gut microbial analysis of the fecal samples, we developed a machine learning model to provide a non-invasive approach to the diagnostic screening of IBD. The machine learning models, developed from an Indian IBD cohort, and tested on a Dutch cohort, achieved 59% accuracy. Conversely, models trained on the Dutch IBD population had an 81% accuracy rate when applied to the Indian IBD population. These results highlight the adaptability and potential of our approach across different ethnicities. Despite the inherent variability in gut microbiome due to factors such as age, sex, diet, and environment, our model’s ability to identify IBD solely through assessing gut microbiota compositions marks a significant leap forward. The robustness and disease specificity of the microbial signature offers a promising prospect for widespread application in diverse populations.
Our study is limited by a small representation of patients with CD within the IBD cohort. Despite this constraint, our study successfully identified several microbiome signals in Indian CD patients, replicated these findings in a European cohort, and provided comparative insights into microbiome differences between CD and UC patients. We did not include additional biomarker testing such as fecal calprotectin for selection of healthy controls. However, our approach aligns with established methodologies used in similar microbiome studies to define healthy controls. We also acknowledge the heterogeneity of the sample due to the inclusion of post-diagnosis patients rather than treatment-naïve individuals. Most studies investigating the gut microbiota in IBD have followed a similar approach due to the practical challenges of obtaining samples at the time of diagnosis or during the preclinical stage. Many patients present after symptom onset and have had received some treatment (antibiotics, antimotility agents, non-steroidal anti-inflammatory agents, etc.), making it difficult to systematically collect pre-treatment samples, especially in real world settings. Despite these limitations, we believe that our findings remain valuable in understanding the microbial signatures in real-world IBD populations, where treatment effects are an inherent part of disease progression. This is the first study to report on an Indian IBD cohort and perform a cross-ethnic variation analyses in the gut microbiota in patients with IBD from Netherlands. Also, an ample sample size would have enabled a detailed resolution of the microbial landscape among individuals with IBD from north India.
In conclusion, this study elucidates both distinctive and the shared microbial and metabolic signatures present within the Indian and Dutch IBD populations. Notably, Indian IBD patients exhibited enrichment in specific pathobionts, particularly the oral microbiota. Future studies incorporating paired oral and gut microbiome analyses to further investigate the role of the oral-gut axis in IBD are required. The machine learning diagnostic models constructed based on microbiome data effectively identified IBD patients across diverse populations. Further validation of these findings, coupled with an exploration of the interactions between microbiota, genetics, and environmental factors, is required.
NOTES
-
Funding Source
The study received partial funding support from multiple sources. Thelma BK, Midha V, and Sood A were partially funded under CoE phase I (BT/01/COE/07/UDSC/2008). Thelma BK also acknowledges funding from the JC Bose Fellowship (phase II, 2016–2021). Juyal G acknowledges support from the Science and Engineering Research Board, New Delhi (F. no. SB/YS/LS-191/2014), and Joshi MC acknowledges financial support from the UGC FRP ((236-FRP)/2015/BSR). The clinical aspects of the study were supported by the Research and Development Center, Dayanand Medical College and Hospital, Ludhiana.
-
Conflict of Interest
Sood A received honorarium for speaker events from Pfizer India and Takeda India. Weersma RK acted as consultant for Takeda Pharmaceuticals, received unrestricted research grants from Takeda, Johnson & Johnson, Tramedico, and Ferring, and received speaker’s fees from MSD, AbbVie, and Janssen Pharmaceuticals. The remaining authors disclose no conflicts.
-
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
-
Author Contributions
Conceptualization: Midha V, Thelma BK, Weersma RK, Sood A. Data curation: Singh A. Formal analysis: Singh A, Juyal G, Gacesa R, Joshi MC, Thelma BK. Funding acquisition: Sood A. Methodology: Singh A, Juyal G, Gacesa R, Joshi MC. Project administration: Sood A. Resources: Singh A, Midha V, Thelma BK, Sood A. Software: Juyal G, Joshi MC, Weersma RK. Supervision: Midha V, Thelma BK, Weersma RK, Sood A. Validation: Singh A, Juyal G, Gacesa R, Weersma RK. Visualization: Singh A, Juyal G, Gacesa R. Writing – original draft: Singh A, Juyal G, Gacesa R. Writing – review & editing: all authors. Approval of f inal manuscript: all authors.
-
Additional Contributions
The authors acknowledge the efforts put in by Dr Dharmatma Singh (Research and Development Center, Dayanand Medical College and Hospital, Ludhiana, Punjab, India) and Mr Vikas Kathuria (Research and Development Center, Dayanand Medical College and Hospital, Ludhiana, Punjab, India) in coordinating the collection of the fecal samples, and Dr Arshia Bhardwaj (Department of Gastroenterology, Dayanand Medical College and Hospital, Ludhiana, Punjab, India) for providing critical inputs to the manuscript.
Supplementary Material
Supplementary materials are available at the Intestinal Research website (https://www.irjournal.org).
Supplementary Fig. 1.
The alpha diversity indices in patients with IBD (UC and CD) and healthy controls. P-values for significant Mann-Whitney U test are shown on the plot, and associations replicated in Dutch cohort are marked with ampersand (&) symbol. IBD, inflammatory bowel disease; UC, ulcerative colitis; CD, Crohn’s disease; HC, healthy controls.
ir-2024-00216-Supplementary-Fig-1.pdf
Supplementary Fig. 2.
The Firmicutes Bacteroidetes ratio in healthy controls and patients with IBD in (A) Indian cohort and (B) Dutch replication cohort. The ratio was 0.84, 2.85, and 2.12 in healthy controls, and patients with UC and CD, respectively. Boxplot Centre line is the median, box limits indicate upper and lower quartiles, whiskers show 1.5× interquartile range, points indicate outliers and the outline displays the distribution of the data. P-values for significant Mann-Whitney U test are shown on the plot. IBD, inflammatory bowel disease; HC, healthy controls; UC, ulcerative colitis; CD, Crohn’s disease.
ir-2024-00216-Supplementary-Fig-2.pdf
Supplementary Fig. 3.
Differentially abundant metabolic pathways in patients with IBD (A) observed in both Indian and Dutch cohorts (B) specific to Indian cohort. IBD, inflammatory bowel disease; UC, ulcerative colitis; CD, Crohn’s disease.
ir-2024-00216-Supplementary-Fig-3.pdf
Supplementary Fig. 4.
Variance in microbiome composition (A) and function (B) explained by phenotypes in the Indian cohort. X-axis show variance explained (R2) value of PERMANOVA analysis, Y-axis lists analyzed phenotypes, and bars are colored based on statistical significance. FDR, false discovery rate.
ir-2024-00216-Supplementary-Fig-4.pdf
Supplementary Fig. 5.
The heatmap revealing differentially abundant taxa either enriched or depleted in north Indian IBD cohort, but not present in Dutch IBD cohort. The blue color represents enrichment and orange color depletion (P<0.05). The “+” and “–” signs indicates associations significant after FDR correction (FDR<0.05). The lowest characterized taxonomic level is written in brackets for species that currently lack species- and genus-level characterization. IBD, inflammatory bowel disease; UC, ulcerative colitis; CD, Crohn’s disease; FDR, false discovery rate.
ir-2024-00216-Supplementary-Fig-5.pdf
Fig. 1.(A) The alpha diversity indices in Indian patients with IBD (UC and CD) and HC. P-values for significant Mann-Whitney U test are shown on the plot, and associations replicated in Dutch cohort are marked with an ampersand (&) symbol. (B) PCoA of the fecal microbiome of IBD patients and HC. The first component shows 18% variance in microbial species between HC and patients with IBD. (C) The f irst component shows 36% variance in metabolic pathways between HC and patients with IBD. Centroids of the groups are depicted with large dots. IBD, inflammatory bowel disease; UC, ulcerative colitis; CD, Crohn’s disease; PCoA, principal coordinate analysis; HC, healthy controls; PCo1, principal coordinate 1; PCo2, principal coordinate 2.
Fig. 2.Gut microbiota composition in Indian patients with inflammatory bowel disease compared to healthy controls at (A) species level, (B) phyla level. HC, healthy controls; UC, ulcerative colitis; CD, Crohn’s disease.
Fig. 3.Heatmap displaying the enriched and depleted microbial species in Indian UC and CD patients, compared to Indian healthy controls. Only the signals replicated in Dutch cohorts at FDR <0.05 are displayed. The blue color represents enrichment and orange color depletion (P<0.05). The “+” and “–” signs indicates associations significant after FDR correction (FDR<0.05). The lowest characterized taxonomic level is written in brackets for species that currently lack species- and genus-level characterization. UC, ulcerative colitis; CD, Crohn’s disease; FDR, false discovery rate.
Fig. 4.(A, B) The microbial genus co-abundance network between healthy controls and patients with inflammatory bowel disease (IBD; Indian cohort). Healthy controls showed a more connected network, while the IBD network had a less connected network indicative of a more dynamic and active microbial community in healthy individuals and less microbiome-microbiome interactions in patients with IBD. Colors depict identified sub-network clusters, edges show significant co-abundances (red for negative correlations and green for positive correlations), and network hubs are highlighted with bold names.
Fig. 5.Receiver operator characteristic curves depicting performance of prediction models to diagnose inflammatory bowel disease (IBD) based on microbiome composition. Shaded area on the curve depicts 95% confidence interval. Sensitivity and specificity are shown for maximal prediction accuracy. AUC, area under the curve.
Table 1.Baseline Characteristics
|
Characteristics |
UC (n = 226) |
CD (n = 28) |
Healthy controls (n = 66) |
|
Age (yr) |
39 ± 13 |
38 ± 13 |
38 ± 14 |
|
Female sex |
107 (47.34) |
10 (35.71) |
21 (31.81) |
|
Body mass index (kg/m2) |
21.56 ± 4.43 |
24.41 ± 4.44 |
25.22 ± 3.78 |
|
Diet |
|
|
|
|
Vegetarian |
118 (52.21) |
18 (64.28) |
28 (42.42) |
|
Non-vegetarian |
108 (47.78) |
10 (35.71) |
38 (57.57) |
|
Disease duration (yr) |
5.12 ± 5.66 |
4.63 ± 5.55 |
- |
|
Age at diagnosis (yr) |
34 ± 13 |
34 ± 13 |
- |
|
Smoking |
- |
- |
- |
|
Disease extent (UC) |
|
- |
- |
|
Proctitis |
33 (14.60) |
|
|
|
Left sided colitis |
150 (66.37) |
|
|
|
Pancolitis |
43 (19.02) |
|
|
|
Age at diagnosis (yr) |
- |
|
- |
|
< 17 |
|
1 (3.57) |
|
|
17–40 |
|
16 (57.14) |
|
|
> 40 |
|
11 (39.28) |
|
|
Disease location (CD) |
- |
|
- |
|
Ileal |
|
12 (42.85) |
|
|
Colonic |
|
8 (28.57) |
|
|
Ileo-colonic |
|
8 (28.57) |
|
|
Disease behavior (CD) |
- |
|
- |
|
Inflammatory |
|
19 (67.85) |
|
|
Stricturing |
|
6 (21.42) |
|
|
Penetrating |
|
3 (10.71) |
|
|
Disease activitya
|
|
|
- |
|
Remissionb
|
94 (41.59) |
13 (46.42) |
|
|
Active disease |
132 (58.40) |
15 (53.58) |
|
|
Concomitant treatment |
|
|
- |
|
5-ASA |
226 (100) |
18 (64.28) |
|
|
Thiopurines |
88 (38.93) |
22 (78.57) |
|
|
Corticosteroids |
80 (35.39) |
16 (57.14) |
|
|
Anti-TNFs |
8 (35.39) |
9 (32.14) |
|
|
Tofacitinib |
28 (12.38) |
- |
|
Table 2.The Metabolic Pathways Altered in Patients with IBD
|
Pathway |
Diagnosis |
Change |
P-value |
FDR |
|
Specific to the Indian IBD cohort |
|
|
|
|
|
PWY-922: Mevalonate pathway I (eukaryotes and bacteria) |
UC |
Increase |
5.49e-13 |
1.29e-09 |
|
PWY-5910: Superpathway of geranylgeranyl diphosphate biosynthesis I (via mevalonate) |
UC |
Increase |
3.27e-12 |
7.64e-09 |
|
PWY-6471: Peptidoglycan biosynthesis IV (Enterococcus faecium) |
UC |
Increase |
6.99e-11 |
1.62e-07 |
|
PWY-6396: Superpathway of 2,3-butanediol biosynthesis |
UC |
Increase |
4.56e-10 |
1.04e-06 |
|
PWY-6471: Peptidoglycan biosynthesis IV (Enterococcus faecium) |
CD |
Increase |
4.56e-08 |
9.96e-05 |
|
METHGLYUT-PWY: Superpathway of methylglyoxal degradation |
UC |
Increase |
1.36e-07 |
2.90e-04 |
|
PWY-822: Fructan biosynthesis |
UC |
Increase |
1.10e-06 |
2.30e-03 |
|
PWY-NAD-BIOSYNTHESIS-II: NAD salvage pathway III (to nicotinamide riboside) |
UC |
Increase |
1.54e-06 |
3.19e-03 |
|
PWY-801: Homocysteine and cysteine interconversion |
UC |
Increase |
1.81e-06 |
3.73e-03 |
|
P122-PWY: Heterolactic fermentation |
UC |
Increase |
1.89e-06 |
3.91e-03 |
|
PWY-5910: Superpathway of geranylgeranyl diphosphate biosynthesis I (via mevalonate) |
CD |
Increase |
4.27e-06 |
8.64e-03 |
|
PWY-922: Mevalonate pathway I (eukaryotes and bacteria) |
CD |
Increase |
7.88E-06 |
1.57E-02 |
|
Shared between the Indian and Dutch IBD cohorts |
|
|
|
|
|
PWY-6895: Superpathway of thiamine diphosphate biosynthesis II |
UC |
Decrease |
1.61E-12 |
3.77E-09 |
|
PWY-6282: Palmitoleate biosynthesis I (from 5Z-dodec-5-enoate) |
UC |
Increase |
9.73e-10 |
2.20e-06 |
|
PWY-6859: All-trans farnesol biosynthesis |
UC |
Increase |
1.29e-09 |
2.92e-06 |
|
PWY-7664: Oleate biosynthesis IV (anaerobic) |
UC |
Increase |
2.91e-09 |
6.52e-06 |
|
PANTO-PWY: Phosphopantothenate biosynthesis I |
UC |
Decrease |
4.61e-09 |
1.03e-05 |
|
FASYN-ELONG-PWY: Fatty acid elongation (saturated) |
UC |
Increase |
5.96e-09 |
1.32e-05 |
|
PWY0-862: 5Z-dodecenoate biosynthesis I |
UC |
Increase |
1.22e-08 |
2.70e-05 |
|
PWY-6282: Palmitoleate biosynthesis I (from 5Z-dodec-5-enoate) |
CD |
Increase |
2.80e-07 |
5.90e-04 |
|
POLYAMSYN-PWY: Superpathway of polyamine biosynthesis I |
UC |
Decrease |
3.32e-07 |
7.00e-04 |
|
PWY-7664: Oleate biosynthesis IV (anaerobic) |
CD |
Increase |
5.17e-07 |
1.09e-03 |
|
PWY0-862: 5Z-dodecenoate biosynthesis I |
CD |
Increase |
6.38e-07 |
1.34e-03 |
|
PWY-5989: Stearate biosynthesis II (bacteria and plants) |
CD |
Increase |
7.47e-07 |
1.57e-03 |
|
FASYN-ELONG-PWY: Fatty acid elongation (saturated) |
CD |
Increase |
8.70e-07 |
1.83e-03 |
|
PANTOSYN-PWY: Superpathway of coenzyme A biosynthesis I (bacteria) |
UC |
Decrease |
9.26e-07 |
1.94e-03 |
|
PWY-6700: Queuosine biosynthesis I (de novo) |
UC |
Decrease |
2.29e-06 |
4.69e-03 |
|
PWY-7977: L-methionine biosynthesis IV |
UC |
Decrease |
7.91e-06 |
1.58e-02 |
|
PWY-6859: All-trans farnesol biosynthesis |
CD |
Increase |
2.51e-05 |
4.84e-02 |
REFERENCES
- 1. Bhat M, Nguyen GC, Pare P, et al. Phenotypic and genotypic characteristics of inflammatory bowel disease in French Canadians: comparison with a large North American repository. Am J Gastroenterol 2009;104:2233–2240.ArticlePubMedPMCPDF
- 2. Walker DG, Williams HR, Kane SP, et al. Differences in inflammatory bowel disease phenotype between South Asians and Northern Europeans living in North West London, UK. Am J Gastroenterol 2011;106:1281–1289.ArticlePubMedPDF
- 3. Hou JK, El-Serag H, Thirumurthi S. Distribution and manifestations of inflammatory bowel disease in Asians, Hispanics, and African Americans: a systematic review. Am J Gastroenterol 2009;104:2100–2109.ArticlePubMedPDF
- 4. Hu S, Bourgonje AR, Gacesa R, et al. Mucosal host-microbe interactions associate with clinical phenotypes in inflammatory bowel disease. Nat Commun 2024;15:1470.ArticlePubMedPMCPDF
- 5. Imhann F, Vich Vila A, Bonder MJ, et al. Interplay of host genetics and gut microbiota underlying the onset and clinical presentation of inflammatory bowel disease. Gut 2018;67:108–119.ArticlePubMed
- 6. Zhou Y, Zhi F. Lower level of Bacteroides in the gut microbiota is associated with inflammatory bowel disease: a meta-analysis. Biomed Res Int 2016;2016:5828959.ArticlePubMedPMCPDF
- 7. Vestergaard MV, Allin KH, Eriksen C, et al. Gut microbiota signatures in inflammatory bowel disease. United European Gastroenterol J 2024;12:22–33.ArticlePubMed
- 8. Prosberg M, Bendtsen F, Vind I, Petersen AM, Gluud LL. The association between the gut microbiota and the inflammatory bowel disease activity: a systematic review and meta-analysis. Scand J Gastroenterol 2016;51:1407–1415.ArticlePubMed
- 9. Sankarasubramanian J, Ahmad R, Avuthu N, Singh AB, Guda C. Gut microbiota and metabolic specificity in ulcerative colitis and Crohn’s disease. Front Med (Lausanne) 2020;7:606298.ArticlePubMedPMC
- 10. Dhakan DB, Maji A, Sharma AK, et al. The unique composition of Indian gut microbiome, gene catalogue, and associated fecal metabolome deciphered using multi-omics approaches. Gigascience 2019;8:giz004.ArticlePDF
- 11. Sheng Y, Wang J, Gao Y, et al. Combined analysis of cross-population healthy adult human microbiome reveals consistent differences in gut microbial characteristics between Western and non-Western countries. Comput Struct Biotechnol J 2023;23:87–95.ArticlePubMedPMC
- 12. Zhang W, Han N, Zhang T, et al. The spatial features and temporal changes in the gut microbiota of a healthy Chinese population. Microbiol Spectr 2022;10:e0131022.ArticlePubMedPMCPDF
- 13. Maaser C, Sturm A, Vavricka SR, et al. ECCO-ESGAR Guideline for Diagnostic Assessment in IBD Part 1: initial diagnosis, monitoring of known IBD, detection of complications. J Crohns Colitis 2019;13:144–164.ArticlePubMed
- 14. Sturm A, Maaser C, Calabrese E, et al. ECCO-ESGAR Guideline for Diagnostic Assessment in IBD Part 2: IBD scores and general principles and technical aspects. J Crohns Colitis 2019;13:273–284.ArticlePubMed
- 15. Lewis JD, Chuai S, Nessel L, et al. Use of the noninvasive components of the Mayo score to assess clinical response in ulcerative colitis. Inflamm Bowel Dis 2008;14:1660–1666.ArticlePubMedPMC
- 16. Harvey RF, Bradshaw JM. A simple index of Crohn’s-disease activity. Lancet 1980;1:514.ArticlePubMed
- 17. Satsangi J, Silverberg MS, Vermeire S, Colombel JF. The Montreal classification of inflammatory bowel disease: controversies, consensus, and implications. Gut 2006;55:749–753.ArticlePubMed
- 18. Beghini F, McIver LJ, Blanco-Míguez A, et al. Integrating taxonomic, functional, and strain-level profiling of diverse microbial communities with bioBakery 3. Elife 2021;10:e65088.ArticlePubMedPMC
- 19. Langmead B, Salzberg SL. Fast gapped-read alignment with Bowtie 2. Nat Methods 2012;9:357–359.ArticlePubMedPMCPDF
- 20. DOE Joint Genome Institute. BBTools [Internet]. c2024 [cited 2024 Apr 14]. https://jgi.doe.gov/data-and-tools/software-tools/bbtools/.
- 21. Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 2014;30:2114–2120.ArticlePubMedPMCPDF
- 22. Blanco-Míguez A, Beghini F, Cumbo F, et al. Extending and improving metagenomic taxonomic profiling with uncharacterized species using MetaPhlAn 4. Nat Biotechnol 2023;41:1633–1644.ArticlePubMedPMCPDF
- 23. Buchfink B, Xie C, Huson DH. Fast and sensitive protein alignment using DIAMOND. Nat Methods 2015;12:59–60.ArticlePubMedPDF
- 24. Suzek BE, Huang H, McGarvey P, Mazumder R, Wu CH. Uni-Ref: comprehensive and non-redundant UniProt reference clusters. Bioinformatics 2007;23:1282–1288.ArticlePubMedPDF
- 25. Dixon P. VEGAN, a package of R functions for community ecology. J Veg Sci 2003;14:927–930.Article
- 26. Kuhn M. Building predictive models in R using the caret package. J Stat Softw 2008;28:1–26.PubMedPMC
- 27. Peschel S, Müller CL, von Mutius E, Boulesteix AL, Depner M. NetCoMi: network construction and comparison for microbiome data in R. Brief Bioinform 2021;22:bbaa290.ArticlePDF
- 28. Imhann F, Van der Velde KJ, Barbieri R, et al. The 1000IBD project: multi-omics data of 1000 inflammatory bowel disease patients; data release 1. BMC Gastroenterol 2019;19:5.ArticlePubMedPMCPDF
- 29. Zhernakova A, Kurilshikov A, Bonder MJ, et al. Population-based metagenomics analysis reveals markers for gut microbiome composition and diversity. Science 2016;352:565–569.ArticlePubMedPMC
- 30. Chen L, Collij V, Jaeger M, et al. Gut microbial co-abundance networks show specificity in inflammatory bowel disease and obesity. Nat Commun 2020;11:4018.ArticlePubMedPMCPDF
- 31. Guo X, Huang C, Xu J, et al. Gut microbiota is a potential biomarker in inflammatory bowel disease. Front Nutr 2022;8:818902.ArticlePubMedPMC
- 32. Stojanov S, Berlec A, Štrukelj B. The influence of probiotics on the Firmicutes/Bacteroidetes ratio in the treatment of obesity and inflammatory bowel disease. Microorganisms 2020;8:1715.ArticlePubMedPMC
- 33. Shen ZH, Zhu CX, Quan YS, et al. Relationship between intestinal microbiota and ulcerative colitis: mechanisms and clinical application of probiotics and fecal microbiota transplantation. World J Gastroenterol 2018;24:5–14.ArticlePubMedPMC
- 34. Zafar H, Saier MH Jr. Gut Bacteroides species in health and disease. Gut Microbes 2021;13:1–20.Article
- 35. Magne F, Gotteland M, Gauthier L, et al. The Firmicutes/Bacteroidetes ratio: a relevant marker of gut dysbiosis in obese patients? Nutrients 2020;12:1474.ArticlePubMedPMC
- 36. Andrews CN, Griffiths TA, Kaufman J, et al. Mesalazine (5-aminosalicylic acid) alters faecal bacterial profiles, but not mucosal proteolytic activity in diarrhoea-predominant irritable bowel syndrome. Aliment Pharmacol Ther 2011;34:374–383.ArticlePubMed
- 37. Benjamin JL, Hedin CR, Koutsoumpas A, et al. Smokers with active Crohn’s disease have a clinically relevant dysbiosis of the gastrointestinal microbiota. Inflamm Bowel Dis 2012;18:1092–1100.ArticlePubMed
- 38. Shin NR, Whon TW, Bae JW. Proteobacteria: microbial signature of dysbiosis in gut microbiota. Trends Biotechnol 2015;33:496–503.ArticlePubMed
- 39. Atarashi K, Suda W, Luo C, et al. Ectopic colonization of oral bacteria in the intestine drives TH1 cell induction and inflammation. Science 2017;358:359–365.ArticlePubMedPMC
- 40. Said HS, Suda W, Nakagome S, et al. Dysbiosis of salivary microbiota in inflammatory bowel disease and its association with oral immunological biomarkers. DNA Res 2014;21:15–25.ArticlePubMed
- 41. Gevers D, Kugathasan S, Denson LA, et al. The treatment-naive microbiome in new-onset Crohn’s disease. Cell Host Microbe 2014;15:382–392.ArticlePubMedPMC
- 42. Elzayat H, Malik T, Al-Awadhi H, et al. Deciphering salivary microbiome signature in Crohn’s disease patients with different factors contributing to dysbiosis. Sci Rep 2023;13:19198.ArticlePubMedPMCPDF
- 43. Elmaghrawy K, Fleming P, Fitzgerald K, et al. The oral microbiome in treatment-naïve paediatric IBD patients exhibits dysbiosis related to disease severity that resolves following therapy. J Crohns Colitis 2023;17:553–564.ArticlePubMedPMCPDF
- 44. Pickard JM, Zeng MY, Caruso R, Núñez G. Gut microbiota: role in pathogen colonization, immune responses, and inflammatory disease. Immunol Rev 2017;279:70–89.ArticlePubMedPMCPDF
- 45. Caballero S, Pamer EG. Microbiota-mediated inflammation and antimicrobial defense in the intestine. Annu Rev Immunol 2015;33:227–256.ArticlePubMedPMC
- 46. Tiwana H, Walmsley RS, Wilson C, et al. Characterization of the humoral immune response to Klebsiella species in inflammatory bowel disease and ankylosing spondylitis. Br J Rheumatol 1998;37:525–531.ArticlePubMed
- 47. Cao X. Intestinal inflammation induced by oral bacteria. Science 2017;358:308–309.ArticlePubMed
- 48. Qi Y, Wu HM, Yang Z, et al. New insights into the role of oral microbiota dysbiosis in the pathogenesis of inflammatory bowel disease. Dig Dis Sci 2022;67:42–55.ArticlePubMedPDF
- 49. Noh JY, Farhataziz N, Kinter MT, Yan X, Sun Y. Colonic dysregulation of major metabolic pathways in experimental ulcerative colitis. Metabolites 2024;14:194.ArticlePubMedPMC
- 50. Aldars-García L, Gisbert JP, Chaparro M. Metabolomics insights into inflammatory bowel disease: a comprehensive review. Pharmaceuticals (Basel) 2021;14:1190.ArticlePubMedPMC
- 51. Scoville EA, Allaman MM, Brown CT, et al. Alterations in lipid, amino acid, and energy metabolism distinguish Crohn’s disease from ulcerative colitis and control subjects by serum metabolomic profiling. Metabolomics 2018;14:17.ArticlePubMedPDF
Citations
Citations to this article as recorded by
KASID
, Garima Juyal2,*
PubReader
ePub Link
Cite
XML Download
