Background
Inflammatory bowel diseases (IBD) are a group of digestive tract disease that affect millions of people worldwide. There are signs that incidence rates of IBD are increasing and presenting earlier in life [
1]. IBD is divided into 2 main disease processes: Crohn’s Disease (CD), which may affect any segment of the gastro-intestinal tract, and Ulcerative Colitis (UC), which is limited to the large intestine [
2]. Patients with CD frequently present with severe abdominal pain, fever, and clinical signs of bowel obstruction or diarrhea [
3]. Inflammation modulation is the mainstay of medical management and ranges from the use of anti-inflammatories, to corticosteroids and immunomodulators [
4]. In severe cases, surgery may be required. Unfortunately, the need for surgery has not decreased despite advancements in diagnostic and treatment protocols [
5].
Despite theorizing that CD arises from an impaired interaction between commensal microbiome and the human host, the distinction between primary driver events and secondary occurrences remains murky. However, the recent focus on gut microbiome dysbiosis in CD has led to the discovery of new diagnostic and therapeutic directions [
6]. Murine models have shown that the disease only manifests in susceptible genotypes and is driven by microbial dysbiosis [
7], while human studies have found a general decrease in alpha diversity with clade-specific changes in CD patients such as increased
Enterobacteriaceae and decreased Firmicutes [
8,
9].
One of the limitations of previous microbiome studies has been the sequencing strategy. Most of the studies employed 16S rRNA sequencing, which limits their findings to bacteria at the genus level. This gap can be addressed with whole genome shotgun (WGS) sequencing, which is able to detect the presence of microbes with better accuracy and provide data at species level [
10]. Additionally, most of these studies were in western populations and with subjects of European descent. This limited the scope of previous results as ethnicity, eating habits and living environment are variables that affect the gut microbiome [
11].
Previous work on the oral-gut axis has demonstrated a connection between oral inflammation and its contribution to gut inflammation in animal models. Oral pathobiont-reactive inflammatory cells arising from oral inflammation were found to migrate to the gut, promoting and contributing to colitis [
12]. The effect of the oral microbiome on CD is by comparison relatively understudied; however, its impact cannot be ignored [
13]. A recent study found that oral
Klebsiella can colonize the gut and result in severe gut inflammation in susceptible individuals, thereby exacerbating inflammatory disease [
14]. Another study found increased presence of species found abundantly in oral communities in the gut microbiomes of CD subjects having diarrhea, suggesting that the oral cavity may serve as a reservoir for opportunistic gut pathogens [
15].
Previous studies examining the oral microbiome were also obfuscated by having different sampling sites, such as sampling from the tongue [
16], plaque [
17] and saliva [
18,
19]. Of these sampling sites, the salivary microbiome appears to offer the most diagnostic value without the excessive influence of local factors [
20]. These studies found enrichment of
Veillonellaceae and depletion of
Haemophilus in CD subjects, highlighting the diagnostic value of the oral microbiome [
18,
19]. Moreover, as a diagnostic tool, saliva samples are non-invasive unlike colonoscopy, and easy to collect at any time unlike fecal samples.
This study aims to characterize, for the first time, the oral and gut microbiome in Asian subjects with CD using whole genome shotgun technique. A case–control analysis was conducted with matched healthy controls to investigate the presence of altered community structure and different community ecotypes in the oral and gut microbiomes of a mixed Asian population consisting of Han Chinese, Malay and Indian.
Discussion
This study represents a metagenomic insight into the oral and gut microbiome in CD patients, conducted in an Asian population, consisting subjects of Chinese, Indian and Malay descent (Table
1), which may have innate differences in gut and oral microbiome. Although some studies have examined the gut microbiome in CD [
10,
21], none of the previous studies on the oral microbiome employed the shotgun metagenomic technique used in this study. Furthermore, this is the first metagenomic study of matched oral and gut samples in CD subjects, allowing matching of the gut and oral microbiome at the strain level.
Crucially, this study found a cluster of CD subjects whose gut microbiome were characterized by
S. salivarius, a prominent oral microbe. In health,
Streptococcus genus typically constitute less than 4% of gut microbiome [
13], thus it was note-worthy to find
S. salivarius enriched in the gut microbiome of CD subjects. Although the ectopic colonization of oral bacteria such as
Klebsiella pneumoniae has been found to induce intestinal inflammation and can result in the progression of CD [
14], it was only demonstrated in mouse models. Other studies suggesting roles in CD by oral bacteria were largely based on finding typically oral bacteria in the gut microbiome of CD subjects. In the current study, matched oral and gut samples were taken and shown for the first time in CD subjects that
S. salivarius were of similar strains. This increases the likelihood that the oral microbiome was the source for ectopic gut colonization in CD. This study also found that subjects in the active phase of disease were more likely to have ectopic gut colonization from the oral microbiome. A previous study found that oral microbes can colonize the gut after diarrheal episodes [
15]. The authors hypothesized that the loss of gut microbes during diarrheal episodes reduces bacterial competition, this coupled with a transient increase in oxygen, allows for the ectopic colonization of oral microbes. Diarrhea is a major symptom during the active phase CD, suggesting that the oral microbiome can serve as a reservoir for pathogenic recolonization of the gut. A factor in ectopic gut colonization is the use of proton pump inhibitors which decreases the gastric acidity, allowing for ectopic gut colonization of oral microbes [
22]. However, this study did not find that the use of proton pump inhibitors and low gastric acid state increases the likelihood of ectopic gut colonization. A limitation of the study is the small number of subjects in which concordant gut and oral strains were found. Furthermore, although not statistically significant, all CD subjects with concordant gut and oral
S. salivarius were Han Chinese. Therefore, it cannot be ruled out that racial differences [
11] contributed to the finding of ectopic gut colonization. Larger follow up studies collecting both oral and gut samples will be required to expand upon this finding.
The gut microbial species found at high abundances (> 5%) in both CD and healthy controls in this study were similar to previous studies employing WGS [
10,
23]. Contrary to previous studies, this study did not find significant reduction of alpha-diversity in the gut microbiome of CD subjects [
24,
25]. However, most of these studies used 16S rRNA sequencing, which may be the reason for the difference. Additionally, another aspect of the study was that most of the CD subjects were in remission (CDAI < 150) and well-controlled with medication, which may be a reason for their microbiome being less divergent from control. However, this would mean that any changes detected are more likely to be key changes driving microbiome dysbiosis.
The differentially abundant species found in gut microbiomes here were similar to previous studies. The depletion of butyrate-producing species such as
F. prausnitzii and
R. inulinivorans has been described in CD subjects [
26].
F. prausnitzii in particular has been proposed to have protective effects in CD [
27,
28], as the butyrate produced by these microbes aid in mucosal barrier function and maintenance of gut health [
23]. Although
Prevotella sp. has been studied extensively and some species have been shown to induce colitis in mice, it has yet been found to be associated with CD in humans [
29]. In this study, it was found that
P. copri was detected in a discreet cluster of healthy controls, similar to the previous WGS study [
10]. This suggests that
P. copri may be protective in some individuals against CD instead.
A. senegalensis is a newly delineated species [
30], which has yet to be implicated in CD. A related species,
Alistipes putredinis has been described to contribute to transcriptional pathways in IBD [
23]. Interestingly, it was also found to be significantly depleted in CD subjects in a previous WGS study [
10] warranting further investigation. Additionally, a commonly implicated microbe
Ruminococcus gnavus [
23,
31] was found to be significantly enriched in CD subjects in this study. A recent mechanistic study into
R. gnavus showed that it produces a complex polysaccharide that can result in the pattern of inflammation seen in CD [
32]. Another species implicated in the pathogenesis of CD is
E. coli, which was found to work in conjunction with other microbes as well as by exhibiting virulence features in CD subjects [
31]. Although
Clostridium nexile was found to be enriched in CD subjects in this study, it was found to be decreased in previous studies [
33]. Comparing this to another WGS study,
C. nexile was also found to be enriched in CD subjects, although not significantly [
10].
C. nexile is able to produce short-chain fatty acids which decreases inflammation and alleviate colitis in experimental models [
34]. However, the exact mechanism of that action is still unknown and does not explain why
C. nexile was enriched instead in CD. This suggests that
C. nexile has a geographically or ethnically specific role and its function warrants further investigation.
Experts in the field have suggested of use the salivary microbiome as a non-invasive tool in the diagnosis of CD [
18,
19]. This study did not find any significant differences between the oral microbiome in CD and controls after adjustment, which could be due to the CD subjects being under treatment, with microbiome alterations due to medications rather than a signature of disease. However, when examining the genus differences, the study found similar enrichment of
Streptococcus, and depletion of
Haemophilus and
Prevotella in CD subjects from previous 16S studies [
18,
19]. Additionally, the ability to detect changes at species level provided better resolution than previous studies showing that the enrichment of
Streptococcus genus involved several species such as
S. anginosus,
S. oligofermentans,
S. vestibularis,
S. cristatus and
S. salivarius, while the depletion of
Prevotella involved
P. salivae and
P. intermedia.
Another interesting finding from the salivary analysis is the enrichment of
Human Herpesvirus 4, also known as Epstein-Barr virus (EBV), found in CD subjects. Recent studies have found a significant correlation between EBV presence in the gut with clinical disease severity [
35]. This could be due to increased susceptibility of the CD patients on immunosuppressive treatment and the ability of EBV to induce inflammation. Being able to detect this in the oral microbiome may allow it to serve as a marker for disease severity.
The number of microbial pathways detected in this study was similar to a previous WGS study which detected 255 pathways in CD subjects [
10]. The microbial pathway analysis found that 4 pathways in the gut microbiome with higher relative abundance levels found in CD compared to control are related to arginine biosynthesis. The arginine pathway is a major contributor to host inflammatory processes via inducible nitric oxide (NO) production, with NO providing protective cytostatic/cytotoxic antimicrobial action [
36]. Administration of L-arginine has even been shown to reduce intestinal inflammation and pathology [
37]. However, the specific role of microbial arginine metabolism has yet to be explored in the context of CD. It could be that these bacteria proliferate in the arginine poor environment in CD subjects as they are able to produce their own arginine. Additionally, isoprene biosynthesis was also found to be at higher relative abundance levels in CD subjects. The increase of isoprene in expired air from IBD subjects has been linked to the activity of disease, and is at a higher level than healthy individuals [
38]. The microbial dysbiosis in CD favors a shift towards arginine and isoprene forming microbes providing a possible target to modify the microbiome.
Methods
Subject inclusion and recruitment
The study was conducted at the National University Hospital, Singapore from May 2017 to January 2019. Subjects with CD (n = 25) were recruited from the IBD clinic, Division of Gastroenterology & Hepatology. Healthy controls (n = 25) without any signs and symptoms of CD were age, sex and ethnicity matched and recruited from the Dental Centre, University Dental Cluster. The control subjects presented at the Dental Centre for routine dental care such as check-up, cleaning and minor restoration. Demographic information, history of tobacco use, other co-morbidities, current drug regimen and current disease activity were recorded. Current disease activity was calculated using the Crohn’s Disease activity index (CDAI) as well as the presence of common signs and symptoms (Diarrhoea/abdominal pain/blood-in-stool/fever and fatigue) [
39]. Subjects who had antibiotic/probiotic/prebiotic use within a month, actively treated for a malignancy with chemotherapy, or diagnosed with an indeterminate colitis were excluded from the study. Non-CD controls were further required to have no known gastrointestinal signs and symptoms such as diarrhoea, abdominal pain and blood-in-stool.
The study received ethics approval from the institutional review boards of the National Healthcare Group, Singapore (DSRB reference E/2016/01,285). Informed consent was obtained from all participants. Additionally, all experiments were performed in accordance with relevant guidelines and regulations.
Oral condition of subjects
Oral involvement of CD is well documented; with studies reporting between 20 and 50% of cases exhibiting oral manifestations including aphthous ulcers, linear deep ulcers, mucosal tags, mucosal “cobblestoning”, mucogingivitis and lip swelling. The presence of these conditions can affect the oral microbiome [
40]. In order to detect differences in the oral microbiome that are key drivers of CD progression, subjects exhibiting oral manifestations were excluded for this study. Other conditions that can affect the oral microbiome such as the presence of caries and periodontal disease were also recorded at sample collection. This was verified by an oral examination conducted by a dentist under artificial light.
Sample collection
Saliva was collected for the examination of the oral microbiome. Subjects were told to refrain from eating, smoking and dental procedures one hour prior to the collection. At least 5 ml of resting whole saliva was collected using the OMNIgene Discover Kit 505 (DNA Genotek, Ottawa, ON, Canada) according to manufacturer’s instructions.
Fecal sample was collected for the examination of the gut microbiome. Subjects were instructed and issued the OMNIgene•GUT (OMR-200) (DNA Genotek, Ottawa, ON, Canada) kit for sample collection. This was completed within two weeks of the saliva collection.
The oral and fecal samples were stored at room temperature as per manufacturer’s recommendation and sent for DNA extraction within 4 weeks from collection.
DNA extraction and purification
Community DNA was extracted from saliva and fecal samples after mechanical lysis via bead-beating using Exgene™ Clinic SV Mini kits (GeneAll Biotechnology, Dongnam-ro, Seoul, Korea), and QIAmp® FAST DNA Stool Mini kits (Qiagen, Hilden, Germany) respectively according to manufacturer’s instructions. Extracted DNA was purified using AMPure XP beads (Beckman Coulter, Indianapolis, IN, USA). The quantity and quality of DNA was examined using a NanoDrop 8000 Spectrophotometer (Thermo Fisher Scientific, Carlsbad, CA, United States). Extracted DNA samples were stored at -80 °C for up to 6 weeks prior to library preparation.
Library preparation and Illumina sequencing
Indexed sequencing libraries were prepared using QIAGEN® QIAseq FX DNA Library Kit (Qiagen, Hilden, Germany) following manufacturer’s instructions and sequenced as paired-end reads 2 × 151 bp on the Illumina HiSeq 4000 sequencer (Illumina, San Diego, CA, USA). On average, 15.5 million raw read pairs were obtained for each sample.
Sequence data processing
Sequenced reads for each library were de-multiplexed into individual fastq file, and analysed using a pipeline (
https://github.com/gis-rpd/pipelines/tree/master/metagenomics/shotgun-metagenomics) for processing paired-end shotgun metagenomic sequencing data. Firstly, raw reads was analyzed using Skewer to trim-off adapter sequences and low-quality bases from each read [
41]. After trimming, reads were decontaminated to remove genomic sequences from the human host by using BWA-MEM to map reads against the hg19 reference [
42]. The remaining reads were regarded as likely of microbial origin (average = 7 million) and were used as input for subsequent taxonomic and functional profiling.
Taxonomic profiling
The MetaPhlAn2 software was used to profile the taxonomic composition of the microbial communities from the post quality-filtered reads [
43]. Reads with sequences that matched microbial clades were used to normalize and calculate relative abundance for taxa from kingdom to species ranks. For reducing noise from false positive identifications, taxa with total relative abundance < 0.1% were excluded from further statistical analysis.
Strain typing
To examine the relationship of the oral microbiome to the gut microbiome, strain analysis on
Streptoccous salivarius, a predominantly oral microbe, was conducted on all gut and oral samples using StrainPhlAn. Using the reads mapped against the marker database of MetaPhlAn2, we extracted the clade specific markers of
S. salivarius. Subsequently multiple sequence alignment of the marker sequences of
S. salivarius strains was performed among the gut and saliva libraries of samples as well as the reference genome (NC_017594), before building the phylogenetic tree [
44].
Statistical analysis
Principal coordinates analysis (PCoA)
Bray–Curtis distance was calculated for PCoA analysis based on the species profiled in the sample set, while the multi-dimensional scaling (MDS) dimensions 1 and 2 were used to visualize the level of similarity between samples based on their microbiome composition.
Diversity analysis
The Vegan package in R was used to calculate Shannon and Simpson’s diversity values at the species level. Subsequently, ggplot2 was used to generate diversity plots. Wilcoxon test in R was used to test for significant difference in the medians of Shannon diversity values between case and control groups.
Association analysis for taxonomic abundances with Crohn’s disease
The Wilcoxon test in R was used to compare the median relative abundance in the cases with Crohn’s disease (CD) versus the non-diseased control group for a significant difference. To analyze the direction of the association, the “alternative” function was used in the model.
Multiple testing correction
We adopted Benjamin Hochberg’s false discovery rate method to correct for multiple testing at a significance threshold of 5%.
Pathway analysis
The HMP Unified Metabolic Analysis Network (HUMAnN2) program was used to determine relative abundance of microbial pathways in gut and saliva microbiomes of CD and control groups. The default Kyoto encyclopedia of genes and genomes (KEGG) orthology catalogue was used as the pathway reference. Subsequently, we extracted the total pathway abundance contributed by the genes present in every taxa of the community for analysis. For association, the Linear discriminant analysis Effect Size (LEfSe) software recognized the relative abundance of each pathway as the feature to test for significant difference in the medians between the two groups. We used the default LDA scores ≥ 2 as the threshold for significance, and to highlight the significance of the association, a more stringent threshold at ≥ 3 was also used for reporting [
45].
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.