Article Text
PDF
Abstract
Objective Our aim was to determine whether or not specific microorganisms were transported selectively to lymph nodes in Crohn's disease (CD) by comparing node and mucosal microbial communities in patients and controls. We also sought evidence of dysbiosis and bacterial translocation.
Design Lymph nodes, and involved and uninvolved mucosal samples were obtained from resections of 58 patients (29 CD, eight ‘other inflammatory bowel disease’ (IBD) and 21 non-IBD). Universal primers targeting V1–V3 regions of bacterial 16S rRNA genes were used to amplify bacterial DNA and amplicons sequenced using high throughput sequencing. 20 patients (eight CD (28%), two other IBD (25%) and 10 non-IBD (48%)) had PCR positive nodes.
Results All samples from an individual were similar: there was no evidence of selective concentration of any microorganism in nodes. No specific microorganism was present in the nodes of all CD samples. Escherichia/Shigella were common in all patient groups but patients with ileal CD had a greater proportion of Escherichia coli reads in their nodes than other CD patients (p=0.0475). Campylobacter, Helicobacter and Yersinia were uncommon; Mycobacterium and Listeria were not detected. Dysbiosis was present in all groups but shifts were specific and no common pattern emerged.
Conclusions It is unlikely that a single bacterium perpetuates inflammation in late stage CD; dysbiosis was common and we found no evidence of increased bacterial translocation. We believe that future studies should focus on early disease and viable bacteria in nodes, aphthous ulcers and granulomas, as they may be more relevant in the initiation of inflammation in CD.
- CROHN'S DISEASE
- BACTERIAL TRANSLOCATION
- SURGICAL RESECTION
- MOLECULAR BIOLOGY
- RIBOSOMAL RNA
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
-
Bacteria are thought to trigger the development of Crohn's disease in susceptible individuals.
-
Microorganisms can be cultured from 18–48% and 5–15% of draining mesenteric lymph nodes from Crohn's disease patients and controls, respectively.
-
Macrophages transport bacteria to lymph nodes, but in Crohn's disease, they may have genetic abnormalities leading to intracellular survival and replication.
What are the new findings?
-
28% and 48% of resected nodes from Crohn's disease patients and surgical disease controls, respectively, contain bacterial DNA.
-
Bacterial species present in nodes reflect those present in mucosa, suggesting a breakdown in the mucosal barrier, not selective bacterial translocation.
-
E coli was found commonly in Crohn's disease patients and was present in greater abundance in patients with ileal disease.
-
Other proposed pathogens, including Campylobacter, Helicobacter, Listeria, Mycobacterial species and Yersinia, were uncommon or not detected.
How might it impact on clinical practice in the foreseeable future?
-
There is a breakdown in barrier function in ‘late stage’ Crohn's disease. Further research may be better directed to studying the early lesion in Crohn's disease, including aphthous ulcers, and examining the role of E coli and its associated virulence factors.
Introduction
Many observations implicate bacteria as a trigger for the development of Crohn's disease (CD): lesions occur in regions with the highest bacterial concentrations1; aphthous ulcers, the earliest lesion in CD, occur in Peyer's patches, the site of bacterial sampling; inflammation resolves when the faecal stream is diverted and is reactivated following reinfusion of bowel contents2; severity of disease is correlated with bacterial density in the mucosa3; granulomas contain bacteria4; and susceptible mice reared under germ free conditions develop inflammation only when non-pathogenic bacteria are introduced.5 Perhaps the strongest argument is that many CD susceptibility loci implicate pathways that involve bacterial recognition, autophagy, barrier function, lymphocyte activation and cytokine signalling.6
There are many mechanisms by which bacteria may trigger CD.7 For example, the presence of a specific pathogen: Mycobacterium avium subspecies paratuberculosis (MAP),8 adherent invasive Escherichia coli (AIEC),4 Campylobacter spp,9 Helicobacter,10 Listeria11 and Yersinia12 ,13 are candidates, but no pathogen is common to all patients or to those of a specific genetic background or disease site. Shifts in microbial composition in the gut lumen affecting homeostasis (dysbiosis) are frequently observed in CD patients, and could trigger CD by mechanisms that are, as yet, not understood. The number of microbes in the adult human gut outnumbers the total number of human cells by an order of magnitude.14 It is estimated that the microbiota comprises as many as 1800 genera and 15 000–36 000 individual species, up to 80% of which cannot be cultured.15 The relationship between the microbiota and the host is symbiotic but the composition of the ‘healthy’ microbiota is difficult to define because of interindividual variation. A few hallmarks are emerging: one of these is the predominance of the phyla Firmicutes (∼50%), Bacteroidetes (∼40%) and Proteobacteria (∼10%).16 This is disturbed in CD, which is characterised by decreases in the proportions of Firmicutes and Faecalibacterium prausnitzii and increases in Bacteroidetes and Enterobacteriaceae.1 ,11 ,17
A breakdown in intestinal barrier function causing bacterial translocation is another proposed mechanism. Bacterial translocation may be a normal physiological event by which luminal antigens are sampled to produce immunocompetent cells,18 and there is evidence that the baseline rate of translocation in humans is 5–10%.19 Several studies have sampled mesenteric lymph nodes of patients suffering from inflammatory bowel disease (IBD) and other conditions requiring surgery using culture dependent techniques. Ambrose et al20 cultured bacteria from 33% and 24% of nodes draining involved and uninvolved CD mucosa, respectively, but from nodes of only two normal controls (5%). Takesue et al21 isolated bacteria from almost half the nodes of CD patients (48%) but only 15% of controls. In contrast, O'Boyle et al22 found that bacterial translocation in 40 IBD patients (18%) was similar to controls (15%). The largest study by MacFie et al23 extended O'Boyle's work and included 927 patients undergoing laparotomy. Positive cultures were found in 14% of patients overall, but IBD was not a risk factor.
Mucosal macrophages and dendritic cells are the cellular mediators of bacterial translocation to draining lymph nodes24 and are able to disseminate viable intestinal bacteria to other sites.25 They are also the cells in which many of the CD susceptibility genes are expressed.6 Gene coexpression network analyses demonstrate an overlapping relationship between pathways that are involved in host responses to intracellular bacteria and those predisposing to IBD.6 Expression of these genes is often increased in macrophages and dendritic cells,26 and defects in these pathways (eg, autophagy) may lead to persistence of viable bacteria.27
We speculated that, in CD, macrophages phagocytose and selectively concentrate specific bacteria responsible for the initiation or perpetuation of inflammation in draining lymph nodes. Our primary aim was to determine whether or not a unique pathogen could be identified by examining the bacteria present in lymph nodes obtained from patients with CD who had undergone surgical resections. The microbiota was characterised using high throughput sequencing (HTS), a sensitive technique capable of detecting cultivable and non-cultivable bacteria, and compared with the composition of adjacent involved and uninvolved mucosa. Results were compared with lymph node and mucosal samples from ‘other IBD’ and non-IBD patients. Disease location and medication use were correlated with the results of the microbial analysis to establish whether or not a bacterial signature or pathogen was associated with these variables. We also sought evidence for dysbiosis and bacterial translocation.
Materials and methods
Subject characteristics
The clinical status of patients was classified as CD, other IBD (ulcerative colitis (UC) and undefined IBD or indeterminate colitis) and non-IBD disease controls. Patient information, including disease status, location and behaviour, age at diagnosis (IBD patients), Montreal classification (CD patients), gender, results of pathology and medication use, including antibiotics, was collected.
Sampling procedures
Immediately following resection, sealed bowel specimens were placed in a sterile container. Prior to opening the bowel lumen, nodes were excised from the mesentery in a sterile environment, and scored with a scalpel to improve penetration of RNAlater. The bowel was opened and 1 cm2 sections of involved and uninvolved bowel were placed in RNAlater. For non-IBD patients, two sites were sampled at opposite ends of the resection. All samples were kept at 4°C for 24 h and transferred to −80°C for storage.
DNA extractions
DNA extractions were performed using Qiagen DNeasy Blood and Tissue kits, with a few modifications. Bead beating (5000 rpm/30 revs/s for 3 min, until visibly homogenised) was performed in buffer ATL using 0.1 mm beads for bacterial cell lysis and 4–6 2.3 mm and 1 mm beads for eukaryotic cell lysis. Debris was removed with proteinase K incubation at 56°C for 15 min. RNase A (5 µL) was added with the first wash buffer, AW1, and left at room temperature for 10 min. Samples were eluted with 100 µL of buffer AE. DNA quantity and quality were checked using a Nanodrop ND-1000 spectrophotometer (Analytical Technologies).
High throughput sequencing of the microbiota
16S rRNA PCR reactions
To determine the sensitivity of the assays, we were able to amplify ovine lymph nodes spiked with 103, but not 102, bacterial cells. Human DNA could be amplified in a sample of PCR negative specimens, indicating that inhibitors were absent and that bacterial DNA was below our detection limit.
V1–V3 hypervariable regions of the 16S rRNA gene were amplified from total DNA extracted from nodes and mucosa using the following primers: 27F (5-CCA TCT CAT CCC TGC GTG TCT CCG ACT CAG MID GAG TTT GAT CMT GGC TCA G-3) and 518R (5CCT ATC CCC TGT GTG CCT TGG CAG TCT CAG WTT ACC GCG GCT GCT GC-3) where the sequences in bold represent the forward and reverse primers of the GS FLX TitaniumMV em PCR (LibL) v2 kit (Roche), MID represents the 8 bp multiplex identifier (MID) and the remainder represents forward primer A and reverse primer B of the V1–V3 region. The MID sequences used, courtesy of Professor Andrew Benson, are provided in online supplementary material (S1). PCR was performed in 50 µL containing 1 U high fidelity platinum Taq DNA polymerase (Invitrogen), 2 mM MgSO4, 5 µL 10× PCR buffer, 0.2 mM of each deoxynucleoside triphosphate (Promega), 10 pmol of each primer, up to 35.8 µL sterile H2O, 20 ng template DNA for mucosa samples and 100 ng for node samples. Reactions were performed using a Palm Cycler (Corbett Research): initial denaturation 94°C for 3 min; 30 cycles of denaturation at 94°C for 15 s, annealing at 55°C for 30 s and extension at 68°C for 60s; and the final extension at 68°C for 10 min. Amplicons were extracted from 2% agarose gels, purified using a Wizard SV Gel and PCR Clean-Up System (Promega), and the Agencourt AMPure XP system (Beckman Coulter). Concentrations were determined using an Agilent 2100 Bioanalyser (Agilent Technologies) with DNA 1000 chips. Equimolar amounts of PCR products were combined (total 500 ng per 48 samples) and used as template in the emulsion based PCR amplification, followed by sequencing on the 454 Genome Sequencer FLX-Titanium system. The emulsion PCR and sequencing steps were performed at the Biological Research Facility, ANU, Australia, according to the manufacturer's instructions (454 Life Sciences, Branford, Connecticut, USA). Signal processing and base calling were performed using 454 Sequencing Software V.2.6 (Roche).
Sequence processing
Sequence curation and processing were performed using Mothur, V.1.23.1.28 Sequences were denoised using the shhh.flows command, a translation of Chris Quince's Pyronoise algorithm29; trimmed of primer and barcode sequences (barcode mismatches allowed, 1 bp, primer mismatches, 2 bp); aligned using the SILVA database; and chimaeras removed using the Uchime code.30 The taxonomy of sequences was determined using the RDP 2011 training sets.
Statistical analysis
Statistical tests for differentially abundant bacterial taxa based on disease status, site and medication use were made using Metastats software31 (http://metastats.cbcb.umd.edu/) with 1000 permutations. This allows a comparison of samples from two groups (eg, lymph nodes from CD and non-IBD patients) and assesses the significance of observed differences. Each hypothesis test includes a p value (individual measure of false positive rate) and a q value (individual measurement of false discovery rate). Any hypothesis test with a q value less than the significance level (p=<0.05) was considered significant and differentially abundant.
Distance matrices (Jaccard and Yue, and Clayton theta coefficients) were generated using a cut-off of 0.15. Files containing operational taxonomic units (OTUs) found across samples (shared files) were used to describe the dissimilarity (1 minus similarity) among samples, and to calculate distances between samples. Species richness estimates were generated using Chao and Ace, community diversity using the Shannon index and sample coverage using Good's coverage. Non-metric dimensional scaling (NMDS) analyses were performed in Mothur with Jaccard and Yue, and Clayton theta distance matrices. The statistical program JMP (V.9) was used to conduct one way ANOVA of Shannon indices between patient groups, and matched pairs t test statistics of mean distances between samples in NMDS plots. We used χ2 tests with Yates’ correction for comparisons of categorical variables.
Results
Subject characteristics
The clinical characteristics of 58 patients who underwent resection are summarised in table 1 and online supplementary table S1. All had intravenous antibiotics with induction of anaesthesia, except where specified. Bacterial DNA was detected in lymph nodes of 28% (8/29) of CD patients, 25% (2/8) of other IBD and 48% (10/21) of non-IBD patients, and in three of 17 (18%) patients with L1, one of four (25%) with L2, four of seven (57%) with L3, zero of one with L4 disease (Montreal classification) and three of six (50%) patients with perineal involvement (NS). Further subgroup analysis based on clinical parameters (eg, treatment other than antibiotics, and underlying surgical disease) was not performed because of the small numbers involved.
Characteristics of patients with PCR positive lymph nodes
Phylotype coverage and richness estimates
The composition of the microbiota in nodes and mucosa of all 20 patients was determined using HTS. Mucosal samples were not available for patient No 13. Table 2 summarises the sequence and OTU numbers, diversity and coverage estimates.
Summary of 16S rRNA sequence numbers, observed diversity richness (OTUs), estimated OTU richness (Chao and Ace), diversity (Shannon) and sample coverage (Good's coverage) for mucosa and lymph node samples
A total of 527 600 raw sequences were generated from all samples. Using Mothur,28 we identified 17 867 unique sequences (average 327 bp). Each sample was covered by an average of 8720 quality sequences, and there was an average of 275 OTUs per sample (236 OTUs per node). The predicted number of OTUs per sample using the Chao and Ace estimators was higher than the observed OTUs. These differences are likely to be due to the effect of rare species, which make up a large proportion of microbial communities.32 Good's coverage estimate was 99% using a 97% species level phylotype threshold. This estimates the completeness of sampling by a probability calculation based on randomly selected amplicon sequences. These results suggest that one additional phylotype would be expected for every 100 additional sequencing efforts and that the vast majority of phylotypes present were captured.
Presence and relative abundance of specific micro-organisms in the lymph nodes and mucosa of CD, other IBD and non-IBD patients
Using Metastats, 39 genera were differentially abundant when CD lymph nodes were compared with non-IBD lymph nodes. However, reads belonging to these genera were sparse and occurred in small numbers of patients (S1), indicating that they were unlikely to be of biological relevance. No bacterial taxon was differentially abundant in the following comparisons: CD nodes and CD affected mucosa, CD nodes and CD unaffected mucosa, non-IBD nodes and non-IBD mucosa.
We investigated the proportions of 16S rRNA gene sequences matching specific pathogens for all node samples (table 4). Escherichia/Shigella was the specific pathogen found most commonly and at the highest levels: 75% of CD patients overall with a relative abundance of 0.2–15.4% of the total microbiota. It was detected in nodes of all three patients with L1 (terminal ileal) CD. Patients with ileal CD also had a greater proportion of E coli reads in nodes than other CD patients (p=0.0475). (We note that two patients in whom it was not detected had been on prolonged courses of antibiotics (weeks)). Escherichia/Shigella was also found in 60% of non-IBD patients. Campylobacteraceae was present in four patients: all three patients with L3p CD and one non-IBD patient. Other specific pathogens were present in one or two patients at low levels (Yersinia, Helicobacteriaceae) or not detected (Mycobacteriaceae, Listeriaceae).
Comparisons of the similarity, diversity and relative abundance of bacterial communities in lymph node and mucosal samples
Given that we could not identify a novel or specific pathogen, we sought to determine whether or not there were characteristic changes in the ecological communities of CD patients in different tissues. HTS results were used to make multiple comparisons of the microbial composition of the nodes, and involved and uninvolved mucosa in an individual CD patient, and in the different patient groups.
The Shannon diversity index takes into account the number and relative abundance of taxa within a community: it averaged 3.09 across all samples. Comparison of diversity estimates showed that non-IBD samples were more diverse than CD samples (ANOVA: F(8,57)=3.1297, p=0.0062), and that samples within the same patient group were similar (figure 1).
Shannon Diversity Indices for each combination of patient disease status and site for the microbial communities, as defined using high throughput sequencing 16S rRNA gene data. The means and 95% CIs for each sample are depicted. Overall, the microbial communities of patients without inflammatory bowel disease (IBD) are significantly more diverse than patients with IBD. For Crohn's disease(CD)/other IBD patients: IM, involved mucosa; LN, lymph node; NM, normal (uninvolved) mucosa. For non-IBD patients, two sites of mucosa, site 1 (S1) and site 2 (S2), were sampled at opposite ends of the resection.
The presence/absence (Jaccard coefficient, see online supplementary figure S1 and figure 2) and relative abundance (Yue and Clayton theta (θYC), see online supplementary figure S2 and figure 3) of OTUs are presented as NMDS plots (figures 2 and 3) and dendrograms (see online supplementary figures S1 and S2). NMDS is a non-parametric ordination based method for reducing community data complexity and identifying relationships among communities. NMDS uses the relative ranks of dissimilarity between microbiome samples in creating the ordination whereas principal coordinates analysis uses the actual dissimilarity values. There is debate over which is superior (http://ordination.okstate.edu/overview.htm) but in this instance we feel that NMDS best preserves the distances between samples. In both comparisons for all patient groups, the node sample from an individual clustered with the corresponding mucosal samples more closely than to any other sample, and each individual was different from another (see online supplementary figures S1 and S2). When comparing the presence/absence of OTUs (see online supplementary figure S1), there was no evidence of clustering by disease state or CD disease site. When comparing relative abundance (see online supplementary figure S2), patients with CD were generally less similar to each other (longer branch length) but three patients with terminal ileal (L1) CD clustered together (patient Nos 3, 34, and 62). The NMDS plots confirmed the similarity of the node and mucosal samples in most patients. A few patients had a dissimilar sample, only one of which was the node—in a non-IBD patient (No 31) (see figures 2 and 3).
Non-metric dimensional scaling (NMDS) plot of the microbial 16S rRNA communities of mucosa and lymph node samples using a distance matrix calculated with the Jaccard similarity coefficient and a threshold of 3%. The Jaccard coefficient is used to calculate the similarity of bacterial sequences in each of the tissue samples. It does not take into account the different proportions (relative abundance) of bacterial sequences. The distance between two points is directly proportional to the Jaccard similarity value for two samples, such that sites positioned close together share a greater fraction of their bacterial taxa than two samples further apart. Tissue samples obtained from the same patient are connected by lines, and on average the lymph node sample clusters closely to the mucosal samples for each patient. Individuals do not cluster according to disease status. Patient numbers precede the tissue type labels: IM, involved mucosa; LN, lymph node; NM, normal (uninvolved) mucosa. For non-IBD patients, two sites of mucosa, site 1 (S1) and site 2 (S2), were sampled at opposite ends of the resection. NMDS stress=0.40.
Non-metric dimensional scaling (NMDS) plot of the microbial 16S rRNA communities of mucosa and lymph node samples using a distance matrix calculated with the Yue and Clayton theta similarity coefficient and a 3% threshold. The Yue and Clayton theta coefficient is used to calculate the similarity in microbial community structure based on the proportions of the different bacterial taxa in each sample. The distance between two points is directly proportional to the Yue and Clayton theta similarity value for two samples, such that sites positioned close together have more similar proportions of the same bacterial taxa than two samples further apart. Proportions of shared and non-shared bacterial sequences of tissue samples from the same patient are highly similar, except for patient Nos 31 and 42, who have one distinct sample each. Patient numbers precede the tissue type labels: IM, involved mucosa; LN, lymph node; NM, normal (uninvolved) mucosa. For non-IBD patients: S1, site 1 mucosa; S2, site 2 mucosa. NMDS stress=0.35.
To determine whether or not there were differences in microbial communities between node and mucosal samples among patients, three matched pair t tests (figure 4, table 3) were carried out using the NMDS values derived from the Jaccard and Yue, and Clayton theta (θYC) distance metrics. The vast majority of the variation observed in microbial communities occurs among individuals, rather than among samples. The match paired t tests eliminate the among patient differences and asks the question, for example, is the average difference (distance) between the position of a lymph node sample and normal mucosal sample for a patient smaller or larger than the distance separating the normal and involved mucosal samples for the same patient (figure 4). Overall, there were no differences between the mean distances of samples from patients with CD (table 3). Although there was evidence for decreased bacterial diversity generally, there was no selection or over-representation of specific bacteria in draining nodes compared with those in adjacent mucosa. There was limited evidence for clustering by disease site, using relative abundance data, in the small number of patients who had disease restricted to the terminal ileum.
The p values for distance comparisons of lymph nodes, and involved and uninvolved microbiota in patients with Crohn's disease
Relative abundance (%) of 16S rRNA sequences of putative Crohn's disease pathogens for lymph node samples taken from Crohn's disease, other IBD and non-IBD bowel resections
As can be seen in figures 2 and 3 (figure 2 is shown above left), the similarity of microbial communities of samples collected from the same patient (eg, patient No 8, circled) are more similar than samples collected from different patients. Standard tests to determine whether lymph node microbial communities are significantly similar or different to mucosal microbial communities would fail, as the differences among individuals would obscure differences between samples from the same individual. To overcome this difficulty, we asked for example: is the microbial community of lymph node more similar to involved mucosa than involved mucosa is to normal mucosa? To do this, we used the t test statistic which compares two means in relation to the variation in the data (distances shown in the non-metric dimensional scaling plots). The mean distances between lymph node and involved mucosa (a) and involved mucosa and normal mucosa (c) were compared for test 1. Test 2 compares the distance between c and b, and test 3, a and b. Table 3 outlines the t test statistics results. IM, involved mucosa; LN, lymph node; NM, normal (uninvolved) mucosa.
Comparison of the predominant phyla in the nodes and mucosa
We sought to explore the nature of the decreased diversity by measuring the relative proportions of Firmicutes and Bacteroidetes for pooled CD and non-IBD mucosa and node samples (figure 5). A small increase in Bacteroidetes was observed relative to Firmicutes for all sample types, but their proportions did not differ between CD and non-IBD nodes (Firmicutes 54% for both).
Proportion of Firmicutes and Bacteroidetes for pooled mucosal and lymph node samples, showing little difference between Crohn's disease (CD) and non-inflammatory bowel disease (IBD) patients. IM, involved mucosa; LN, lymph node; NM, normal (uninvolved) mucosa. For non-IBD patients: S1, site 1 mucosa; S2, site 2 mucosa.
These pooled figures did not reflect individual differences (figure 6). Most sequences in each sample were associated with the three dominant phyla, Firmicutes, Bacteroidetes and Proteobacteria, found in mucosal tissues.15 The majority of patients had similar, but ‘abnormal’, proportions of each phylum for each of their samples. (A healthy microbiota is thought to comprise around 10% Proteobacteria.15 ,16 ,33) We note that three patients (Nos 49, 59, and 60) with the most abnormal proportions of the various phyla had either been on antibiotics for weeks (patient Nos 49 and 60) or for up to 5 days prior to surgery. We did not observe a specific microbial signature in the node in association with an individual disease site in CD patients (figure 7).
Relative abundance of the predominant bacterial phyla of lymph nodes, and uninvolved and involved mucosa for patients with Crohn's disease (CD) or other inflammatory bowel disease (IBD), and lymph nodes, site 1 and site 2 mucosa for non-IBD patients. There is significant variation between, but little difference within, a patient. Sample label: patient ID No, tissue type. IM, involved mucosa; LN, lymph node; NM, normal (uninvolved) mucosa. For non-IBD patients: S1, site 1 mucosa; S2, site 2 mucosa.
Pie charts showing the relative abundance of 16S rRNA sequences classified at the phylum level in Mothur using the Ribosomal Database Project database with a 3% threshold. Patients with the same disease location differ in the proportions of different phyla, and Crohn's disease (CD) patients are as different to one another as controls. L1, terminal ileal CD; L3p, ileocolonic CD with perineal involvement. Refer to table 1 for antibiotic usage.
Antibiotics and other medications have the potential to alter microbial communities,34 so we determined whether or not patients on antibiotics or immunosuppressant medications had differentially abundant genera using the full data set of 58 patients. An analysis in Metastats showed that no comparison of genera between the two groups had p and q values <0.05 (S5): we conclude that immunosuppressives and short term antibiotic treatment did not affect specific genera. We note that two patients who had been on prolonged courses of antibiotics (weeks) had no detectable E coli.
Discussion
We hypothesised that bacterial agents that are able to generate an abnormal immune response in genetically susceptible individuals would be selectively concentrated in lymph nodes: we characterised the microbial communities of resected nodes and mucosa from patients suffering from CD and ‘other disease’ controls. Using a detection rate of 103 bacteria per node, we detected bacterial DNA in nodes of 28% of patients with CD, 25% with other IBD and 48% patients who had a range of conditions, including diverticulitis, intestinal obstruction, ischaemic gut and volvulus. There were no significant differences in the proportion of positive nodes by disease site or the presence of perineal disease.
Although no specific microorganism was present in all bacterial DNA positive nodes of patients with CD, or the associated mucosa, Escherichia/Shigella was detected at high levels in patients with DNA positive lymph nodes (75% of CD and 60% of non-IBD patients, respectively). We also found that patients with terminal ileal CD had a greater proportion of E coli reads in their nodes than patients with colonic or ileocolonic disease. Campylobacteraceae were present in three patients who had L3p CD disease and bacterial DNA positive nodes. We feel that there is an argument for further investigation of these microorganisms in CD. Other specific pathogens were not detected or found in small numbers at low levels.
E coli is the dominant species cultured from nodes of patients with CD20 ,21 ,35 and there is a significant body of work implicating E coli in CD, particularly AIEC.4 ,36 ,37 Long polar fimbriae (lpf) assist AIEC in binding to M cells overlying Peyer's patches38: enhanced binding to M cells and the ability to survive M cell transcytosis may explain why numbers are higher in L1 patients. AIEC are able to evade a major host immune response,39 which could result in enhanced growth and survival.
Many studies have reported the detection of MAP in CD tissues.40 ,41 It has been suggested that MAP occurs in a paucibacillary form in histological lesions42 but that high numbers are present in the lymphatics and nodes draining involved bowel, the mesentery, including adipocytes, and the walls of fistulas. We did not detect Mycobacterium in any sample from 58 patients using methods that included enzymatic and mechanical cell wall disruption.
Our analysis of microbial communities of mucosa and nodes from patients with and without IBD showed that mucosal and node samples from the same patient were similar—that is, the microorganisms found in draining nodes in patients who have late stage disease (requiring surgery) are representative of the microorganisms in the adjacent mucosa. We observed that involved and uninvolved tissues contain similar bacterial communities despite one site having obvious macroscopic inflammation and ulceration. Most studies of involved and uninvolved tissue biopsies from patients with CD found them to be similar,43–45 although some show differences.17 ,46 Given that only 28% of draining nodes from CD patients contained detectable bacterial DNA compared with 48% of nodes from non-IBD resections for conditions associated with gut infection, obstruction and ischaemia, we argue that the translocation of bacterial DNA to draining nodes most likely reflects a breakdown in mucosal barrier function associated with inflammation and ulceration, and the non-selective translocation of locally invasive bacteria to draining nodes by intestinal phagocytes or by lymphatic fluid. We found no evidence for increased bacterial translocation in CD. Others report similar findings: MacFie et al23 performed intraoperative mesenteric node cultures in 927 patients and used multivariate analysis to show that the rate of translocation was increased in those who underwent emergency surgery or who received preoperative total parenteral nutrition, but not in those who had a diagnosis of IBD.
Bacterial diversity in mucosal and node samples from patients with CD was decreased compared with non-IBD patients. Previous studies also reported decreased diversity in patients with CD compared with controls.11 ,47 ,48 Despite differences in diversity, patients did not cluster by disease, but by individual. All patient groups displayed abnormal proportions of the dominant phyla; non-IBD patients had microbiota compositions similar to IBD patients. Overall, pooled CD and non-IBD mucosal and node samples had a moderate increase in the proportion of Bacteroidetes. There are shifts towards an increase in Bacteroidetes in studies of faecal material11 ,47 and in intestinal biopsies17 ,49 of patients with CD. However, individuals vary immensely in the proportions of the dominant phyla,50 and while an increase in Bacteroidetes is often observed for groups of patients, this finding may not apply to an individual. Changes in relative abundance of bacterial phyla residing in the gut may be important in the causation of IBD but may also reflect changes due to inflammation or medications.
HTS can be used to examine the role of bacteria in CD but the results must be interpreted with caution because of interindividual variation and the large number of rare OTUs. Culture dependent techniques have a role in determining whether or not bacteria are viable. Although we were unable to determine whether or not the bacteria detected in the nodes in our study were viable, we know that many species can be cultured from CD nodes.20 In our study, antibiotics were administered to all patients at induction of anaesthesia: this would affect the viability of bacteria and their detection using culture dependent techniques. In contrast, HTS depends on the presence of bacterial DNA, which is unlikely to be affected by short courses of antibiotics. Further work is needed to determine which species are viable, and their effect in patients who have variants in genes involved in bacterial clearance. We also recognise that our non-IBD disease controls were not ‘normal’: they required surgery for conditions including infectious, ischaemic and obstructive disease, but it was not possible to obtain a ‘physiological’ comparator. Normal mucosa is available from patients undergoing resection for colorectal cancer but all nodes are required for staging purposes.
In summary, we compared the microbiota present in nodes, and in affected and unaffected mucosal samples from surgical resections of patients with CD, and non-IBD surgical disease controls. Bacterial DNA was present in the draining nodes of 28%, 25% and 48% of patients suffering from CD, other IBD and non-IBD, respectively. The proportion of positive nodes did not vary by disease site or the presence of perineal disease. Analysis of the microbial communities showed no differences between the microorganisms present in nodes and those in the mucosa of the same patient, suggesting that there was non-selective translocation of bacteria following the breakdown of the mucosal barrier in ‘advanced’ disease. We found that E coli was more abundant in positive nodes from CD patients than from controls, especially in those who had terminal ileal disease, and that Campylobacter was more common in ileocolonic disease with perineal involvement (although the numbers were small). None of the other candidate pathogens or any other microorganism was detected consistently or at significant levels. Although there were significant changes in the degree of bacterial diversity, there was interindividual variation and no characteristic pattern for CD. We believe that future studies should focus on early disease and viable bacteria in nodes, aphthous ulcers and granulomas of patients with CD, as they may be more relevant in the initiation of inflammation in CD.
Acknowledgments
We would like to thank the patients who agreed to be involved in this study, and the surgeons (Drs Ian Davis, David Rangiah, Phil Jeans and Peter Barry), nurses and pathology staff who also assisted. Particular thanks to Professor Jane Dahlstrom of ACT Anatomical Pathology.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
- Data supplement 2 - Online table
Footnotes
-
Contributors CLO: study concept and design; acquisition of the data; analysis and interpretation of the data; drafting of the manuscript; critical revision of the manuscript for important intellectual content; and statistical analysis. DMG: statistical analysis; and critical revision of the manuscript for important intellectual content. GEA: study concept and design; analysis and interpretation of the data; critical revision of the manuscript for important intellectual content; and administrative and study supervision. PP: study concept and design; analysis and interpretation of the data; critical revision of the manuscript for important intellectual content; obtained funding; and administrative and study supervision.
-
Competing interests None.
-
Ethics approval The study was approved by: Australian Capital Territory Health Protocol Nos ETH.6/03.276, ETH.9/99.252 and ETH.5/07.464; and Australian National University Protocol Nos 2007/2213 and 2007/131 (LESC-CMHS).
-
Provenance and peer review Not commissioned; externally peer reviewed.
-
Data sharing statement The raw data (sff files) from the 454 sequencer and all metadata, as prescribed by MIMARKS (Yilmaz et al 2011), will be made publicly available via the following link on acceptance: http://dx.doi.org/10.4225/13/519428341DBA6. Use of these data should be acknowledged.