A Distinct Gut Microbiota Exists Within Crohn's Disease–Related Perianal Fistulae

j o u r n a l o f s u r g i c a l r e s e a r c h  o c t o b e r 2 0 1 9 ( 2 4 2 ) 1 1 8 e1 2 8

Available online at www.sciencedirect.com

ScienceDirect journal homepage: www.JournalofSurgicalResearch.com

A Distinct Gut Microbiota Exists Within Crohn’s DiseaseeRelated Perianal Fistulae Bryce E. Haac, MD,a Nicholas C. Palmateer, BS,b Max E. Seaton, MD,a Ryan VanYPeren, PhD,b Claire M. Fraser, PhD,b and Andrea C. Bafford, MDa,* a b

Department of Surgery, University of Maryland, Baltimore, Maryland Institute for Genome Sciences, University of Maryland, Baltimore, Maryland

article info

abstract

Article history:

Background: Gut bacteria are strongly suspected to play a key role in the pathogenesis of

Received 7 February 2019

Crohn’s disease (CD). Studies have demonstrated alterations in the gut microbiota in this

Received in revised form

patient population. The purpose of this study was to characterize the gut microbiota of

27 March 2019

fistulizing perianal CD.

Accepted 9 April 2019

Materials and Methods: Stool and fistula samples were obtained from patients undergoing

Available online 7 May 2019

surgery for CD-related anorectal fistulae. Microbial compositions of matched stool and fistula samples were characterized using 16S rRNA gene profiling. The effect of sample

Keywords:

type, patient gender, disease classification (Montreal A/B), disease activity (Harvey Brad-

Microbiome

shaw Index), antibiotic use, and presence of active proctitis on microbial composition was

Microbiota

assessed.

Crohn’s disease

Results: Samples were obtained from 18 patients. Bacteroides was the most abundant

Perianal

genera across all samples collected, followed by Streptococcus and Bifidobacterium. Bifido-

Fistula

bacterium was present at significantly higher levels in fecal samples than fistula samples, whereas Achromobacter and Corynebacterium were present at significantly higher levels in fistula samples. Antibiotic, but not thiopurine or antitumor necrosis factor medication, exposure affected the gut microbial composition. Patient gender, disease classification, disease activity, and presence of active proctitis did not alter stool or fistula microbiota. Conclusions: Our data show that the gut microbiota within CD-related anorectal fistulae is distinct from that in stool samples obtained from the same patients. We also observe a dysbiosis in patients treated with antibiotics compared with those not treated with antibiotics. ª 2019 Elsevier Inc. All rights reserved.

Introduction Inflammatory bowel disease (IBD) is a chronic, relapsing inflammatory condition of the gastrointestinal tract, which affects 1.0 to 1.5 million Americans.1 Two subtypes exist, ulcerative colitis and Crohn’s disease (CD). Although the exact

etiology of IBD is unclear, the prevailing model of pathogenesis has long been a dysregulated mucosal immune response triggered by an environmental factor in genetically susceptible hosts. Gut bacteria are strongly suspected to be the inflammatory trigger in most cases. Indeed, there is a growing body of literature showing both abnormalities in the gut

* Corresponding author. Department of Surgery, University of Maryland School of Medicine, 22 South Greene Street, Baltimore, MD 21201. Tel.: þ1 410 328 6187, fax: þ1 410 328 5919. E-mail address: [email protected] (A.C. Bafford). 0022-4804/$ e see front matter ª 2019 Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.jss.2019.04.032

119

haac et al  gut microbiome of perianal crohn’s

mucosal immune response and alterations in the gut microbiota, in patients with IBD. Several facts support the hypothesis that intestinal bacteria play an important role in the pathogenesis of IBD. First, IBD affects areas of the intestinal tract with the highest concentration of bacteria. Diverting the fecal stream away from these affected areas leads to disease remission in most patients.2 Furthermore, antibiotics, in particular, ciprofloxacin, metronidazole, and rifaximin, have been shown to induce disease remission in patients with CD.3 Possible mechanisms of disease include expansion of proinflammatory species, a reduction in protective microbes, and/or the presence of a classic pathogen for IBD, such as Helicobacter pylori in peptic ulcer disease.4 Mutations in the NOD2/CARD15 gene, a gene involved in macrophage activation in response to intracellular lipopolysaccharides, are thought to account for 15%-30% of cases of CD.5,6 Variations in autophagy genes ATG16L1 and IRGM have also been linked to CD.7,8 These genetic associations further support the belief that aberrations in the innate host defense against enteric bacteria leads to IBD. Population-based studies show a cumulative incidence of perianal fistulae in CD of 23% to 38%.9,10 Symptoms vary from anal pain and purulent discharge to bleeding and incontinence and can be associated with significant morbidity and impaired quality of life.11-13 A fistula is defined as an abnormal communication between two epithelial surfaces, in this case, skin and anus or rectum. Fistulae are thought to develop either because of extension of a deep penetrating ulcer of the

rectum or anus or anal gland infection leading to local sepsis and tract formation.14 Studies have demonstrated improvement in fistulizing perianal CD with metronidazole15 and ciprofloxacin16 treatment, arguing for a pathologic role of the microbiota located in and around the fistula. The aim of this study was to characterize the gut microbiota in patients with CD-related anorectal fistulae, specifically comparing the luminal bacterial population with those seen within fistula tracts, to begin to understand possible underlying microbial patterns in disease causation across a diverse set of patients.

Materials and methods Study design and patient identification This study was approved by the University of Maryland Medical Center Institutional Review Board and was carried out in accordance with The Code of Ethics of the World Medical Association. After institutional review board approval was granted, consecutive patients with perianal CD undergoing surgery for anorectal fistulae at the University of Maryland Medical Center were approached for study participation. Once informed consent was obtained, stool and fistula samples were collected from each patient intraoperatively. Fecal specimens were collected via digital rectal examination. Fistula samples were collected from within the fistula tracts using a single passage of a small curette.

Table 1 e Patient and disease characteristics. Gender

Montreal A*

Montreal By

HBIz

Thiopurinesx

Anti-TNFk

Antibiotics{

Proctitis

M

A3

B1p

4

Y

Y

Y

Y

2

F

A2

B2p

5

N

Y

Y

Y

3

M

A3

P

1

N

Y

N

N

Patient 1

4

F

A2

B2p

3

N

N

N

Y

5

M

A2

B3p

1

N

Y

N

N

6

F

-

B1p

3

Y

Y

Y

-

7

F

A2

B2p

7

N

Y

Y

N

8

M

A2

B1p

15

Y

Y

N

N

9

M

A3

B1p

-

N

N

Y

Y

10

F

A1

B1p

6

N

Y

N

Y

11

F

A1

B1p

-

N

N

Y

N

12

M

A1

B1p

4

N

N

N

Y

13

M

A3

B1p

2

Y

Y

Y

Y

14

M

A2

B1p

9

N

N

Y

N

15

F

A1

B2p

9

N

Y

N

Y

16

F

A2

B2p

7

N

Y

Y

N

17

M

A2

B1p

5

N

Y

Y

Y

18

M

A1

B2p

0

N

Y

N

Y

*

Montreal A (age at diagnosis): A1: 16 years; A2: 17-40 years; and A3: >40 years. Montreal B (disease behavior): B1p: nonstricturing and nonpenetrating with perianal disease; B2p: stricturing with perianal disease; B3p: penetrating with perianal disease; P: perianal disease alone. z HBI (Harvey-Bradshaw Index): <5: remission; 5-7: mild disease; 8-16: moderate disease. x Thiopurines: azathioprine, 6-mercaptopurine. k Anti-TNF agents: infliximab, adalimumab, certolizumab pegol, vedolizumab. { Antibiotics (ciprofloxacin, metronidazole, ampicillin/sulbactam, clindamycin, or combination) initiated 24 h before surgery. y

120

j o u r n a l o f s u r g i c a l r e s e a r c h  o c t o b e r 2 0 1 9 ( 2 4 2 ) 1 1 8 e1 2 8

Fig. 1 e Sample composition in antibiotic-treated and untreated patients. Based on the 25 taxa with the highest relative abundance in all samples, the proportion of these taxa that are found in each sample collected from patients enrolled in the study is shown. Each bar represents a single sample paired based on the patient it was derived from and separated based on antibiotic status.

Patients did not receive preoperative bowel preparation. Samples were transferred to sterile containers and immediately placed on ice. Samples were then stored at 70 until DNA processing. Patient demographic and CD-related data were collected from the institution’s electronic medical

record (Table 1). Patients included in the study had a mix of acute and chronic presentations. Eight patients had a symptomatic fistula. Six patients had an acute abscess. Four patients had an inactive fistula, and surgery was for fistulotomy or seton placement.

haac et al  gut microbiome of perianal crohn’s

Table 2 e Antibiotic treatment. Patient

Antibiotic

Duration of treatment (d)

1

Ciprofloxacin

>30

2

Metronidazole

10

6

Ciprofloxacin þ metronidazole

8

7

Ampicillin/sulbactam þ metronidazole

1

9

Ciprofloxacin þ metronidazole

2

11

Ciprofloxacin þ metronidazole

>30

13

Ciprofloxacin þ metronidazole

14

14

Ciprofloxacin þ metronidazole

18

16

Ciprofloxacin

6

17

Clindamycin

10

121

spike.19 Raw reads were processed with trimmomatic, and paired ends were assembled using fast length adjustment of short reads20 with error correction on the w90 bp overlapping region. We used dual barcoding information with one barcode on each paired end read to index individual samples. The sequences were then demultiplexed by binning sequences with the same barcode in QIIME (version 1.8.0).21 QIIME quality trimming was performed using the following criteria: 1) truncation of sequences before three consecutive low-quality bases and re-evaluation for length, 2) no ambiguous base calls, and 3) minimum sequence length of 230 bp after trimming. Barcode sequences, heterogeneity spacers, and primers were further trimmed off. Similar sequences with less than 3% dissimilarity were clustered together using USEARCH (version 5.2.32),22 and de novo chimera detection was conducted in UCHIME (version 5.1).22,23

DNA extraction

Data analysis

For DNA extraction, equal amounts of each sample type were transferred to DNA/RNA-free sterile tubes and 1 mL of phosphate-buffered saline was added to each sample. Cells were lysed by adding 5 mL of lysozyme (100 mg/mL) and 15 mL of mutanolysin (5000 units/mL). After a 30-min incubation period, 10 mL of Proteinase K (at 20 mg/mL) and 50 mL of 10% sodium dodecyl sulfate were added, and cells were incubated for an additional 45 min. The samples were treated by bead beating performed in the FastPrep instrument FP120 at 6 m/s for 40 s using 0.1 mm silica spheres (QBiogene Lysis Matrix B, MP Biomedicals, Santa Ana, CA). Total DNA was extracted from the lysate using the ZR fecal DNA isolation kit (D6010; Zymo Research Corp., Irvine, CA). Negative control extractions were processed in parallel to monitor for potential contamination.17

Sample composition and diversity were evaluated using the phyloseq analysis package.24 Agglomeration of operational taxonomic units (OTUs) was at the genus rank to merge species that have the same taxonomy at the genus level or below. To make OTU abundance more easily comparable, total abundance was converted to relative abundance. Finally, taxa that were determined to be at very low abundance (<1  105) were removed. To compare statistical significance of variation due to sample type or patient, the statistical tests adonis and analysis of similarities (ANOSIM) of the QIIME package were used to evaluate weighted UniFrac distance matrices.25 Sample diversity was measured using the Shannon Diversity Index, which takes into account both abundance and evenness of the species present.26 The effect of sample type (fecal versus fistula), patient gender, disease classification (Montreal A/B), disease activity (Harvey Bradshaw Index), thiopurine use, antietumor necrosis factor (TNF) use, antibiotic use, and presence of active proctitis were assessed using the JensenShannon divergence between the microbial communities of interest to perform principal coordinate analyses.27 Comparisons were made between sample groups (sample type, antibiotic status) by calculating the linear discriminant analysis score and determining its effect size using the linear discriminant analysis effect size (LEfSe) tool.28 This tool allows for the discovery of biomarkers between two populations using relative abundance.

16S rRNA gene sequence processing and analysis The V3-V4 regions of the 16S rRNA gene, which are of optimal size for sequencing on the Illumina sequencing platform, were amplified as described by Fadrosh et al, 2014,18 using the following cycling parameters: 5 min of denaturation at 95 C, followed by 25 cycles of 30 s at 95 C (denaturing), 30 s at 58 C (annealing), and 90 s at 72 C (elongation), with a final extension at 72 C for 5 min. Resulting amplicons were diluted 1:20 and used as templates for a second polymerase chain reaction step to add a sample specific dual-index barcode and flow cell linker adaptors. Phusion Taq Master Mix was added to 3% dimethyl sulfoxide and primers diluted to 0.4 mM under the following cycling conditions: initial denaturation at 95 C for 30 s, 10 cycles consisting of denaturation at 95 C for 30 s, annealing at 58 C for 30 s, and elongation at 72 C for 60 s, followed by a final elongation step at 72 C for 5 min. Amplicons were cleaned with the SequalPrep kit (Invitrogen, Carlsbad, CA). Negative controls were tested in parallel to monitor for potential contamination. The success or failure of sequence library preparation was measured by the presence or absence of an amplicon band on an agarose gel.

Sequencing by Illumina HiSeq and sequence processing Samples were sequenced using HiSeq 2500 Rapid Run 300 bp paired end sequencing with 8pM targeted loads and 5% phiX

Results Sample composition The composition of matched fecal and fistula tract samples from 18 patients diagnosed with perianal CD was determined using 16S rRNA gene profiling. A total of 34,192 OTUs were identified in all the samples from this study. The 25 most abundant genera represented between 50% and 95% of every sample were examined (Fig. 1). Bacteroides was the most abundant OTU in 11 of the 36 samples (7 fecal and 4 fistula biopsy) followed by Streptococcus (2 fecal and 2 fistula biopsy). Twelve taxa from the Bacteroidetes, Firmicutes, and

122

j o u r n a l o f s u r g i c a l r e s e a r c h  o c t o b e r 2 0 1 9 ( 2 4 2 ) 1 1 8 e1 2 8

Fig. 2 e Linear discriminant analysis effect size (LEfSe) analysis. The linear discriminant analysis (LDA) score of taxa are significantly different when comparing biopsy and fecal samples. The length of the bar represents the LDA score (log scale), which corresponds to the differentially abundant taxa of statistical significance. (Color version of figure is available online.)

123

haac et al  gut microbiome of perianal crohn’s

in fecal samples than fistula biopsy samples, whereas Achromobacter and Corynebacterium were present at significantly higher levels in biopsy samples (Fig. 2). Although there were certain taxa that differed significantly between the two types of samples, overall, the differences in the composition of the microbiota varied more because of the patient than sample type. Based on the adonis statistical method, only 6.2% of the variation in a UniFrac distance matrix for all samples is due to the sample type (P-value ¼ 0.016), whereas 67% of variation is from the individual (P-value ¼ 0.001). This was also supported by the ANOSIM statistical test. The ANOSIM test yielded an Rvalue of 0.11 (P ¼ 0.011) when grouping OTU reads by sample type and an R-value of 0.52 (P ¼ 0.001) when grouping by patient (Table 3). Because an R-value at or near zero indicates no significant dissimilarity between the groups, we conclude that variability is more due to interindividual differences than sample type.

Table 3 e Overlap in the 25 most abundant genera in biopsy and fecal samples. Biopsy

Biopsy and fecal

Fecal

Bacteroides

Akkermansia

Actinomyces

Bifidobacterium

Collinsella

Anaerococcus

Blautia

Dorea

Clostridium

Enterococcus

Corynebacterium

Eubacterium

Dialister

Lactobacillus

Achromobacter

Cloacibacterium Finegoldia Fusobacterium Parvimonas

Megasphaera

Oscillospira

Sneathia

Peptoniphilus

Parabacteroides

Staphylococcus

Peptostreptococcus Prevotella Ruminococcus Streptococcus

Some operational taxonomic units (OTUs) are not identified at the genus level.

Impact of preoperative immunosuppressive and antibiotic therapy Four of 18 (22%) and 13 of 18 (72%) patients were exposed to thiopurine and/or anti-TNF immunomodulating therapy, preoperatively. Ten of 18 (56%) patients received antibiotics for at least 24 h before surgery (Table 4). Thiopurine and antiTNF medication exposure did not appear to affect the composition of the gut microbiota. However, a pattern of sample dissimilarity was observed between patients who received preoperative antibiotics and those who did not. Samples taken from patients receiving preoperative antibiotics exhibited a greater level of alpha diversity, as measured by the Shannon Diversity Index (Fig. 3). As shown in Figure 4A and B, where more than 50% of variation is observed in the first two axes of principal coordinate analyses, patients clustered based on antibiotic exposure. This was observed in both fecal and fistula biopsy samples collected from the patients.

Actinobacteria phyla were found among the 25 most abundant genera in both fecal and fistula biopsy samples (Table 2). Differing microbial composition was observed based on sample type (stool versus fistula) and the use of preoperative antibiotics in study participants; however, patient gender, disease classification, disease activity, preoperative thiopurine and anti-TNF use, and presence of active proctitis did not discriminate across the samples.

Sample type comparison We observed several taxa that were significantly more abundant in one sample type than the other. Using the linear discriminant analysis effect size (LEfSe) score, we observed that Bifidobacterium was present at significantly higher levels

Table 4 e Statistical analysis of sample groupings using weighted distance matrices. Df

SumsOfSqs

MeanSqs

R2

Pr (>F)

0.06208

0.016

y

0.001

*

F.Model

ADONIS comparison by sample type 1

0.5263

0.52634

Residuals

Sample type

34

7.9517

0.23387

Total

35

8.4781

2.2505

0.93792 1

ADONIS comparison by patient Patient

17

5.6815

0.3342

Residuals

18

2.7966

0.15537

Total Method name

35 R statistic

8.4781 P-value

0.1131

0.011

999

0.001

999

ANOSIM comparison by patient ANOSIM *

0.5196

Significance code: 0.001. y Significance code: 0.05.

0.67014 0.32986 1

Number of permutations

ANOSIM comparison by sample type ANOSIM

2.1511

124

j o u r n a l o f s u r g i c a l r e s e a r c h  o c t o b e r 2 0 1 9 ( 2 4 2 ) 1 1 8 e1 2 8

Fig. 3 e Shannon Diversity. The Shannon Diversity Index characterizes species diversity. The graph shown examines the level of diversity in each sample type (biopsy or fecal) based on the antibiotic status of the patient. Fecal samples from patients treated with antibiotics had a higher Shannon Diversity Index value, whereas biopsy samples had similar Shannon Index values for samples from patients treated with antibiotics and those untreated.

haac et al  gut microbiome of perianal crohn’s

125

Fig. 4 e (A-B) Principal coordinate analysis (PCoA) examining similarity of patients treated with antibiotics. Based on the Jensen-Shannon divergence (JSD) of (A) biopsy and (B) fecal samples, the similarity of samples from patients treated with or without antibiotics is shown. The above PCoAs capture (A) 52.2% and (B) 55% of the variation within the first two axes. (Color version of figure is available online.)

Discussion and conclusions A number of published studies have demonstrated alterations in the gut microbial community in patients with IBD. These include a higher total number of bacteria on the mucosal surface of patients with IBD than non-IBD controls,29 a reduction in overall bacterial diversity due to loss of several butyrate-producing bacteria, including Bacteroidetes and Firmicutes, in particular, the anaerobes, Clostridia and Lactobacilli, and an increase in Enterobacteriaceae and other Proteobacteria in

patients with IBD.30-32 Previous studies evaluating the microbiota within CD-related and idiopathic anal fistula tracts showed low levels of bacteria, prompting study authors to conclude that bacterial infection is unlikely to play a significant role in fistula persistence.33,34 In the present study, we characterized the gut microbiota in fecal and fistula tract samples from patients with active perianal CD undergoing surgery. Bacteroides, Streptococcus, and Bifidobacterium were the most abundant genera in both stool and fistula samples. Bifidobacterium, a common colonic microbe, was found at significantly higher levels in fecal than fistula samples. We also

126

j o u r n a l o f s u r g i c a l r e s e a r c h  o c t o b e r 2 0 1 9 ( 2 4 2 ) 1 1 8 e1 2 8

Fig. 4 e (continued).

observed an unexpected abundance of Achromobacter in fistula biopsy specimens relative to fecal specimens (Fig. 2). Achromobacter, previously considered a rare opportunistic pathogen, has recently been implicated in airway infection and loss of airway microbial diversity in patients with cystic fibrosis.35,36 Similar to those with cystic fibrosis, patients with fistulizing perianal CD frequently receive immunosuppressive and antibiotic therapy. Based on our data, it could be of value to further investigate Achromobacter for a potential role in CDrelated perianal infections. Similar to past studies,37 we also found an increased abundance of Corynebacterium in the biopsy samples from our patients (Fig. 2). This is not unexpected given the persistent communication between perianal skin and the anorectal

lumen created by fistulae. Corynebacterium, frequently found as part of the normal skin microbiota, has been increasingly recognized for its role in causing opportunistic infection in immunocompromised patients.38 Finally, Fusobacteria and Actinomyces, microbes previously linked to colorectal cancer,39,40 were also abundant in our patients. Multiple studies have demonstrated an elevated risk of colorectal cancer in patients with CD.41-43 Our data are consistent with the possibility that this risk may be related to colonic dysbiosis and this is also an avenue for future investigation. Studies have demonstrated improvement in perianal CD with antibiotic treatment, in particular ciprofloxacin and metronidazole.15,16 Indeed, we found that antibiotic exposure correlated with a distinct difference in both stool and fistula

haac et al  gut microbiome of perianal crohn’s

samples. Antibiotic efficacy may be related to decreased levels of proinflammatory bacteria or improved growth of protective species. In this study, unexpectedly high levels of Achromobacter and Corynebacteria were found in fistula tissue; neither of which are optimally treated with ciprofloxacin and/or metronidazole. Alternative antibiotic regimens may be indicated in patients with refractory perianal CD. It remains unclear whether IBD-associated alterations in the gut microbiota are the cause, consequence of, or unrelated to the disease. Our data show that differences exist between the gut microbiota of stool samples and that of the fistula tracts themselves in perianal CD. We further discovered a difference in the composition of the microbiota in patients treated with antibiotics compared with those not. In the future, further description of the microbial changes associated with antibiotic exposure in patients with CD, particularly if related to disease activity, may reveal organisms, which either drive or protect against inflammation. Our study has some limitations. This is a pilot study with a small number of enrolled patients. In addition, we would have ideally obtained stool samples from our patients at times of disease remission to serve as controls. However, given that patients with CD often have some level of active disease present, this was not deemed feasible. Rather, we elected to compare fistula with fecal bacteria presuming that alterations found may play a role in disease pathogenesis. Another option would have been to use healthy patients as controls; however, studies have shown that the gut microbiota varies significantly between individuals and even within a single individual over time.44,45 Owing to these significant differences, comparing microbiota between individuals is not as helpful as one might think. Instead, patients are likely their own best controls. In addition, because there is evidence that bacteria can vary in the same patient over time, samples obtained at future time points may also be inadequate controls. In this study, we examined differences between the microbiota of the gut lumen and the fistula tract, with the lumen serving as the control. In addition, unlike stool samples acquired by digital examination, fistula samples were obtained via gentle curettage because of their limited size. As a result, these samples may contain some mucosa-associated bacteria in addition to fecal bacteria. Yasuda et al. demonstrated that stool microbial composition accurately reflects the colonic lumen and mucosa in the rhesus macaque, an animal with a gut microbiome fundamentally similar to that of humans.46 However, although fecal samples are generally thought to be representative of the background mucosal bacterial biofilm, some studies have shown otherwise.47 Nonetheless, our data show a distinct microbiota within CD-related perianal fistulae. Further characterizing these changes may lead to better understanding of disease pathogenesis and improvements in antimicrobial therapy.

Acknowledgment Authors’ contributions: A.C.B. and C.M.F. conceived of the presented idea. A.C.B., N.C.P., and R.V. carried out the

127

experiment and analyzed the data. A.C.B., B.E.H., and N.C.P. wrote the manuscript with support from M.E.S. C.M.F. supervised the project. All authors discussed the results and contributed to the final manuscript.

Disclosure The authors reported no proprietary or commercial interest in any product mentioned or concept discussed in this article.

references

1. Kappelman M, Rifas-Shiman SL, Porter CQ, et al. Direct health care costs of Crohn’s disease and ulcerative colitis in US children and adults. Gastroenterology. 2008;135:1907e1913. 2. Winslet MC, Allan A, Poxon V, Youngs D, Keighley MR. Faecal diversion for Crohn’s colitis: a model to study the role of the faecal stream in the inflammatory process. Gut. 1994;35:236e242. 3. Khan KJ, Ullman TA, Ford AC, et al. Antibiotic therapy in inflammatory bowel disease: a systematic review and metaanalysis. Am J Gastroenterol. 2011;106:661e673. 4. Marshall BJ, Warren JR. Unidentified curved bacilli in the stomach of patients with gastritis and peptic ulceration. Lancet. 1984;1:1311e1315. 5. Hugot JP, Chamaillard M, Zouali H, et al. Association of NOD2 leucine-rich repeat variants with susceptibility to Crohn’s disease. Nature. 2001;411:559e603. 6. Ogura Y, Bonen DK, Inohara N, et al. A frameshift mutation in NOD2 associated with susceptibility to Crohn’s disease. Nature. 2001;411:603e606, 129e137. 7. Rioux JD, Xavier RJ, Taylor KD, et al. Genome-wide association study identifies new susceptibility loci for Crohn disease and implicates autophagy in disease pathogenesis. Nat Genet. 2007;39:596. 8. Parkes M, Barrett JC, Prescott NJ, et al. Sequence variants in the autophagy gene IRGM and multiple other replicating loci contribute to Crohn’s disease susceptibility. Nat Genet. 2007;39:830. 9. Hellers G, Bergstrand O, Ewerth S, Holmstro¨m B. Occurrence and outcome after primary treatment of anal fistulae in Crohn’s disease. Gut. 1980;21:525e527. 10. Schwartz DA, Loftus EV, Tremaine WJ, et al. The natural history of fistulizing Crohn’s disease in Olmsted County, Minnesota. Gastroenterology. 2002;122:875e880. 11. Rankin GB, Watts HD, Melnyk CS, Kelley Jr ML. National Cooperative Crohn’s Disease Study: extraintestinal manifestations and perianal complications. Gastroenterology. 1979;77(4 Pt 2):914. 12. McKee RF, Keenan RA. Perianal Crohn’s disease–is it all bad news? Dis Colon Rectum. 1996;39:136. 13. Platell C, Mackay J, Collopy B, Fink R, Ryan P, Woods R. Anal pathology in patients with Crohn’s disease. Aust N Z J Surg. 1996;66:5. 14. Safar B, Sands D. Perianal Crohn’s disease. Clin Colon Rectal Surg. 2007;20:282e293. 15. Bernstein LH, Frank MS, Brandt LJ, Boley SJ. Healing of perineal Crohn’s disease with metronidazole. Gastroenterology. 1980;79:357e365. 16. West RL, Van der Woude CJ, Hansen BE, et al. Clinical and endosonographic effect of ciprofloxacin on treatment of perianal fistulae in Crohn’s disease with infliximab: a doubleblind placebo-controlled study. Aliment Pharmacol Ther. 2004;20:1329e1336.

128

j o u r n a l o f s u r g i c a l r e s e a r c h  o c t o b e r 2 0 1 9 ( 2 4 2 ) 1 1 8 e1 2 8

17. Zupancic ML, Cantarel BL, Liu Z, et al. Analysis of the gut microbiota in the old order Amish and its relation to the metabolic syndrome. PLoS One. 2012;7:e43052. 18. Fadrosh DW, Ma B, Gajer P, et al. An improved dual-indexing approach for multiplexed 16S rRNA gene sequencing on the Illumina MiSeq platform. Microbiome. 2014;2:6. 19. 16S Metagenomic sequencing library preparation. 2013. Available at: https://www.illumina.com/content/dam/ illumina-support/documents/documentation/chemistry_ documentation/16s/16s-metagenomic-library-prep-guide15044223-b.pdf. 20. Magoc T, Salzberg SL. FLASH: fast length adjustment of short reads to improve genome assemblies. Bioinformatics. 2011;27:2957e2963. 21. Caporaso JG, Kuczynski J, Stombaugh J, et al. QIIME allows analysis of high-throughput community sequencing data. Nat Methods. 2010;7:335e336. 22. Edgar RC. Search and clustering orders of magnitude faster than BLAST. Bioinformatics. 2010;26:2460e2461. 23. Edgar RC, Haas BJ, Clemente JC, Quince C, Knight R. UCHIME improves sensitivity and speed of chimera detection. Bioinformatics. 2011;27:2194e2200. 24. McMurdie PJ, Holmes S. phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data. PLoS One. 2013;8:e61217. 25. Navas-Molina JA, Peralta-Sa´nchez JM, Gonza´lez A, et al. Advancing our understanding of the human microbiome using QIIME. Methods Enzymol. 2013;531:371e444. 26. Shannon CE. A mathematical theory of communication. ACM SIGMOBILE Mobile. Comput Commun Rev. 2001;5:3e55. 27. Fuglede B, Topsoe F. “Jensen-Shannon divergence and hilbert space embedding”. In: Proc Int Symp Inf Theory. 2004:31. 28. Segata N, Izard J, Waldron L, et al. Metagenomic biomarker discovery and explanation. Genome Biol. 2011;12:R60. 29. Kleessen B, Kroesen AJ, Buhr HJ, Blaut M. Mucosal and invading bacteria in patients with inflammatory bowel disease compared with controls. Scand J Gastroenterol. 2002;37:1034e1041. 30. Manichanh C, Rigottier-Gois L, Bonnaud E, et al. Reduced diversity of faecal microbiota in Crohn’s disease revealed by a metagenomic approach. Gut. 2006;55:205e211. 31. Gophna U, Sommerfeld K, Gophna S, Doolittle WF, Veldhuyzen van Zanten SJ. Differences between tissue-associated intestinal microfloras of patients with Crohn’s disease and ulcerative colitis. J Clin Microbiol. 2006;44:4136e4141. 32. Chen L, Wang W, Zhou R, et al. Characteristics of fecal and mucosa-associated microbiota in Chinese patients with inflammatory bowel disease. Medicine (Baltimore). 2014;93:e51.

33. Tozer PJ, Rayment N, Hart AL, et al. What role do bacteria play in persisting fistula formation in idiopathic and Crohn’s anal fistula? Colorectal Dis. 2015;17:235e241. 34. van Onkelen RS, Mitalas LE, Gosselink MP, et al. Assessment of microbiota and peptidoglycan in perianal fistulas. Diagn Microbiol Infect Dis. 2013;75:50e54. 35. Parkins MD, Floto RA. Emerging bacterial pathogens and changing concepts of bacterial pathogenesis in cystic fibrosis. J Cyst Fibros. 2015;14:293e304. 36. Talbot NP, Flight WG. Severe Achromobacter xylosoxidans infection and loss of sputum bacterial diversity in an adult patient with cystic fibrosis. Paediatr Respir Rev. 2016;20(Suppl l):27e29. 37. West RL, Van der Woude CJ, Endtz HP, et al. Perianal fistulae in Crohn’s disease are predominantly colonized by skin flora: implication for antibiotic treatment. Dig Dis Sci. 2005;50:1260e1263. 38. Funke G, von Graevenitz A, Clarridge III J, Bernard KA. Clinical microbiology of coryneform bacteria. Clin Microbiol Rev. 1997;10:125e159. 39. Kostic AD, Gevers D, Pedamallu CS, et al. Genomic analysis identifies association of Fusobacterium with colorectal carcinoma. Genome Res. 2012;22:292e298. 40. Castellarin M, Warren RL, Freeman JD, et al. Fusobacterium nucleatum infection is prevalent in human colorectal carcinoma. Genome Res. 2012;22:299e306. 41. Weedon DD, Shorter RG, Ilstrup DM, Huizenga KA, Taylor WF. Crohn’s disease and cancer. N Engl J Med. 1973;289:1099e1103. 42. Gyde SN, Prior P, Macartney JC, Thompson H, Waterhouse JA, Allan RN. Malignancy in Crohn’s disease. Gut. 1980;21:1024e1029. 43. Bernstein CN, Blanchard JF, Kliewer E, Wajda A. Cancer risk in patients with inflammatory bowel disease: a populationbased study. Cancer. 2001;91:854e862. 44. Eckburg PB, Bik EM, Bernstein CN, et al. Diversity of the human intestinal microbial flora. Science. 2005;308: 1635e1638. 45. Costello EK, Lauber CL, Hamady M, Fierer N, Gordon JI, Knight R. Bacterial community variation in human body habitats across space and time. Science. 2009;326:1694e1697. 46. Yasuda K, Oh K, Ren B, et al. Biogeography of the intestinal mucosal and luminal microbiome in the rhesus macaque. Cell Host Microbe. 2015;17:385e391. 47. Durba´n A, Abella´n JJ, Jime´nez-Herna´ndez N, et al. Assessing gut microbial diversity from feces and rectal mucosa. Microb Ecol. 2011;61:123e133.