Distinct microbiome in pouchitis compared to healthy pouches in ulcerative colitis and familial adenomatous polyposis.

Abstract

Pouchitis occurs in up to 50% of patients with ulcerative colitis (UC) undergoing ileal pouch anal anastomosis (IPAA). Pouchitis rarely occurs in patients with familial adenomatous polyposis (FAP) who undergo IPAA. Our aim was to compare mucosal and luminal flora in patients with UC-associated pouchitis (UCP), healthy UC pouches (HUC), and healthy FAP pouches (FAP).
Nineteen patients were enrolled in this cross-sectional study (nine UCP, three HUC, seven FAP). Patients with active pouchitis were identified using the Pouchitis Disease Activity Index (PDAI). Ileal pouch mucosal biopsies and fecal samples were analyzed with a 16S rDNA-based terminal restriction fragment length polymorphism (TRFLP) approach. Pooled fecal DNA from four UCP and four FAP pouches were sequenced for further speciation.
TRFLP data revealed statistically significant differences in the mucosal and fecal microbiota between each group of patients. UCP samples exhibited significantly more TRFLP peaks matching Clostridium and Eubacterium genera compared to HUC and FAP pouches and fewer peaks matching Lactobacillus and Streptococcus genera compared to FAP. DNA Sanger sequencing of a subset of luminal samples revealed UCP having more identifiable sequences of Firmicutes (51.2% versus 21.2%) and Verrucomicrobia (20.2% versus 3.2%), and fewer Bacteroidetes (17.9% versus 60.5%) and Proteobacteria (9.8% versus 14.7%) compared to FAP.
The pouch microbial environment appears to be distinctly different in the settings of UC pouchitis, healthy UC, and FAP. These findings suggest that a dysbiosis may exist in pouchitis which may be central to understanding the disease.

Full text

Page 1

Distinct Microbiome in Pouchitis Compared to Healthy Pouches in Ulcerative Colitis and
Familial Adenomatous Polyposis
Garrett C. Zella, MD, Elizabeth J. Hait, MD, MPH, Tiffany Glavan, BA, Dirk Gevers,
PhD, Doyle V. Ward, PhD, Christopher L. Kitts, PhD, and Joshua R. Korzenik, MD
Background: Pouchitis occurs in up to 50% of patients with ulcerative colitis (UC)
undergoing Heal pouch anal anastomosis (IPAA). Pouchitis rarely occurs in patients with
familial adenomatous polyposis (FAP) who undergo IPAA. Our aim was to compare mucosal
and luminal flora in patients with UC-associated pouchitis (UCP), healthy UC pouches (HUC),
and healthy FAP pouches (FAP).
Methods: Nineteen patients were enrolled in this cross-sectional study (nine UCP, three HUC,
seven FAP). Patients with active pouchitis were identified using the Pouchitis Disease
Activity Index (PDAI). Heal pouch mucosal biopsies and fecal samples were analyzed with a
16S rDNA-based terminal restriction fragment length polymorphism (TRFLP) approach. Pooled
fecal DNA from four UCP and four FAP pouches were sequenced for further speciation.
Results: TRFLP data revealed statistically significant differences in the mucosal and fecal
microbiota between each group of patients. UCP samples exhibited significantly more TRFLP
peaks matching Clostridium and Eubacterium genera compared to HUC and FAP pouches
and fewer peaks matching Lactobacillus and Streptococcus genera compared to FAP. DNA
Sanger sequencing of a subset of luminal samples revealed UCP having more identifiable
sequences of Firmicutes (51.2% versus 21.2%) and Verrucomicrobia (20,2% versus 3,2%), and
fewer Bacteroidetes (17,9% versus 60.5%) and Proteobacteria (9,8% versus 14,7%) compared to
FAP.

Page 2

Conclusions: The pouch microbial environment appears to be distinctly different in the settings
of UC pouchitis, healthy UC, and FAP. These findings suggest that a dysbiosis may exist in
pouchitis which may be central to understanding the disease.
Abdominal colectomy with ileal pouch anal anastomosis .(IPAA) is the preferred surgery for
patients with ulcerative colitis (UC) unresponsive to medical therapy or with dysplasia and for
those with familial adenomatous polyposis syndrome (FAP),I Inflammation of the ileal pouch,
termed pouchitis, is the most common long-term complication in patients with UC who undergo
IPAA.2 Pouchitis occurs in up to 50% of these patients3'4 and is the major factor leading to
morbidity after IPAA surgery in patients with UC.3't In comparison to individuals with a history of
UC, those undergoing IPAA for FAP are far less likely to develop pouchitis.5
The pathogenesis of pouchitis is incompletely understood. Proposed etiologies include fecal
stasis, bacterial overgrowth, dysbiosis, genetic susceptibility, short-chain fatty acid deficiency,
ischemic complications of surgery, bile-acid toxicity, immune alteration, a missed diagnosis of
Crohn's disease (CD), or a recurrence of UC in the pouch.1.6,1 Pouch bacteria appear central to the
pathophysiology of the disease. The efficacy of antibiotics in ameliorating pouchitis underscores the
role of mucosal or luminal bacteria in the development of pouchitis." The role of gut microbes in
pouchitis is further supported by studies revealing fewer pouchitis relapses and postoperative
prophylactic benefit with probiotic therapy.1° A therapeutic utility of both antibiotics and
probiotics suggests there may be subtle shifts in pouch microflora that culminate in pouch
inflammation.
Recently, Komanduri et all' described a pouch dysbiosis in five patients with UC and pouchitis
when compared to patients with UC and no pouchitis and to healthy ileal tissue in continuity with
the colon. Inflamed pouch mucosa was found to have greater bacterial species diversity than pouch

Page 3

control or healthy ileal tissue. Fusobacterium varium was more abundant in pouchitis tissue,
while Streptococcus species were dramatically reduced. Healthy UC pouches possessed a unique
floral pattern, with more representatives from the Clostridium, Enteric, and Streptococcus groups
compared to ileal tissue from healthy subjects.
In our study we aimed to characterize both mucosal and luminal microflora in inflamed UC
pouches compared to healthy UC and FAP pouches. We designed the study with FAP pouch
control subjects in an effort to find shifts in flora that were distinct to UC and inflammation
rather than fecal stasis. We utilized 16S ribosomal DNA gene-based terminal restriction fragment
length polymorphism (TRFLP) techniques to characterize luminal and mucosal microbial
environments, followed by DNA sequencing for further characterization and validation.
MATERIALS AND METHODS
Patient Population
Individuals over age 18 years were enrolled from the Massachusetts General Hospital (MGH)
Gastrointestinal Unit in one of three cohorts with pouches after IPAA: those with a history of
UC and subsequent pouchitis (UC pouchitis, UCP), those with a history of UC without pouchitis
(healthy UC pouch, HUC), or those with a history of FAP without pouchitis (healthy FAP pouch,
FAP). The diagnosis of UC and FAP in these patients was based on standard clinical, radiologic,
and histologic diagnostic criteria. UCP patients were eligible if they had experienced at least two
episodes of antibiotic-responsive pouchitis symptoms in the previous 2 years or had persistent
pouchitis symptoms despite two courses of antimicrobial therapy. A Pouch Disease Activity Index
(PDAI, comprising clinical, endoscopic, and histologic elements) score >7 was required for
inclusion in the UCP group.I2 Eligible UCP patients underwent colectomy with IPAA more than
3 months prior to enrollment. Eligible HUC and FAP patients underwent colectomy with IPAA more

Page 4

than 1 year prior to enrollment without experiencing pouchitis symptoms. Exclusion criteria for each
group included any indication of CD and the use of probiotics, biologic or other immunosuppressive
therapy, rectal therapy, or new antibiotics during the previous 2 months. Chronic antibiotic use for
more than 3 months was permitted in the UCP group if the patient's symptoms persisted without
the use of new antibiotics.
Data Collection
Key clinical data collected from each patient included age, gender, disease duration and
extent, pouch duration, number of pouchitis episodes in the previous year, medication use, and
the clinical components of the PDAI. The PDAI clinical score takes into account stooling
frequency and the presence of hematochezia, stooling urgency, abdominal cramping, and fever.
Endoscopic and histologic findings were noted for each patient and incorporated into the PDAI
score.
Specimen Collection
All patients underwent pouchoscopy after clinical data were collected. Patients received a
short course of bowel preparation 1 day prior to the procedure with phospho-soda. Luminal pouch
contents were suctioned through the colonoscope into a sterile container. Approximately I cc of
the suctioned stool was aliquoted into each of three vials containing 1 cc of 10% glycero1/90%
sterile water. Tubes were inverted to generate a homogeneous mixture then immediately placed
in dry ice. The pouch was examined for elements of the PDAI endoscopic score: edema,
granularity, friability, loss of vascular pattern, mucous exudates, and ulceration. Then three
biopsies were taken at the most inflamed sites or at random locations from normal-appearing
tissue. Each tissue specimen was immediately rinsed in normal saline then placed in a vial
containing 1 cc of 10% glycerol/90% sterile water. The vials were inverted to submerge the tissue

Page 5

specimen then immediately placed in dry ice. All specimens were then stored at —80°C within
30 minutes.
TRFLP Analysis
All stool and tissue specimens were analyzed by TRFLP at the Environmental
Biotechnology Institute at California Polytechnic State University (CPSU). DNA was isolated
using MoBio's Power Soil DNA extraction kit according to the manufacturer's protocol (MoBio
Laboratories, Carlsbad, CA). Triplicate 0.1-g samples were used for the extraction of stool
samples, while the entire biopsy sample was used for the isolation of DNA from tissue samples.
Bacterial 16S ribosomal RNA genes were amplified from both types of sample using the
forward primer 8DF (5'-AGA GTT TGT TCM TGG CTC AG-3') and the reverse primer 536-
K2R (5'-GTA TTA CCG CGG CTG CTG G-3'). The forward primer was fluorescently labeled
with a phosphamide dye. Fifty microliter reactions were assembled using 1 FL of undiluted
DNA, 5 pl., of 10x buffer, 3 pl. of 10 mM dNTPs, 2 tL of 20 mg/mL BSA, 7 pL of 25 mM
MgCl2, 1 tL of each primer, and 0.3 ,td, of 5 U/L TaqGold (Applied Biosystems, Foster City,
CA).
Reaction temperatures and times were as follows: 95°C for 10 minutes followed by 30 cycles of
95°C for 1 minute, 60°C for 1 minute, and 72°C for 2 minutes, followed by a final extension of
72°C for 10 minutes. The quality of the extractions and polymerase chain reaction (PCR) reactions
were confirmed using gel electrophoresis. PCR triplicates were combined during a PCR cleanup
performed using MoBio's PCR Cleanup Kit following the manufacturer's protocol (MoBio
Laboratories). PCR products were quantified using a FLX800 microplate fluorescence reader
tuned to the labeling dye (Bio-Tek Instruments, Winooski, VT).

Page 6

Three enzyme digests were independently performed using cleaned-up PCR product and the
restriction endonucleases HaeIH, Hpall, and Alul (New England Biolabs, Beverly, MA). Each
40-1,tL digest included 75 ng DNA, 1 U enzyme, and 4 pi, buffer. Samples were digested for
4 hours at 37°C and the enzyme was inactivated for 20 minutes at either 65°C (HpaII
and Alul) or 80°C (Haell1). Resulting fragments were ethanol precipitated and resuspended in 20
[IL formamide and 0.25 tL CEQ 600 basepair standard. Terminal restriction fragments were
separated using capillary gel electrophoresis and profiles were obtained using a Beckman
Coulter (Fullerton, CA) CEQ8000X DNA analysis system.
Terminal restriction fragment lengths and relative peak areas were exported from the
CEQ8000 into Excel (Microsoft, Seattle, WA). TRF data consists of peak sizes, which indicate the
length of the DNA fragment in nucleotides, and the area under each peak, which provide a mea-
sure of abundance for each DNA fragment. Since the amount of DNA loaded on the capillary
cannot be accurately controlled, the sum of the total peak area varies between TRF patterns.
Therefore, peak areas are normalized by converting the value to parts per million to standardize
the data for comparison." Peaks with an area less than 10,000 ppm (<1.0% of the total) were
excluded from the analysis to reduce excess noise. Data from three independent restriction enzyme
digests were included to reduce the incidence of distinct sequences with equivalent TRF lengths.
TRF fragments determined to differ most between sample groups were compared to available
GenBank (Bethesda, MD) sequences to tentatively identify potential bacterial populations. TRFLP
data were analyzed for peaks that could be identified by database analysis as representing particular
groups of organisms.
16S Ribosomal DNA-based DNA Sequencing
Four fecal samples from each of the UCP group and the FAP group were selected for
sequencing analysis of pooled DNA. Samples with the greatest intersample similarities in TRF

Page 7

profiles (as determined with Bray—Curtis similarity indices) were selected in an effort to create
the most uniform pooled specimens representative of each patient group. DNA from the
samples was processed as detailed for TRF analysis, using full-length 16S PCR primers
without fluorescent label, then sent to the Broad Institute (Cambridge, MA) for sequencing.
Samples were cloned into the pCR2.1-TOPO vector and sequenced on an ABI3730 DNA sequencer.
Sequences were trimmed using LUCY.14 Read pairs from each clone were then assembled
using the following alignment-assisted assembly method implemented at the Broad Institute. A
reference sequence for use as an alignment template for the read pairs was first selected from a
core set of nonchimeric 16S rRNA sequences obtained from Greengenes.15 The core set
reference sequence sharing the greatest number of matching k-mers with the aggregate k-mer set
of both forward and reverse reads was selected. Forward and reverse reads were each aligned to
the core reference sequence using Blast.I6 Aligned forward and reverse reads were then assembled
based on the alignment with base quality scores used and preserved through assembly.
For purposes of classification, only assembled sequences of greater than 1100 nucleotides
were considered. Sequences were classified using a naïve Bayesian rRNA classifier." The
classifier was trained on the Ribosomal Database Project core set, RDP10.I8
Statistics
Differences in patient characteristics were assessed using Fisher's Exact test for categorical
variables and the Exact Kruskal—Wallis test for continuous variables. TRFLP datasets were
transformed by taking the square root of the area under each peak to deemphasize large TRFLP
peaks while still accounting for relative abundance. Transformed data from UCP, HUC, and FAP
groups were compared using Bray—Curtis similarity, multidimensional scaling, and analysis of
similarity (ANOSIM) (Primer E, Plymouth, UK). ANOSIM was also used to compare

Page 8

intersubject and intrasubject TRFLP profile similarities. Analysis of variance (ANOVA) was
used to compare relative abundance of specific TRFLP pattern elements between each group of
subjects. Among the pooled fecal DNA samples, differences in the percentage of total identifiable
clones represented by each bacterial genus were calculated using Fisher's Exact test,
considering that the number of identifiable clones in a sample is proportional to species
abundance.
Ethical Considerations
This study was approved by the Institutional Review Boards at MGH and CPSU.
[Insert Table 1]

RESULTS
Patients
A total of 19 patients were enrolled and divided into three groups: UCP (n = 9), FAP (n
=. 7), and HUC (n =- 3) (Table 1). Overall, 37% of patients were female and the average age was
40.6 years. There were no significant differences in baseline demographics or disease characteristics
between the three groups. PDAI values were significantly higher in the UCP group (P < 0.001).
All mucosal biopsies in the UCP group displayed histologic evidence of active pouchitis. One UCP
patient received chronic ciprofloxacin therapy for the prior 8 years and one received chronic
rifaximin therapy for the prior year, both with continued episodes of pouchitis. Three UCP
patients had received short courses of ciprofloxacin ranging from 3 months to 1 year prior to
enrollment. The remaining four UCP patients received ciprofloxacin briefly —2 years prior to
enrollment.

Page 9

TRFLP Analysis
Significantly more peaks were observed in TRFLP data from mucosal samples (average total
number of peaks/ sample in three digests = 64.8, standard deviation [SD] = 12.2) compared to fecal
samples (average total number of peaks/sample in three digests = 41.3, SD = 8.2) with P < 0.001
(Fig. 1). The greatest amount of similarity existed within samples from the same person, with
both mucosal and fecal sample similarity greater within a subject than between subjects (P =
0.027). There was no significant difference between patient groups in the degree of similarity
between fecal and mucosal samples (P > 0.05). Between individuals, mucosal TRFLP profiles
were found to be more consistent than fecal samples (P < 0.001).
TRFLP data from mucosal and fecal samples were significantly different between each of
the three patient groups. Figure 2 displays the results of multidimensional scaling analysis
comparing TRFLP data. Significant differences (P < 0.05) in bacterial TRFLP data were
revealed between the UCP group, HUC group, and FAP group when comparing mucosal samples,
fecal samples, and when both sample types were combined. Samples from the two UCP patients
receiving a single long-term antibiotic were not significantly different compared to other UCP
samples. Samples from the three patients who received prior short courses of antibiotics within 1
year were also not significantly different, as all nine UCP samples grouped together apart from
HUC and FAP.
Approximately 15% of the total number of TRFLP peaks were matched to peak sets in the
database. In an analysis of TRFLP elements that represent bacterial genera, a set of TRFs
matching Lactobacillus and Streptococcus (Hae 264-5, Hpa 97-9, Alu 76, 532) were present at
higher relative abundance in both mucosal and fecal samples from FAP patients compared to UCP
(ratio 5:1 in mucosa, 3:1 in stool). A second set of TRFLP peaks (Hae 272-4, Hpa 222-3, Alu
440) matching Clostridium, Eubacteriurn and Roseburia genera were present at five times the

Page 10

relative abundance in stool from UCP patients compared to FAP patients. Fecal samples from
HUC pouches also had fewer peaks matching Clostridium, Eubacterium, and Roseburia
compared to UCP (ratio 1:15). HUC stool revealed fewer Escherichia, Streptococcus, and
various sulfur-oxidizing bacteria (Hpa 496, Hae 205, Alu 74) at a ratio of 1:2 compared to UCP.
Mucosal samples from HUC pouches revealed a similar smaller quantity of Escherichia,
Streptococcus, and various sulfur-oxidizing bacteria compared to UCP (ratio 1:2).
[Insert Figure 1 and Figure 2]
DNA Sequencing
In all, 2304 clones were processed for each pooled sample (4608 total clones). After data
processing and analysis, good quality forward and reverse reads were identified with a reliable
classification at the genus level (>80%). A total of 712 sequences were identified in the UCP
pooled sample and 1015 in the FAP pooled sample.
At the phylum level, the UCP pooled sample revealed significantly more Firmicutes
and Verrucomicrobia (52% and 22% of clones, respectively) compared to the FAP group
comprised of 19% Firmicutes and 3% Verrucomicrobia (P < 0.001 and P < 0.001, respectively).
However, numbers of clones in the Bacteroidetes phylum were significantly higher in the FAP group
(71% of clones) compared to UCP (20% of clones, P < 0.001). There were no significant
differences between pooled samples among the Fusobacteria, Proteobacteria, and Actinobacteria
phyla (Table 2). [Insert Table 2]
Multiple bacterial genera within the class Clostridia were significantly more prominent in the
UCP group compared to the FAP group (Table 3). Among Clostridia, the UCP group had more
clones represented by the genera Roseburia (8% of clones in UCP versus near 0% in FAP, P <
0.001), Lachnospiraceae Incertae Sedis (28% versus 10%, P < 0.001), Clostridium (9% versus
1%, P < 0.001), Lachnospira (2% versus 0, P < 0.001), and Veillonella (3% versus near 0%,

Page 11

P < 0.001). The UCP pooled sample also had more clones from the genus Prevotella than the
FAP group (6% versus 1%, P < 0.001). Finally, the UCP group revealed 22% of its pooled clones
from the genus Akkermansia in the phylum Verrucomicrobia, compared to only 3% in FAP (P <
0.001).
Members of the genus Bacteroides predominated in the pooled fecal sample of the FAP
group, with 678 of 1015 (67%) clones representing Bacteroides, compared to 99 of 712 (14%)
in UCP (P < 0.001). Other genera that were statistically more prominent in the FAP group were
Faecalibacterium in the family Ruminococcaceae (6% of clones in FAP, 0% in UCP, P <
0.001), Parabacteroides (2% in FAP, 0% in UCP, P < 0.001), and Escherichia (2% in FAP,
0% in UCP, P < 0.001).
DISCUSSION
The results of our study suggest that the pouch microbial environment is distinctly different
between patients with UC-associated pouchitis and healthy UC or FAP pouches. Using 16S ribosomal
gene-based TRFLP data, we identified significant differences in bacterial communities
between all three patient cohorts in both stool and mucosa. These broad differences in TRFLP
profiles were further explored using DNA sequencing. Sequencing revealed multiple statistically
significant variations in specific bacterial genera between pooled fecal DNA from a subgroup of
patients with UCP and a subgroup of patients with FAP.
We report data that supports previously published work suggesting a dysbiosis is central to
the development of pouchitis. The efficacy of both antibiotics and probiotics in ameliorating the
symptoms of pouchitis underscores the theory that shifts in bacteria may be important in the gener-
ation and cessation of pouch inflammation."" However, few details are known about the precise
bacterial populations in the pouch and the nature of the dysbiosis associated with pouchitis.

Page 12

Studies based on classic culturing techniques have suggested the presence of more anaerobes in
pouch effluent compared to ileostomy effluent and that pouch fecal bacterial populations are
similar to rectal flora.2022 A later study using similar stool culture techniques noted a greater
number of facultative anaerobes in the inflamed pouch lumen compared to the noninflamed UC
pouch or normal ileum.23 More recently, Duffy et a124 used culture-based techniques to show that
sulfate-reducing bacteria were present in the stool of 8 of 10 healthy UC pouches but not in any
of seven FAP pouches.
Molecular techniques have been used to define the pouch microbiota more precisely. Falk et
a125 used TRFLP to demonstrate that pouch mucosal species in two healthy UC pouches
resembled those from the colon. However, populations of Clostridium perfringens in both
patients were unique to the pouch. To our knowledge, only one study, by Komanduri et al,E I used
molecular techniques to compare pouch microbiota between UC patients with and without
pouchitis. Using length-heterogeneity PCR followed by sequencing, they concluded that
inflamed pouch mucosa had greater species diversity and more Fusobacterium varium than
healthy UC pouches or healthy deal tissue in continuity with the colon. Noninflamed UC pouches
revealed more representatives from the Clostridium, enteric, and Streptococcus groups compared
to pouchitis and healthy nonpouch Heal tissue. These differences between UC and health could
be explained by the fact that the healthy ileal tissue was in continuity with the colon without
fecal stasis. In each of these studies the healthy pouch control subjects suffered from UC, without
any non-inflammatory bowel disease (IBD) pouch controls.
Our study was designed to utilize molecular techniques to identify bacterial populations
unique to inflamed or healthy UC pouches compared to a non-IBD pouch, keeping pouch fecal
stasis as a consistent factor across all patient groups. This design and our findings differed from
the results of Komanduri et al.1' We found no increase in Fusobacterium species among inflamed

Page 13

pouches and instead noted the largest numbers of Clostridium species in the UCP group. We
also noted a unique reduction in Bacteroides in the inflamed pouch compared to FAP. Our
findings suggest that a mucosal and luminal dysbiosis exists in pouchitis, not only when
compared to the healthy UC pouch but also when compared to a non-IBD pouch. Additionally,
healthy UC pouches differed significantly from FAP pouches. This suggests an alteration in
ileal pouch microbiota that may be unique to the UC disease state, with or without inflammation.
Our TRFLP results supported an overall increase in fecal Clostridium in UCP patients
compared to FAP patients. This finding may be consistent with those of Falk et al.25 DNA
sequencing confirmed our findings, with the genus Clostridium representing 9% of identifiable
clones in the UCP pooled sample compared to only 1% in the pooled FAP sample. More broadly,
the class Clostridia accounted for 53% of identifiable clones in the UCP group compared to 21%
in the FAP group. The genus Eubacterium was also more prevalent in the UCP stool by TRFLP, but
no known Eubacterium species were revealed by sequencing. This may be due to the
difficulty of TRFLP to distinguish between Eubacterium and Clostridium species. The
decrease in Lactobacillus and Streptococcus genera in UCP compared to FAP pouches also was
not reflected in the sequencing data. However, this is not surprising since the number of clones
identified as representing Lactobacillus and Streptococcus was less than 0.5% of the total
identified. These disparities between TRFLP and sequencing results could be explained by
our sampling only four patients from each group for sequencing. Importantly, TRFLP is best
used to compare bacterial community structure between samples and is less precise than
sequencing analysis for bacterial speciation.26
Sequencing results revealed an overall decrease of the phylum Bacteroidetes in the
inflamed pouch. This finding is consistent with multiple molecular surveys of intestinal microbiota
in both CD and UC.27-3° Bacteroidetes play a key role in maintaining gut health, so it has been

Page 14

proposed that a relative reduction in this population may favor the development of inflammation.29
Consistent with our TRFLP data, DNA sequencing also demonstrated a significant increase in
clostridia in the inflamed pouch, namely among the genera Clostridium, Lachnospiraceae, and
Rosehuria. While multiple studies have shown a relative decrease in clostridia species among
patients with CD,30-32 our findings are consistent with previous work in UC revealing more
clostridia in inflamed UC rectal mucosa compared to noninflamed mucosa.33
While evidence to date supports a general shift in multiple species as being linked to the
development of IBD, we note two particular genera that reflect specific theories of IBD
pathogenesis. We found significantly greater numbers of the genus Roseburia among inflamed
pouches compared to FAP pouches. Roseburia are flagellated commensal inhabitants of the colon.
Flagellin have been shown to induce proinflammatory gene expression by activating Toll-like
receptor 5 (TLR5), and patients with TLR5 polymorphisms and low levels of antiflagellin
antibodies may be protected from developing CD.34-36 High levels of antiflagellin antibodies,
specifically anti-CBirl antibodies to flagellin of Clostridium species, are present in 50% of CD
patients, 6% of UC patients, and may be associated with the development of pouchitis.37'38 It is
unknown whether Roseburia species may play a role in TLR5 activation leading to increased
mucosal inflammation. Additionally, we found the mucin-degrading genus Akkermansia of the
phylum Verrucomicrobia to be significantly more prevalent in pouchitis. Several theories exist
regarding the role of mucin in the protection of intestinal epithelium in IBD, with mutations,
alterations, and degradation of mucins being associated with CD and UC.39-42 A recent large-scale
genome-wide association study revealed that a gene encoding for mucin proteins, MUC19, was
mutated significantly more often in patients with CD.43 While the mucin-degrading
Verrucomicrobia species and Akkermansia in particular have been identified within the human

Page 15

colon,44'45 there have been only two reports of DNA from these bacteria in patients with CD" and no
previous reports of Verrucomicrobia in UC.
This study has certain limitations. TRFLP is best used to compare bacterial community
structure between cohorts and is not precise in identifying species. Thus, the results of TRFLP
should be interpreted as suggestive of bacterial groups, rather than specific species, and emphasis
should be placed on the separation between each group. The study may have found more
differences between patient groups if the sample size had been larger. Nevertheless, we were able
to identify statistically significant differences in TRFLP data and clone libraries between groups.
Another limitation was the pooling of DNA into groups for DNA sequencing. Individual
sequencing of each sample was limited due to expense. This may have led to an apparent
difference in flora between groups if only one sample had a unique bacterial profile. However, we
grouped samples with similar TRFLP data in an effort to minimize the chance of one sample
skewing the analysis. Additionally, the use of antibiotics in UCP patients may have altered gut
flora. However, TRFLP profiles were similar in all UCP patients and distinct compared to HUC
and FAP, regardless of whether the subjects had received antibiotics chronically or briefly 1-2
years before enrollment. This suggests that the dysbiosis may be unique to pouchitis and not a
direct result of exposure to antibiotic medications.
In conclusion, we identified differences in the pouch microbiome between the inflamed pouch in
UC, the healthy pouch in UC, and the healthy pouch in FAP. These differences in bacterial
populations were evident in stool and mucosa by TRFLP analysis. 16S rDNA sequencing
revealed specific differences in luminal bacterial genera, with the pouchitis group having
substantially fewer Bacteroidetes and more Clostridia compared to the healthy FAP group. These
findings reinforce other surveys of the micro-biome in both UC and CD. Additionally, one genus
with flagellated species and one genus with mucin-degrading species were more prevalent in the