Faecal microbiota profile of Crohn’s disease determined by terminal restriction fragment length polymorphism analysis


Dr A. Andoh, Department of Internal Medicine, Shiga University of Medical Science, Seta-Tsukinowa, Otsu 520-2191, Japan.
E-mail: andoh@belle.shiga-med.ac.jp


Background  Terminal restriction fragment length polymorphism (T-RFLP) analyses are powerful tools to assess the diversity of complex microbiota. T-RFLPs permit rapid comparisons of microbiota from many samples.

Aim  To perform T-RFLP analyses of faecal microbiota in Crohn’s disease (CD) patients to investigate potential alterations in faecal microbial communities and furthermore to analyse the effects of elemental diet on faecal microbiota profiles.

Methods  Thirty-four patients with CD and 30 healthy individuals were enrolled in the study. DNA was extracted from stool samples and 16S rRNA genes were amplified by PCR. PCR products were digested with BslI restriction enzymes and T-RF lengths were determined.

Results  Faecal microbial communities were classified into seven clusters. Almost all healthy individuals (28/30) were included in cluster I, II and III, but the majority of CD patients (25/34) could be divided into another four clusters (cluster IV–VII). Prediction of bacteria based on the BslI-digested T-RFLP database showed a significant decrease in Clostridium cluster IV, Clostridium cluster XI and subcluster XIVa in CD patients. In contrast, Bacteroides significantly increased in CD patients. Significant increases in Enterobacteriales were also observed in CD patients. Furthermore, elemental diets modulated faecal bacterial communities in CD patients.

Conclusions  Terminal restriction fragment length polymorphism analyses showed that the diversity of faecal microbiota in patients with CD differed from that of healthy individuals. Furthermore, elemental diets modulated faecal microbiota composition, and this effect may be involved in mechanisms of clinical effects of elemental diet.


Inflammatory bowel disease (IBD), ulcerative colitis (UC) and Crohn’s disease (CD) are chronic intestinal disorders of an unknown aetiology. It is generally accepted that commensal enteric bacteria provide constant antigenic stimulation that activates pathogenic T cells to cause chronic intestinal injury in genetically susceptible individuals.1–4 Recent lines of evidence from a variety of disciplines, including genetics, molecular microbiology, immunology, translational research and clinical trials have combined to implicate abnormal host–microbial interactions in the pathogenesis of IBD.5–7 For example, intestinal lesions of IBD are predominant in distal parts of the gastrointestinal tract where commensal bacteria are most abundant. The presence of intestinal bacteria is essential for the development of experimental colitis in several animal models.8–10 In CD, faecal stream diversions reduce gut inflammation and induce mucosal healing in excluded intestinal segments, whereas infusions of intestinal contents rapidly provoke disease flare-ups.11

Most commensal enteric bacteria are hard to cultivate even using recent culture techniques; cultivable bacteria are limited to 20–30% of the total because of the strict anaerobic and complex environment.12, 13 However, recent molecular techniques in microbial composition of faecal samples using 16S ribosomal DNA and RNA have increased previous culture-based estimates of 200–300 colonic species to as high as 1800 geneta and between 15 000 and 36 000 individual species.4 These bacteria increase in both concentration and complexity from the proximal gastric and duodenal population of 102–103 aerobic organisms/g luminal contents to 1011–1012 predominantly anaerobic bacteria/g in the ceacum and colon.

Terminal restriction fragment length polymorphism (T-RFLP) is an alternative molecular approach that allows the assessment of the diversity of a bacterial community structure and diversity of different ecosystems.14–16 Although T-RFLP analyses do not provide high resolution data like a 16S rRNA library, the distribution of micro-organisms including uncultivables can be divided into clusters or groups. T-RFLP analyses are potentially useful for assessing the diversity of faecal microbiota and rapid comparisons of community structures among individuals.14, 15, 17 In addition, bacterial species of major T-RFs can be estimated by means of computer simulations with T-RFLP analyses programmes such as TAP T-RFLP by Marsh et al.18 (http://www.cme.msu.edu/RDP/html/analyses.html). Recently, we reported that faecal microbial communities of UC patients are different from those of healthy volunteers by using T-RFLP analyses.19

Dietary therapies such as an elemental diet (ED) play an important role in the treatment of CD both to induce and to maintain remission.20, 21 However, the precise mechanism underlying its clinical effects, such as modulation of gut microbiota, remains unclear.

In this study, we performed T-RFLP analyses of faecal microbiota in CD patients to investigate potential alterations in faecal microbial communities. Furthermore, we analysed the effects of ED on faecal microbiota profiles.

Materials and methods


Thirty-four patients with CD (14 females and 20 males, median age 29 years) and 30 healthy individuals (15 females and 15 males, median age 34 years) were enrolled in this study (Table 1). CD patients consisted of 22 ileocolic type, eight colic type and four ileal type. Seventeen patients were active phase [Crohn’s disease activity Index (CDAI reported by Best et al.22) >150] and 17 patients were inactive phase (CDAI <150). Twenty-one patients with CD were placed on an ED with Elental (Ajinomoto pharmaceutical Ltd., Tokyo, Japan; more than 1200 kcal/day for over 6 months) as their primary therapeutic treatment,23 and these patients were regarded as the ED group. All patients were managed at the Division of Gastroenterology of the Hospital of the Shiga University of Medical Science. The baseline characteristics from the healthy controls were matched with CD patients. All patients were not treated with antibiotics and/or probiotics, which might affect their faecal microbial structures. The ethics committee of Shiga University of Medical Science approved this project.

Table 1.   Baseline clinical characteristics
 Control (n = 30)CD (n = 34)
  1. Each data indicates mean ± s.d.

  2. ED, elemental diet; CD, Crohn’s disease; CDAI, Crohn’s Disease Activity Index reported by Best et al.

Gender (M/F)  15/15  20/14
Age (years)34.4 ± 5.529.4 ± 7.5
Disease duration (years)5.3 ± 3.3
5-ASA treatment34
Corticosteroid use10
Purine analogues use18
Type of CD
ED over 6 months21
CDAI >15017

DNA extraction

Faecal sample was suspended in 4 m guanidium thiocyanate, 100 mm Tris–HCl (pH 9.0) and 40 mm EDTA, then beaten in the presence of zirconia beads using a FastPrep FP100A Instrument (MP Biomedicals, Irvine, CA, USA). Thereafter, DNA was extracted from the bead-treated suspension using Magtration System 12GC and GC series Magtration–MagaZorb DNA Common Kit 200 N (Precision System Science, Chiba, Japan), and final concentration of DNA sample was adjusted to 10 ng/μL.

PCR amplification and T-RFLP analysis

The 16S rRNA gene was amplified from human faecal DNA by using the fluorescently labelled 516f primer (5′-TGCCAGCAGCCGCGGTA-3′; Escherichia coli positions 516–532) and 1510r primer (5′-GGTTACCTTGTTACGACTT-3′; E. coli positions 1510–1492).24 The 5′-ends of the forward primers were labelled with 6′-carboxyfluorescein (6-FAM), which was synthesized by Applied Biosystems (Tokyo, Japan). The PCR amplifications of DNA samples (10 ng of each DNA) were performed according to a protocol described by Nagashima et al.24 The amplified 16S rDNA genes were purified using GFX PCR DNA and Gel Band Purification Kit (GE Healthcare Bio-Sciences, Tokyo, Japan) and redissolved in 30 μL distilled water.

The restriction enzymes were selected according to Nagashima et al.24 The purified PCR products (2 μL) were digested with 10 U of BslI at 55 °C for 3 h. The length of the T-RF fragment was determined using an ABI PRISM 3100 genetic analyser (Applied Biosystems) in GeneScan mode. Standard size markers, GS 2500 ROX (Applied Biosystems), were used. The fragment sizes were estimated, using the local southern method in genescan 3.1 software (Applied Biosystems).The T-RF fragments were divided into 30 operational taxonomic units (OTUs) according to the methods described by Nagashima et al.24 The OTUs were quantified as the percentage values of individual OTU per total OTU areas, and this was expressed as the % area of the under-peak curve (%AUC). Cluster analyses were performed using the software jmp (SAS Institute, Cary, NC, USA) based on the BslI T-RFLP patterns and a dendrogram was constructed using a minimal variance algorithm (Ward’s method).

Statistical analysis

The relative abundances of specific bacterial groups reflected through their T-RF peak areas and their percentage values were compared between the control and CD patients using the Aspin–Welch t-test.


Based on analyses of faecal bacterial communities by BslI-digested T-RF patterns in CD and healthy individuals, faecal microbial diversities were compared by a dendrogram constructed by using a minimal variance algorithm (Figure 1). A setting of similarity generated seven major clusters. Almost all healthy individuals (21/30) were included in cluster I. Seven of the residual nine healthy individuals were involved in cluster II and III. Thus, 28 of the 30 healthy individuals were involved in either of the clusters I, II and III, but only nine of the 34 CD patients were included in these clusters. On the other hand, a majority of CD patients (25/34) could be divided into four additional clusters (cluster IV–VII). There were no specific clusters for disease activity, medical treatment and clinical types. Only two healthy individuals were involved in these clusters, indicating that faecal bacterial communities differed between healthy individuals and CD patients.

Figure 1.

 Dendrogram of the faecal bacterial structure in Crohn’s disease patients and healthy individuals. T-RFLP patterns by BslI digestions were analysed using the software jmp (SAS Institute, Cary, NC, USA), and a minimal variance algorithm (Ward’s method) was used to construct a dendrogram.

The %AUC (percentage values of individual OTU area per total OTU area) of representative OTUs after BslI digestion is shown in Table 2. The %AUC is associated with the predominance of bacterial species which compose each peak. Predicted bacteria based on the BslI-digested T-RFLP database are also presented.24 Some OTUs (110, 168, 338, 494, 749, 754, 919 and 955 bp) predicting the Clostridium family decreased in CD samples compared with those from healthy individuals. In particular, a significant decrease in OTUs of 168 bp (Clostridium cluster IV), 494 bp (Clostridium subcluster XIVa), 919 bp (Clostridium cluster XI, subcluster XIVa) and 955 bp (Clostridium subcluster XIVa) was observed in CD patients. In contrast, 469- and 853-bp OTUs predicting Bacteroides significantly increased in CD patients. A significant increase in the 940-bp OUT predicting Enterobacteriales was also observed in CD patients. There were no significant differences of Bifidobacterium predicted by 124-bp OUT and Lactobacillales predicted by 332- and 657-bp OTUs between CD patients and healthy individuals.

Table 2.   Terminal restriction fragment length polymorphism profiling of faecal samples from healthy individuals and Crohn’s disease
OTUBacteria predicted by T-RF lengthControl (n = 30) (%)CD (n = 34) (%)
  1. Data were expressed as mean ± s.d. of %AUC (area under the peak curve).

  2. %AUC, per cent of total area; OTU, operational taxonomic unit. OTUs with significant changes are underlined.

  3. P < 0.05.

  4. ** P < 0.01.

110Clostridium cluster IX10.2 ± 6.04.2 ± 4.0
124Bifidobacterium6.5 ± 6.47.6 ± 7.2
168Clostridium cluster IV2.8 ± 3.20.7 ± 3.2*
317Prevotella1.4 ± 3.31.6 ± 4.9
332Lactobacillales1.6 ± 2.12.4 ± 4.6
338Clostridium cluster Xl1.5 ± 2.92.6 ± 6.4
370Bacteroides, Clostridium cluster IV5.4 ± 3.94.4 ± 5.5
469Bacteroides12.2 ± 6.621.6 ± 14.5**
494Clostridium subcluster XIVa8.2 ± 3.05.5 ± 5.6*
657Lactobacillales0.2 ± 0.94.3 ± 11.3
749Clostridium cluster IV4.8 ± 5.02.8 ± 5.5
754Clostridium subcluster XIVa2.1 ± 4.60.4 ± 1.3
853Bacteroides0.3 ± 0.61.2 ± 1.5*
919Clostridium cluster Xl, subcluster XIVa12.0 ± 8.14.9 ± 4.4**
940Enterobacteriales0.8 ± 1.22.6 ± 2.9*
955Clostridium subcluster XIVa10.2 ± 5.75.2 ± 5.0**
Others 22.5 ± 10.025.3 ± 13.7

To investigate the effects of ED on faecal bacterial communities, faecal microbial structures were compared between CD patients with ED (= 21) and those with free diets (= 13). As shown in Figure 2, CD patients were divided into seven major clusters. A majority of CD patients with a free diet (10/13) were involved in cluster I–III, whereas the majority of CD patients with an ED (15/21) were included in cluster IV–VII. These data suggest that ED affects faecal bacterial communities in CD patients.

Figure 2.

 Dendrogram of faecal bacterial structures in Crohn’s disease patients with an elemental diet (ED) or those with a free diet. T-RFLP patterns by BslI digestions were analysed using the software jmp (SAS Institute, Cary, NC, USA), and a minimal variance algorithm (Ward’s method) was used to construct a dendrogram.


Culture-based approaches have been previously used to analyse intestinal microbiota in IBD patients.25–27 However, most intestinal microbiota cannot be cultured because of anaerobic requirements and complex environments.12, 13 So, molecular approaches which enable the evaluation of alterations in uncultivable bacteria have been employed for analysing microbiota in IBD.4 For example, a phylogenetic approach based on 16S rDNA sequences has been used to investigate the diversity of faecal microbiota.12, 28–30 Manichanh et al.31 recently reported a metagenomic study to evaluate the microbial diversity of CD patients.

In this study, we used T-RFLP analyses to investigate faecal bacterial communities in CD patients. T-RFLP analyses provide rapid and reproducible comparisons of microbial communities and allow for the assessment of diversity in complex microbial communities.32 T-RFLP reflects differences in microbial composition patterns, which cannot be detected by culture-based analyses. Another benefit of T-RFLP is an estimation of bacterial species with major T-RFs by computer simulations based on 16S rRNA gene sequences registered into a database.33 Thus, T-RFLP analyses may be an ideal tool for investigating the diversity of bacterial communities in many samples. Recently, we reported results from T-RFLP analyses of faecal microbial communities in CD and healthy controls.19

Dendrogram analyses showed that CD patients clustered separately from healthy individuals, indicating that faecal bacterial communities differed between CD patients and healthy individuals. These differences are associated with a significant decrease in the Clostridium family such as Clostridium cluster IV and cluster XIV and a significant increase in Bacteroides in CD patients. Significant increases in Enterobacteriales in CD patients were also observed. Decreased Clostridium species in CD are a common feature of many studies,27, 30, 31, 34, 35 although the increased Bacteroides conflict with Bacteroides species being decreased in mucosal IBD specimens.29, 31, 35, 36 Increased Enterobacteriales is compatible with most previous studies demonstrating increased numbers of Enterobacteriaceae, including E. coli.27, 29, 36 Frank et al.36 analysed 190 resected tissue samples and showed decreased numbers of the phyla Firmicutes and Bacteroidetes with concomitant increases in Proteobacteria and Actinobacteria. Decreases in Firmicutes were largely because of decreases in Clostridium XIVa and IV groups. Metagenomic approach by Manichanh et al.31 demonstrated a reduced diversity of faecal microbiota in CD in association with decrease in Firmicutes. Similar observation of reduced diversity has been also reported by Ott et al.35 Thus, our observations of a decrease in the Clostridium family and an increase in Enterobacteriaceae by T-RFLP analyses are compatible with previous studies in CD patients, but increases in Bacteroides require further investigation.

Firmicutes and Bacteroides are predominant species of intestinal bacteria,4 and metabolize dietary fibre to short-chain fatty acids (SCFAs).31, 37 SCFAs account for up to 10% of the human energy source in the intestine,4 and facilitate epithelial growth.4 SCFAs, in particular butyrate, also function as a potent anti-inflammatory factor through inhibition of NF-κB activation.38, 39 Physiological concentration of butyrate blocks TNF-α- and IL-1β-induced IL-8 secretion from intestinal epithelial cells.38, 39 Previously, decrease in SCFAs in faecal extracts of IBD patients has been reported.40 These suggest that a decrease in Clostridium species may induce reduction in intestinal SCFAs, resulting in an inhibition in the mucosal repair process and an aggravation in intestinal inflammation.

Increased Enterobacteriaceae by T-RFLP analyses may be supported by a previous theory of functional alterations of commensal bacteria in IBD.4 Barnich and Darfeuille-Michaud described a role of adherent/invasive E. coli (AIEC), which were functionally transformed under inflammatory conditions. AIEC persist within epithelial cells and macrophages and selectively colonize the ileum of CD patients.41 AIEC were recovered from 65% of chronically inflamed ileal resections and 36% of mucosal biopsy specimens of the neoterminal ileum of patients with early postresection recurrent CD and 22% of endoscopically normal CD biopsy specimens in contrast to 3.7% of colonic biopsy specimens from the same patients and 6% of normal control ileal biopsy specimens. Combined with defective antimicrobial peptide function in CD,42 increases in Enterobacteriaceae species in our study may be explained by these functional alterations in bacteria, including E. coli.

Faecal microbial communities differed between CD patients with ED and those with free diets. The detailed mechanism by which ED maintains patients with CD in remission remains unknown. One possible explanation is the response related to the low content of lipids in the diet. Reduced antigenic load, low fat contents, the supply of trophic amino acids, intestinal permeability, faecal pH and bowel rest, among others, have been proposed as potential factors contributing to clinical effects of ED.43, 44 Although modulation of gut microbiota by ED has been suspected for long periods, there are no data regarding modulation. This is the first observation that ED can modulate faecal microbiota including uncultivable bacteria, suggesting that parts of the clinical effect of ED may be associated with changes in intestinal microbiota compositions. One may suppose that such results might be associated with the selection of patients for ED, but we did not apply ED according to patients’ clinical activity and clinical types. Application and continuance of ED for 6 months was dependent on only patients’ acceptance. In the future, changes in microbiota profiles in the same patient before and after starting ED should be investigated to get much more precise effects. Furthermore, to address a direct mechanism of ED to exert its clinical effects via modulation of microbiota compositions, functional studies of altered microbiota should be performed.

In conclusion, using T-RFLP analyses, we showed that faecal microbial communities in patients with CD differed from those in healthy individuals. Differences were characterized by decreased Clostridium species and increased Bacreroides and Enterobacteriaceae. However, no consensus has emerged on general alterations in microbiota from CD and whether these changes are primary or secondary events. Further studies on these matters are required. The identification of mechanisms and/or factors contributing to microbiota structures may contribute to the development of new therapeutic strategies for patients with CD.


The authors thank Dr Tsutomu Maeta and Takayoshi Hisada (FloraInformatics, LLP., Tokyo, Japan) for their technical assistance. Declaration of personal and funding interests: None.