Analysis of the large bowel microbiota of colitic mice using PCR/DGGE


R. Bibiloni, Department of Agricultural, Food and Nutritional Science, University of Alberta, Room 4-10 Ag/For Centre, Edmonton, AB, Canada T6G 2P5 (e-mail:


Aim:  To test combined polymerase chain reaction amplification of 16S rRNA gene sequences and denaturing gradient gel electrophoresis (PCR/DGGE) as an analytical method to investigate the composition of the large bowel microbiota of mice during the development of colitis.

Methods and Results:  The colonic microbiota of formerly germfree interleukin 10 (IL-10)-deficient mice that had been exposed to the faecal microbiota of specific pathogen-free animals was screened using PCR/DGGE. The composition of the large bowel microbiota of IL-10-deficient mice changed as colitis progressed. DNA fragments originating from four bacterial populations (‘Bacteroides sp.’, Bifidobacterium animalis, Clostridium cocleatum, enterococci) were more apparent in PCR/DGGE profiles of colitic mice relative to non-colitic animals, whereas two populations were less apparent (Eubacterium ventriosum, Acidophilus group lactobacilli). Specific DNA:RNA dot blot analysis showed that bifidobacterial ribosomal RNA (rRNA) abundance increased as colitis developed.

Conclusions:  PCR/DGGE was shown to be an effective method to demonstrate changes in the composition of the large bowel microbiota of mice in relation to progression of inflammatory disease. The intensity of staining of DNA fragments in DGGE profiles reflected increased abundance of bifidobacterial rRNA in the microbiota of colitic animals. As bifidobacterial fragments in PCR/DGGE profiles generated from microbiota DNA showed increased intensity of fragment staining, an increase in bifidobacterial numbers in colitic mice was indicated.

Significance and Impact of the Study:  PCR/DGGE analysis demonstrated an altered composition of the large bowel microbiota of colitic mice. This work will allow specific groups of bacteria to be targeted in future research concerning the pathogenesis of colitis.


Inflammatory bowel diseases of humans (Crohn's disease, ulcerative colitis) are of complex aetiology but involve an aggressive immune response directed by a dysfunctional immune system at microbial inhabitants of the gut (the gut microbiota) (Duchmann et al. 1995; Van Heel et al. 2001). Although the immunology of inflammatory bowel diseases has been studied intensively, relatively little is known of the associated microbiology (Sartor 2004). Studies of the gut microbiota of humans in health and disease are confounded by the diversity in composition of the bacterial community from one human to another, and access to bowel sites where disease occurs is limited (Stebbings et al. 2002). In contrast, samples from various bowel regions of experimental animals, and commercial reagents for concomitant immunological analysis, are readily obtained. For these reasons, rodent models of colitis have been used to investigate chronic intestinal inflammation (Elson et al. 1995).

Interleukin-10 (IL-10)-deficient mice, in which the gene encoding the cytokine has been deleted, provide one such model (Kuhn et al. 1993). IL-10-deficient mice colonized with a conventional microbiota develop a lethal enterocolitis but the disease is attenuated and confined to the large bowel when the animals are colonized with gut microbiota from specific pathogen-free (SPF) animals, and is absent when the mice are maintained germfree. Colitis is evident by 1 week following bacterial colonization but becomes progressively more severe and involves the full thickness of the intestinal wall. Mortality is about 50% (Kuhn et al. 1993; Sellon et al. 1998; Sartor 2004). Antibiotic treatment can prevent or treat colitis in IL-10-deficient mice, further documenting the role of the gut microbiota in the pathogenesis of the disease (Madsen et al. 2000).

The gut microbiota of mice is not identical to that of humans but conforms to the general features of bacterial communities in the large bowel of mammals: obligately anaerobic bacteria are the numerically predominant members of the microbiota, many of which cannot yet be cultivated in the laboratory (Zoetendal et al. 1998; Salzman et al. 2002). Thus bacteriological culture-based investigations are of limited value until an overview of the microbiota has been obtained by nucleic acid-based methodologies, allowing particular species to be targeted subsequently. Comprehensive knowledge of the composition of the gut microbiota in relation to colitis in IL-10-deficient mice is not available (Madsen et al. 1999). Therefore, we used polymerase chain reaction coupled with denaturing gradient gel electrophoresis (DGGE) to screen the gut microbiota of formerly germfree IL-10-deficient mice that had been exposed to the faecal microbiota of SPF animals. PCR/DGGE has been used for the analysis of complex bacterial communities inhabiting the gut of healthy humans and other animals (Zoetendal et al. 1998; Tannock et al. 2000; Walter et al. 2000; Knarreborg et al. 2002). This method permits the detection of the numerically predominant members of the community, cultivable or not-yet-cultivated (Zoetendal et al. 1998). PCR-associated artefacts such as differential amplification of DNA templates, 16S rRNA gene copy number, primer specificity, template concentration and formation of chimeric sequences can skew assessments of biodiversity (Hugenholtz et al. 1998) but are of minimal significance in a comparative study where bacterial communities are screened to detect perturbations in composition. While PCR-associated artefacts could, in theory, affect quantitative assessment of bacterial populations, real-time quantitative PCR is a widely used method for this purpose, emphasizing that these theoretical objections are of small significance in comparative studies (Huijsdens et al. 2002; Requena et al. 2002; Bartosch et al. 2004; Gueimonde et al. 2004; Matsuki et al. 2004). In comparative studies, the analytical method remains constant and the communities are drawn from a common background. The aim of our study was to determine whether PCR/DGGE could detect changes in the composition of the large bowel microbiota of mice with a dysfunctional immune system during the development of intestinal disease. Changes in composition could point to bacterial species that might be used in future investigations of the interactions that occur between gut residents and the immune system.

Material and methods


Germfree IL-10-deficient mice on an inbred 129 Sv/Ev background were maintained in Trexler isolators. Wild type 129 Sv/Ev mice (Jackson Laboratories, Bar Harbor, ME, USA) were maintained under SPF conditions in filter top micro-isolators. The animals were shown to be Helicobacter-free by PCR (Sellon et al. 1998). Fresh faecal samples from the wild type mice were crushed in phosphate-buffered saline and used to gavage, and to inoculate per rectum, germfree mice that were 10–12 weeks old and which had been acclimated to sterile micro-isolators for 2 d. The formerly germfree mice were also exposed to bedding containing faeces that had been obtained from cages housing wild type mice.

Study design

Two experiments were carried out 4 months apart. In the first experiment, the animals were killed and gut samples were collected 2 weeks (three mice) or 5 weeks (four mice) after exposure to gut microbiota from SPF animals. Samples were also obtained from one wild type mouse so that the gut microbiota of donor and recipient animals could be compared. In the second experiment, samples were collected from animals killed 1, 3 or 6 weeks after exposure. As in the first experiment, samples were also obtained from wild type animals. Five mice were killed at each time point except at 6 weeks when six mice were examined. These experiments were approved by the Institutional Animal Care and Use Committees of the University of North Carolina and North Carolina State University. The caecum and colon were removed from each mouse, the caecal tip and a segment of the distal colon were retained for histological scoring, and the remainder snap-frozen in iso-pentane and shipped on dry ice to the analytical laboratory. The samples were stored at −80°C until analysis was carried out.

Histological scoring

Blinded histological scoring of haematoxylin and eosin-stained tissues was based on the degree of mononuclear cellular infiltration, crypt hyperplasia, goblet cell depletion, and architectural distortion using a validated scale (Sellon et al. 1998).

Nucleic acid extraction, PCR amplification and DGGE analysis

Caecal, proximal colon and distal colon contents were removed from the samples and DNA and RNA were extracted as described previously (Tannock et al. 2000, 2004). Although there is a stratification of the gut microbiota apparent in Gram-stained cryosections of the mouse colon (Savage et al. 1968), the composition of the microbiota of extruded bowel contents provides a composite picture of both mucus-associated and luminal bacterial populations (McBurney 2003). The hypervariable V3 region (corresponding to nucleotides 339 to 539 in the Escherichia coli gene) of the bacterial 16S rRNA gene was amplified using PCR or RT-PCR with universal bacterial primers (HDA1-GC, HDA-2), and DGGE was carried out as described previously (Tannock et al. 2000, 2004). In brief, DGGE was performed with a DCode universal mutation system (Bio-Rad, Hercules, CA, USA) utilizing 16-cm by 16-cm by 1 mm gels. Eight per cent polyacrylamide gels were prepared and run with TAE buffer (2 mol l−1 Tris base, 1 mol l−1 glacial acetic acid, and 50 mmol l−1 EDTA). The gels contained a 30–50% gradient of urea and formamide increasing in the direction of electrophoresis. A 100% denaturing solution contained 40% (v/v) formamide and 7·0 mol l−1 urea. The electrophoresis was conducted with a constant voltage of 130 V at 60°C for about 4 h 30 min. Gels were stained with ethidium bromide solution (5 μg ml−1; 20 min), washed with deionized water, and viewed by UV transillumination. Identification of bacterial origin of DNA sequences in the gels was performed as described previously (Knarreborg et al. 2002). Sequencing was carried out by the Centre for Gene Research, University of Otago, by the didoexy method of Sanger et al. (1977) by using PRISM BigDye Terminator Cycle Sequencing Ready Reaction kit (Applied Biosystems Inc., Foster City, CA, USA) in combination with an Applied Biosystems model 377A automated sequencing system. A scanned image of an electrophoretic gel was used to measure the staining intensity of the fragments. These measurements were made using Quantity One software (version 4·2; Bio-Rad Laboratories). The intensity of fragments was expressed as a proportion (%) relative to the sum of the intensities of all of the fragments in the same lane of the image so as to control for possible variations in the amount of DNA per PCR product used to load the gel.

Dot blot hybridization

To provide confirmatory evidence of alterations to the abundance of Bifidobacterium and Bacteroides during the development of colitis, we used labelled DNA probes that targeted 16S rRNA sequences and quantitative dot blot hybridization (Dore et al. 1998). Briefly, two to three replicates of RNA extracted, as described previously (Tannock et al. 2004), from gut contents, dilutions of RNA extracted from pure cultures of Bacteroides vulgatus ATCC 8482T and Bifidobacterium animalis DSM 20104T, and RNA from E. coli (strain W, Sigma, St Louis, MO, USA), as well as appropriate negative controls, were blotted onto nylon membranes. Hybridization was carried out using 32P-labelled probes [Bifidobacterium probe = PCR product amplified from DNA extracted from Bif. animalis DSM 21004T according to Satokari et al. (2001); Bacteroides sp. probe = appropriate fragment eluted from DGGE gel and amplified using HDA primers, nucleotides 319–500 of GenBank sequence AF157056 (Dewhirst et al. 1999); universal bacterial probe = Eub338 (Stahl et al. 1988)] and hybridization temperatures of 65, 65 and 42°C respectively. Washing temperatures were 65, 56, and 62°C for the respective probes. Washes were performed twice for 30 min with 0·5% SDS in 2X SSC buffer (1X SSC buffer for Bifidobacterium hybridization). Measurement of hybridization signals on dot blots was by autoradiography at room temperature and films were scanned and analysed using Quantity One software. The abundance of ribosomal RNA from each group was expressed as a proportion of that of the total microbiota (%) as determined by using the bacterial domain probe (Eub338).


Development of colitis

Inflammation of the large bowel tissue progressed with time (Fig. 1) and alterations to PCR/DGGE profiles of large bowel samples of IL-10-deficient mice were observed in association with the developing colitis. Distal colon microbiota profiles are shown as an example in Fig. 2. The same alterations were observed in caecal and proximal colon profiles, and in samples collected during the course of both experiments. The PCR/DGGE profiles generated from DNA or RNA did not differ (data not shown). Alignment of 16S rRNA gene sequences obtained from DNA fragments cut from DGGE gels with sequences in the NCBI (GenBank) database resulted in identification of the bacterial origins. As shown in Fig. 2, lactobacilli belonging to the Acidophilus group were commonly detected in the microbiota of wild type mice but rarely in that of IL-10-deficient mice (Fig. 2, fragments 6 and 10 were sequenced, 98% identity AF243165), whereas Lactobacillus murinus was present in all of the animals (Fig. 2, fragments 16 and 22, 100% identity AF157049). Eubacterium ventriosum was easily detected in wild type mice and in formerly germfree animals exposed to SPF gut microbiota 1 week previously (Fig. 2, fragment 5, 99% identity L34421), but not in animals with more aggressive colitis at 3 and 6 weeks. Enterococcus sp. was represented in the profiles of some IL-10-deficient mice (Fig. 2, fragment 9, 99% identity, AF145258) but was less obvious in wild type mice. Similarly a DNA fragment originating in Clostridium cocleatum (Fig. 2, fragment 19, 99% identity, Y18188) was more intensely stained in the profiles of IL-10-deficient mice compared with wild type. Uncultured bacteria were detected in wild type mice that appeared to be missing, or had reduced staining intensity, from the profiles of IL-10-deficient mice (Fig. 2, fragments 2, 3 and 4, 99–100% identity AF132240, AF371479, AJ400262 respectively). Most striking, however, was the increased intensity of staining of DNA fragments representing Bacteroides sp. (Fig. 2, fragment 18, 99% identity AF157056) and Bif. animalis (Fig. 2, fragments 8 and 12, 99% identity AB050138) in the profiles of IL-10-deficient mice at weeks 3 and 6 relative to those of 1-week-inoculated and wild type animals.

Figure 1.

Average histological scores (mean and SEM) for caecal and distal colonic tissues obtained from wild type (WT) and IL-10-deficient mice (1, 3 and 6 weeks after exposure to WT faecal microbiota). Five to six animals per group. *IL-10-deficient mouse groups significantly different from WT, Mann–Whitney test, P < 0·01. +IL-10-deficient 3-week group significantly different from 1 week, P < 0·01. #IL-10-deficient 6-week group significantly different from 1- and 3-week groups, P ≤ 0·01

Figure 2.

DGGE profiles of the distal colonic microbiota of wild type (WT) and IL-10-deficient mice (1, 3, and 6 weeks after exposure to faecal microbiota of WT mice). Numbers indicate DNA fragments that were cloned and sequenced. Identifications of bacterial species of origin are given in the text of the article. Circled areas and/or bacterial names indicate alterations in detection or intensity of staining of fragments in profiles

Relative abundance of rRNA of Bacteroides and Bifidobacterium measured by dot blot hybridizations

There was a correlation between the intensity of staining of the bifidobacterial fragments in DGGE profiles and the proportion of bifidobacterial rRNA that was detected relative to that of the total microbiota as determined by dot blot hybridization (Fig. 3; Spearman Rank Correlation r = 0·7162; P = 0·0003). Moreover, although there was variation in the abundance of the bifidobacterial rRNA from mouse to mouse, the proportion that these bacteria formed of the total microbiota was larger at 6 weeks compared with the populations in the wild type mice and in IL-10-deficient mice sampled 1 week after exposure to the wild type microbiota (Mann–Whitney test, P = 0·0173, 0·0087 respectively). The abundance of Bacteroides sp. rRNA measured by dot blot hybridization were not different between mouse groups (Mann–Whitney test, P > 0·05), probably because there may have been other members of this phylogenetic group present in the microbiota in addition to that represented by fragment 18, and some cross-hybridization may therefore have occurred. Measurement of the relative intensity of staining of fragment 18 in proportion to the total fragments in the profile, however, showed that this specific Bacteroides sp. population was more apparent in mice 3 and 6 weeks after inoculation relative to the wild type mice (Mann–Whitney test P = 0·0079, 0·0043 respectively).

Figure 3.

Comparison of staining intensity of DNA fragments originating from Bifidobacterium animalis (arrow) in DGGE gel (a) with quantitative data obtained by dot blot hybridization (b) for each corresponding colonic sample from individual mice. Five to six animals per group. WT = wild type mice; IL-10 = IL-10-deficient mice 1, 3, and 6 weeks after exposure to faecal microbiota of WT mice. Spearman Rank Correlation r = 0·7162; P = 0·0003


The PCR/DGGE detected changes in the composition of the large bowel microbiota of IL-10-deficient mice as caecal and colonic inflammation increased. DNA fragments originating in members of the Lactobacillus Acidophilus group were less apparent or missing from the DGGE profiles of IL-10-deficient mice 3 and 6 weeks after microbial exposure. This observation is of interest because, as reported by others, the administration of lactobacilli to colitic IL-10-deficient mice attenuated the development of colitis (Madsen et al. 1999; McCarthy et al. 2003). In contrast, fragments representing enterococci were more apparent in the profiles of IL-10-deficient mice at 3 and 6 weeks after microbial exposure. Enterococcus faecalis has been shown to induce colitis in monoassociated IL-10-deficient mice (Balish and Warner 2002). Alterations in microbiota composition with regard to Eu. ventriosum and Cl. cocleatum have not been investigated previously but could be worthy of further consideration in the light of our results.

The changes in intensity of staining of DNA fragments representing Bacteroides sp. and bifidobacterial populations, which was observed in profiles generated from both DNA and RNA templates, was particularly striking. Bacteroides vulgatus has been reported to induce colitis in HLA-B27 transgenic rats, but does not produce this effect in mice (Rath et al. 1999, 2001). The Bacteroides that we detected did not represent Bac. vulgatus according to 16S rRNA gene sequence analysis, but was characteristic (99% identity) of strain ASF 519 detected in the ‘Schaedler flora’ and ‘Altered Schaedler flora’ used as a standard inoculum in the derivation of SPF mouse colonies. This strain has resemblances to Bact. distasonis but belongs to an unnamed genus, closely related to Porphyromonas, in the CytophagaFlavobacterBacteroides phylum that is a common inhabitant of the murine gut (Dewhirst et al. 1999; Salzman et al. 2002).

Bifidobacteria are generally considered to be harmless members of the gut microbiota (Gibson and Roberfroid 1995) but the Bif. animalis rRNA abundance increased as a proportion of the total microbiota, as measured by quantitative dot blot hybridization, during the progression of colitis in IL-10-deficient mice. Data concerning relative rRNA abundance cannot be directly translated into bacterial cell numbers and may reflect altered metabolic activity rather than increased population size (Amann et al. 1995; Tannock et al. 2004). Nevertheless, taken together with the altered PCR/DGGE profiles detected using DNA as template, an increase in the size of the bifidobacterial population seems likely (Felske et al. 1997). Our study does not, of course, demonstrate cause and effect: the inflamed conditions of the bowel doubtless provide an altered environment in which some bacterial species may flourish while others might diminish in numbers. Nevertheless, Bif. animalis has also been detected in the inflamed bowel of HLA-B27 transgenic rats but not in that of their nontransgenic (and therefore nondiseased) counterparts (Schultz et al. 2004) and hence must be worthy of further investigation in relation to colitis.

Our study has shown that PCR/DGGE is an effective means of monitoring the composition of the gut microbiota during the development of intestinal disease and that, using dot blot hybridization as the comparative standard, the intensity of staining of fragments in DGGE profiles was correlated with the relative abundance of 16S ribosomal RNA representing the bifidobacterial population. PCR/DGGE thus provides an excellent means of screening the composition of bacterial communities of initially unknown composition to obtain targets for further study. While much of the species comprising the gut microbiota of mice have not yet been cultivated in the laboratory, the altered populations that we detected and their interaction with the immune system of gnotobiotic mice could, in future studies, likely be studied by culture-dependent methods.


The support of the Crohn's and Colitis Foundation of America is gratefully acknowledged, as are NIH grants RO1 DK53347 and P30 DK34987. R. Bibiloni was the recipient of a postdoctoral fellowship from the Consejo Nacional de Investigaciones Científicas y Técnicas, Argentina.