Mucosa-associated bacteria in ulcerative colitis before and after antibiotic combination therapy

Authors


Dr N. Sato, Department of Gastroenterology, Juntendo University School of Medicine, 2-1-1 Hongo, Bunkyo-ku, Tokyo 113-8421, Japan.
E-mail: nsato@med.juntendo.ac.jp

Summary

Background : We proposed that Fusobacterium varium is one of the causative agents in ulcerative colitis.

Aim : To examine the efficacy of antibiotic combination therapy against F. varium and to investigate the mucosa-associated bacteria before and after the therapy using a new molecular approach.

Methods : Twenty patients with ulcerative colitis were randomly assigned into the antibiotic treatment group (amoxicillin, tetracycline and metronidazole for 2 weeks) and no-antibiotics group. Clinical assessment, colonoscopic and histological evaluations were performed at 0 and 3–5 months after the treatment. DNA from mucosal bacteria was isolated from biopsy specimens. We investigated the mucosa-associated bacterial components by terminal restriction fragment length polymorphism with the restriction enzyme HhaI and MspI, and quantified the change in the number of bacteria by real-time polymerase chain reaction. Immunohistochemical detection of F. varium in biopsy specimens was also performed.

Results : After the treatment, the clinical assessment, colonoscopic and histological scores improved in the antibiotic group compared with the control group. Three peaks of terminal restriction fragment length polymorphism decreased after treatment only in the antibiotic group. Eubacterium rectale, Dorea formicigenerans, Clostridium clostridioforme and F. varium were included in these peaks. Based on the real-time polymerase chain reaction study, only F. varium was significantly reduced after treatment. In the immunostaining, post-treatment scores in treatment group were significantly lower than that in control group.

Conclusions : Antibiotics combination therapy was effective for ulcerative colitis. The number of mucosa-associated F. varium significantly decreased after the treatment.

Introduction

Although the aetiology of ulcerative colitis (UC) is poorly understood, UC shares histological features with infectious colitis.1 Luminal bacteria have been examined in detail, and their changes in patients with inflammatory bowel disease (IBD) have been reported by Benno and Mitsuoka.2 Because infections usually start with adherence of the micro-organism to host cells, the mucosal bacteria may play a more important role than the luminal bacteria.3 Recently, Swidsinski et al. and Fujita et al. investigated mucosal bacteria in IBD by polymerase chain reaction (PCR) of 16S rRNA.4, 5

Sakamoto et al. investigated the diversity of oral bacterial flora in saliva and the human faecal microbiota with a new molecular approach, terminal restriction fragment length polymorphism (T-RFLP) method.6, 7 T-RFLP allows an assessment of the diversity of complex bacterial communities and a rapid comparison between subjects.8, 9

Recently, we have shown that Fusobacterium varium was present in the mucosa of a significant number of patients with UC,10 and butyric acid in culture supernatants from cultures of F. varium caused UC-like lesions in mice.11 In this study, we investigated the efficacy of antibiotic combination therapy (amoxicillin, tetracycline and metronidazole were selected based on susceptibility tests with F. varium) for UC, and the changes of mucosa-associated bacterial components before and after treatment using T-RFLP, and also investigated quantitative changes in bacterial numbers by real-time PCR.

Materials and methods

Patients

Twenty-three patients with UC were enrolled consecutively between August and October 2001 at the Juntendo University Hospital. Entry criteria were high serum titres of immunoglobulins (IgG + IgM + IgA) to F. varium using enzyme-linked immunosorbent assay (ELISA) as described in a previous study,10 and mild-to-moderate disease with a score of at least 2 [on a scale of 1 (normal) to 5 (ulcers, erosion or necrosis of the mucosa with cellular infiltration)] on rectal histology.12 Patients were excluded if they had low serum titres of immunoglobulins (optical densities <0.14) to F. varium using ELISA, if they were newly presenting patients, or if they required hospitalization or intravenous corticosteroid therapy. Patients treated with antibiotics within 4 weeks before entry were also excluded. Three patients were excluded from the trial, one for each of the following reasons: low titre of antibody to F. varium, no inflammatory findings in the mucosa, and patient refusal. Patients were randomly assigned to one of two groups using a random number table and numbered sequentially. Therefore, 20 patients were eventually enrolled, and were randomized to treatment with amoxicillin 500 mg t.d.s., tetracycline 500 mg t.d.s. and metronidazole 250 mg t.d.s. for 2 weeks or no-antibiotics. These antibiotics were selected based on susceptibility tests for F. varium (minimum inhibitory concentrations (MICs): amoxicillin 1.56 μg/mL, tetracycline <0.39 μg/mL, metronidazole <1.56 μg/mL). Therapy was started after the initial endoscopy at entry. A follow-up endoscopy was carried out 3–5 months after the treatment to assess whether the antibiotic combination therapy against F. varium was effective for induction of remission of UC. Samples were taken from the appendix, caecum, ascending colon, transversal colon, descending colon, sigmoid colon and rectum at each endoscopic examination. Biopsy samples were immediately placed into formalin for histopathological assessment and liquid nitrogen to store at −80 °C for subsequent isolation of DNAs. Histopathology was assessed by two pathologists in a blind fashion. Clinical assessments were evaluated by the Lichitiger et al.'s symptom score.13 Endoscopic and histological scores were graded according to Matts's score.12 Histological scores were the sum of the scores of the appendix, caecum, ascending colon, transversal colon, descending colon, sigmoid colon and rectum. Patients who underwent colonoscopy because of occult stool blood, with normal colonoscopic and histological findings were also included as normal controls (n = 6, mean age 63, range: 51–71, male/female; 5/1). This study was approved by the institutional review board of our hospital, and written informed consent was obtained from all patients.

Isolation of DNAs from mucosal bacteria

Each biopsy sample was washed by gently shaking in sterile water to remove loosely adherent faecal material and mucus. DNAs from biopsy samples were extracted using the UltraClean Soil DNA Isolation Kit (Mo Bio Laboratories, Solana Beach, CA, USA) with some modifications.14 Each biopsy sample was put into 2 mL Bead Solution tubes (containing 550 μL Bead Solution) and mixed by vortexing on a FastPrep instrument (Bio 101, Vista, CA, USA) for 10 s at 4 m/s, followed by addition of lysozyme (final concentration of 1 mg/mL) and N-acetylmuramidase (final concentration of 1 mg/mL), and incubation at 37 °C for 1 h. After incubation, 60 μL of Solution S1 and 200 μL of Solution IRS (inhibitor removal solution) were added and the samples were vortexed on a FastPrep for 10 s at 4 m/s. The supernatant was collected by centrifugation at 12 000 g for 5 min, and mixed with Solution S2, vortexed gently, and put on ice for 5 min. The mixture was centrifuged at 13 000 rpm for 1 min and the supernatant was transferred to a clean microcentrifuge tube. After addition of 900 μL of Solution S3, the sample was vortexed gently. The mixture was loaded onto a spin filter tube and centrifuged at 12 000 g for 1 min. The flow-through solution was discarded, 300 μL Solution S4 was added, and the spin filter was centrifuged at 13 000 rpm for 1 min. The flow-through solution was discarded and the spin filter was placed in a new tube. About 50 μL of Solution S5 was added and after the sample was centrifuged at 13 000 rpm for 1 min, the spin filter was discarded and the DNA was collected in the tube.

PCR amplification of 16S rDNA and T-RFLP analysis

The PCR amplification of 16S rDNA for T-RFLP was carried out as described previously with some modifications.6 The primers used for the PCR amplification of 16S rDNA sequences were 27F (5′-AGAGTTTGATCCTGGCTCAG-3′) and 1492R (5′-GGTTACCTTGTTACGACTT-3′).15 Primer 27F was labelled at the 5′-end with 6′-carboxyfluorescein (6-FAM), which was synthesized by Applied Biosystems Japan (Tokyo, Japan). Amplification reactions were performed in a total volume of 50 μL containing 5 μL of dissolved DNA (100 ng), 1.25 U of TaKaRa Ex Taq (Takara Shuzo, Shiga, Japan), 5 μL of 10X Ex Taq buffer, 4 μL of dNTP mixture (2.5 mm each), and 10 pmol of each primer. The 16S rDNAs were amplified in a Biometra Thermocycler TGradient (Biometra, Goettingen, Germany) using the following programme: 95 °C for 3 min, followed by 30 cycles consisting of 95 °C for 30 s, 50 °C for 30 s and 72 °C for 1.5 min, with a final extension period at 72 °C for 10 min. This programme was repeated a second time. Amplified DNA was verified by electrophoresis of aliquots of PCR mixtures (2 μL) on 1.5% agarose gels in 1X TAE buffer. PCR products were purified by the polyethylene glycol (PEG) precipitation method16 with some modifications. A 50 μL aliquot of the 16S rDNA solution was mixed with 30 μL of a PEG solution (40% PEG 6000 and 10 mm MgCl2) and 12 μL of 3 m sodium acetate, and centrifuged at 14 000 rpm for 15 min. The supernatant was carefully removed by pipetting, and then the precipitated DNA was washed twice with 70% ethanol and redissolved in 20 μL of sterile distilled water.

The restriction enzymes were selected according to Moyer et al.17 Purified PCR product (2 μL) was digested with 20 U of either HhaI or MspI (Takara Shuzo) in a total volume of 10 μL at 37 °C for 3 h. The restriction digestion product (1 μL) was mixed with 12 μL of deionized formamide and 1 μL of DNA fragment length standard. The standard size marker was a 1:1 mixture of the size standard GS 500 ROX (including 35, 50, 75, 100, 139, 150, 160, 200, 300, 350, 400, 450, 490 and 500 bp) and GS 1000 ROX (including 29, 33, 37, 64, 67, 75, 81, 108, 118, 244, 275, 299, 421, 539, 674, 677 and 926 bp; Applied Biosystems). Each sample was denatured at 95 °C for 2 min and then immediately placed on ice. The length of the T-RF was determined on an ABI PRISM 310 genetic analyzer (Applied Biosystems) in GeneScan mode (15 kV, 8 mA and 60 °C for 48 min for each sample). Fragment sizes were estimated by using the Local Southern Method in the genescan 3.1 software (Applied Biosystems).

Cloning, DNA sequencing and computer simulations

The PCR was performed using two primers, 27F (without 6-FAM label) and 1492R. The amplification reaction and programme were as described in ‘PCR amplification of 16S rDNA for T-RFLP’. The PCR products were analysed by 1.5% agarose gel electrophoresis in Tris-acetate ethylenediaminetetraacetic acid (EDTA) buffer, and stained with 0.5 μg/mL ethidium bromide and visualized under a 300 nm UV light. The target bands were cut from the gel, and the 16S rDNA genes present in the gel were purified using the Ultra Clean GelSpin DNA Purification Kit (Mo Bio Laboratories). Purified DNA was ligated into the plasmid vector pUC118. PCR was performed using two primers, 27F and M3 (5′-GTAAAACGACGGCCAGT-3′). The PCR and programme were as described in ‘PCR amplification of 16S rDNA for T-RFLP’. The PCR product was purified using the Ultra Clean PCR Clean-up Kit (Mo Bio Laboratories), ligated into the plasmid vector pCR 2.1, and transformed into One Shot (Invitrogen Corp, Carlsbad, CA, USA) INVαF’ competent cells.18

Plasmid DNAs were prepared from randomly selected recombinants and used as templates for sequencing. Sequencing was conducted using the 520F (5′-GTGCCAGCAGCCGCGG-3′), 520R (5′-ACCGCGGCTGCTGGC-3′), T7 (5′-TAATACGACTCACTATAGGG-3′), RV (5′-CAGGAAACAGCTATGAC-3′) primers, a Big Dye Terminator Cycle Sequencing Kit (Applied Biosystems) and an ABI PRISM 3100 Genetic Analyzer (Applied Biosystems). All sequences were compared with sequences similar to the reference organisms by blast analysis.19

The computer-simulated digestion with HhaI and MspI was done with the genetyx-mac program (Soft Developing Co., Tokyo, Japan).

Real-time PCR

The primer sets except for those specific for Dorea formicigenerans and F. varium were designed as previously reported.20, 21 The primer sets for D. formicigenerans and F. varium were designed by aligning their 16S rDNA gene sequences (obtained from GenBank database) with those from closely related bacteria using the clustal w program.22 The specificity of the primers was tested using bacterial strains in taxonomically related genera and species. The primer set for D. formicigenerans was tested against D. formicigenerans (ATCC 27755), D. longicatena (JCM 11232), Clostridium oroticum (JCM 1492) and C. aminovalericum (JCM 11016), and the primer set for F. varium was tested against F. varium (JCM 3722), F. nucleatum ssp. nucleatum (JCM 6328), F. necrophorum ssp. necrophorum (JCM 3716) and F. necrophorum ssp. fundiforme (JCM 3717). The primer set for human β-actin was purchased from Maxim Biotech Inc (South San Francisco, CA, USA).

DNAs isolated from biopsy samples were amplified with species-specific primers and the primer set for human β-actin. PCR was done in a 50 μL volume containing 2.5 μL of DNA sample, 25 μL of SYBR Green PCR Master Mix (Applied Biosystems) and 15 pmol of each primer. Amplification and detection were done with a sequence detection system (ABI PRISM 7900HT; Applied Biosystems). The protocol for C. clostridioforme, D. formicigenerans, Eubacterium rectale and F. varium was 50 °C for 2 min, 95 °C for 10 min, and then 50 cycles at 95 °C for 15 s and 60 °C for 1 min. The protocol for human β-actin was 50 °C for 2 min, 95 °C for 10 min, and then 50 cycles at 95 °C for 30 s and 60 °C for 1 min. The number of bacterial cells or human β-actin gene copies in tissue samples was estimated with a standard curve for each species or human β-actin gene. The standard curves were drawn using results from five concentrations of bacterial cells (107, 106, 105, 104 and 103 cells/μL) or human β-actin gene (106, 105, 104, 103 and 102 copies/μL). Finally, the results of real-time PCR for each sample were expressed as the number of bacterial cells per 106 copies of human β-actin gene.

Immunohistochemical detection of F. varium

Immunohistochemical detection of F. varium in biopsy specimens was performed using the avidin–biotin complex method.10, 23 The primary antibodies (Protein Refining Co., Maebashi, Japan) used for antigen detection were from rabbits immunized with whole cell preparation of F. varium that had been treated with 3% buffered formalin. The antibodies were affinity purified and peroxidase-labelled. The histological sections were incubated for 30 min with the primary antibody, diluted 100-fold, and then processed with the Dako Envision peroxidase kit system (Dako Envision, peroxidase, Dako Corp., Carpinteria, CA, USA). Normal rabbit serum, diluted 100-fold, was used for the negative control. Smears of isolated bacteria were treated the same way as the biopsy samples in order to confirm the reactivity of the primary antibody. Grading the density of the organism present in the specimens was given a score of 0, 1, 2 and 3. Severe colonization, graded 3, indicated the presence of large groups of organisms on the surface and upper pits of more than two-thirds of the mucosal surface examined. Mild colonization, graded 1, meant individual organisms, or small groups, covering less than one-third of the mucosal surface. Moderate colonization, graded 2, was midway between these two. Fusobacterium varium staining scores were the sum of the scores of appendix, caecum, ascending colon, transversal colon, descending colon, sigmoid colon and rectum.

The doctors doing endoscopic and pathological examinations did not know the treatment history of the patients. In addition, samples for T-RFLP, PCR and immunohistochemical studies were all numbered for blindness of the treatment of the patients in this study until the statistical analyses were carried out.

Statistical analysis

The Mann–Whitney U-test was used for statistical comparison of clinical, pathology, colonoscopic grade, and the number of bacterial cells at entry and 3–5 months later in the two treatment groups, and the number of bacterial cells between UC and normal controls. stat view software, version J 5.1 (SAS Institute, Inc., Cary, NC, USA), was used for all analyses. Differences with P < 0.05 were considered to be statistically significant.

Results

Efficacy of the treatment

Of the 20 patients enrolled in the study, mucosal DNA in the rectum was available in 18 patients (Table 1). All patients were undergoing oral therapies of sulfasalazine, mesalazine, steroid and/or probiotics therapy (Streptococcus faecalis and Bacillus subtilis) before entry. At entry, all patients except two in each group had symptoms such as diarrhoea or visible blood in the stool. There were no differences in age, gender, duration of disease, duration of current activity, extent of disease or concomitant medications between the antibiotic treatment and no-antibiotics groups. Before the treatment, there were also no differences in clinical assessment, colonoscopic or histological scores between the two groups (Table 2). After the treatment, the symptom score in the antibiotic treatment group was improved when compared with the no-antibiotics group (P = 0.022). Two patients in the no-antibiotics group had relapsed up to 5 months after entry; however, their symptoms were mild and sustaining during the trial. The colonoscopic and histological scores in the antibiotic treatment were also lower than those of the no-antibiotics group (P = 0.001 and 0.036 respectively). One patient in the treatment group had reversible urticaria and fever. No other serious drug-related toxicity was observed during the trial.

Table 1.  Entry data in antibiotics and no-antibiotics groups
CharacteristicsAntibiotics group (n = 10)No-antibiotics group (n = 8)P-value
  1. 5-ASA, mesalazine.

Mean age, year (range)39 (21–58)43 (16–73)>0.2
Gender, male/female5/56/2>0.2
Mean duration of disease, year (range)14 (3–45)13 (3–26)>0.2
Mean duration of current activity, week (range)1.75 (0–4)1.80 (0–4)>0.2
Extent of disease, number of patients (%)
 Pancolitis7 (70)6 (75)>0.2
 Left-sided colitis2 (20)2 (25)
 Proctitis1 (10)0 (0)
Concomitant medications before and during the trial, number of patients (%)
 Sulfasalazine9 (90)7 (88)>0.2
 5-ASA1 (10)1 (13)
 Glucocorticoid1 (10)2 (25)
 Probiotics4 (40)4 (50)
Table 2.  Symptom score, endoscopic finding score and pathological score in antibiotics and no-antibiotics groups
VariableAntibiotics group (n = 10)No-antibiotics group (n = 8)P-value
  1. * Symptom grades were scored according to Lichtiger et al.’s13 symptom score.

  2. † Endoscopic and histological findings were scored according to Matts's12 grading score.

Mean symptom score (range)*
 Before4.8 (2–10)3.1 (2–9)0.113
 After2.4 (0–3)4.3 (2–8)0.022
Mean endoscopic finding score (range)†
 Before2.9 (2–4)2.1 (1–4)0.066
 After1.3 (1–2)2.8 (1–4)0.001
Mean pathological score (range)†
 Before15.5 (13–21)15.4 (11–22)>0.2
 After14.0 (11–18)16.4 (11–20)0.036

T-RFLP analysis

The T-RFLP analysis was used to characterize the differences in the mucosal flora between UC and normal controls, and also before and after antibiotic treatment. The peak heights of each T-RF length (bp) show the relative quantity of 16S rDNA fragments of each T-RF length. Representative T-RFLP patterns are shown in Figure 2. There were no characteristic peak patterns in UC at entry compared with the normal control (Figure 2a,b). Peak heights of >1000 bp and approximately 200 bp fragments for the restriction enzyme HhaI (Figure 1a) and the approximately 200 bp fragment for MspI (Figure 2b) were reduced after treatment in all patients in the antibiotic treatment group, but no changes were seen in the no-antibiotics group. These results indicate that bacteria whose 16S rDNA fragment lengths after restriction by HhaI were larger than 1000 bp, and approximately 200 bp and for MspI were approximately 200 bp, respectively, were reduced after antibiotic treatment.

Figure 2.

(a) Terminal restriction fragment length polymorphism (T-RFLP) patterns of 16S rDNAs from mucosal bacteria digested with HhaI. 16S rDNAs were amplified with universal primers 27F [with 6′-carboxyfluorescein (6-FAM)] and 1492R. The peaks of more than 1000 bp (arrows) and approximately 200 bp (dot arrows) in the antibiotic group generally tended to decrease after treatment. This reduction was not found in the no-antibiotics group after treatment. There were no characteristic peak patterns in ulcerative colitis (UC) compared with normal controls. (b) T-RFLP patterns of 16S rDNAs from mucosal bacteria digested with MspI. 16S rDNAs were amplified with universal primers 27F (with 6-FAM) and 1492R. The peak of approximately 200 bp (arrows) in the antibiotic group generally tended to decrease after treatment. This reduction was not found in the no-antibiotics group after treatment. There were no characteristic peak patterns in UC compared with normal controls.

Figure 1.

Trial profile.

To identify the bacteria included in these peaks, cloning and DNA sequencing were done. DNAs included in the larger than 1000 bp fragment with HhaI/and approximately 200 bp fragments with MspI showed high similarities with 16S rDNA from D. formicigenerans, C. clostridioforme and uncultured bacterium HuCB21. Those corresponding to the approximately 200 bp fragment with HhaI/around 200 bp fragment with MspI showed high similarities with 16S rDNA of E. rectale and uncultured E. NB2A11. The similarity rate was more than 98%.

In the previous study, we detected 20 bacterial species by the culture method,11 and the T-RF lengths of these 20 bacterial species by computer-simulated digestion were determined with HhaI and MspI. Clostridium clostridioforme (1090 bp with HhaI/147 bp with MspI) and F. varium (199 bp with HhaI/151 bp with MspI) were included in these peaks.

Real-time PCR

We investigated the quantitative changes of these bacteria before and after treatment in the antibiotic treatment group (n = 10), no-antibiotics group (n = 8) and normal control group (n = 6). The species-specific primers are shown in Table 3. Real-time PCR was done for C. clostridioforme, D. formicigenerans, E. rectale and F. varium because of the results of the cloning/sequencing analysis or the computer-simulated T-RF length. Uncultured bacterium HuCB21 and E. NB2A11 were excluded because we could not design specific primers for their detection or examine the specificity of the primers used for these strains.

Table 3.  Sequences of primers
TargetSequences (5′–3′)
Clostridium clostridioforme 16S rDNACCGCATGGCAGTGTGTGAAA
CTGCTGATAGAGCTTTACATA
Dorea formicigenerans 16S rDNACAAGTCGAGCGAAGCACATA
CAGACTTGCACAGCCGTCTA
Eubacterium rectale 16S rDNATACAGGGGGATAACAGTTGG
GTCACCCTCTCAGGTCGGC
Fusobacterium varium 16S rDNACCGTTGCTTATATGGGGTTG
CCTCGCAGATTCACACAGAA
Human β-actinGCCAACCGCGAGAAGATGACC
CTTTAGCACGCACTGTAATTCCTC

The numbers of bacterial cells per 106 copies of human β-actin gene of all four of the species examined in UC were statistically higher than those of the normal control at entry, C. clostridioforme: the median number was 132.9 (with the 25th and 75th percentiles was 6.5 and 365.2) in UC compared with 0.2 (0.1 and 1.5) in the normal control (P = 0.0051); D. formicigenerans: 36.9 (2.5 and 439.6) in UC compared with 0.8 (0.2 and 7.4) in the normal control (P = 0.0113); E. rectale: 710.9 (460.7 and 2677.4) in UC compared with 49.2 (21.3 and 102.7) in the normal control (P = 0.0069); F. varium: 129.3 (21.1 and 281.0) in UC compared with 2.3 (1.6 and 2.8) in the normal control (P = 0.0011; Figure 3).

Figure 3.

The numbers of mucosa-associated bacterial cells of Clostridium clostridioforme, Dorea formicigenerans, Eubacterium rectale and Fusobacterium varium per 106 copies of human β-actin gene detected in the samples of UC and NL were examined at entry. All four bacterial species examined in UC were statistically higher than those of NL. UC, ulcerative colitis; NL, normal control.

Figure 4 shows the number of bacterial cells per 106 copies of human β-actin before and after treatment of the antibiotics and no-antibiotics groups. The numbers of C. clostridioforme in the antibiotics and no-antibiotics groups were 172.7 (6.5 and 1617.0), 106.4 (36.0 and 317.5) at entry, and 81.5 (8.1 and 742.1), 182.2 (22.9 and 1162.9) at 3 months after the treatment, respectively. The number of C. clostridioforme in the antibiotic treatment group tended to decrease, but not significantly, after the treatment when compared with that in the no-antibiotics group (P = 0.9292). The numbers of D. formicigenerans in the antibiotics and no-antibiotics groups were 137.3 (23.3 and 748.7), 31.8 (1.7 and 50.2) at entry, and 102.6 (8.2 and 821.8), 79.7 (10.2 and 1322.5) after treatment, respectively. The numbers of E. rectale in the antibiotics and no-antibiotics groups were 1038.3 (460.7 and 3926.8), 1449.3 (959.4 and 3018.6) at entry, and 1984.0 (599.3 and 4797.2), 2833.6 (688.4 and 14821.5) after the treatment, respectively. The numbers of D. formicigenerans and E. rectale in the antibiotics group were similar to those in the no-antibiotics group before and after the treatment (P > 0.2). The numbers of F. varium in the antibiotics and no-antibiotics groups were 90.0 (7.9 and 208.5), 200.3 (73.0 and 524.5) at entry, and 30.5 (6.5 and 85.9), 552.4 (142.1 and 1060.7) after the treatment, respectively. The number of F. varium in the antibiotics group decreased significantly after the treatment when compared with that in the no-antibiotics group (P = 0.0410). When the postantibiotic bacterial loads were compared with preantibiotic bacterial loads, only F. varium was significantly decreased after the antibiotic treatment (P = 0.0218, analysed by the Wilcoxon signed rank test). The numbers of all four bacteria after antibiotic treatment were significantly higher than in the normal controls (C. clostridioforme, P = 0.0062; D. formicigenerans, P = 0.0093; E. rectale, P = 0.0124 and F. varium, P = 0.0041), however, the numbers of F. varium in four of 10 patients in treatment group were similar to those in the normal control.

Figure 4.

The numbers of mucosa-associated bacterial cells of Clostridium clostridioforme, Dorea formicigenerans, Eubacterium rectale and Fusobacterium varium per 106 copies of human β-actin gene detected in the samples of ulcerative colitis (UC) in the antibiotics group and no-antibiotics group were examined at entry and after treatment. Only the number of F. varium decreased significantly after the treatment.

Immunohistochemical assessment

In the F. varium immunostaining, the median score for the antibiotics group was 17.5 (the 25th and 75th percentiles: 13.5 and 19.0), and the score for the no-antibiotics group was 17 (15 and 18) before treatment (Figure 5). The post-treatment score for the treatment group (median, the 25th and 75th percentiles: 7.0, 3.0 and 11.5) was significantly lower than that of the no-antibiotics group (19.0, 17.0 and 20.0, P = 0.0006). In the treatment group, a decrease, no change and an increase in the F. varium scores were found in nine patients, none, and one patient, respectively. By contrast, in the control group, a decrease, no change and an increase in the F. varium density were found in one, one and six patients, respectively. However, the scores of the single patient who had a decrease in the F. varium density were still twofold higher compared with those of the patients in the treatment group.

Figure 5.

Cumulative Fusobacterium varium density in the mucosa with immunostaining. The score of F. varium density in the mucosa in the antibiotics group after the treatment was lower than that in the no-antibiotics group (P = 0.0006).

Discussion

The T-RFLP was first introduced in 1997,8 and many research articles were published that included this powerful tool in their reports.6, 7 T-RFLP allows an easy assessment of the diversity of complex bacterial communities, because the fluorescently labelled T-RFs are visualized. This report is the first to use T-RFLP to investigate mucosa-associated bacteria before and after antibiotic combination therapy. The change in the peak pattern after antibiotic treatment for UC is clearly shown in this study. The peak heights of the >1000 bp fragment and the approximately 200 bp fragment produced with the restriction enzyme HhaI and the approximately 200 bp fragment with MspI decreased after treatment only in the antibiotic treatment group. There are many species of bacteria whose predicted 16S rDNA lengths after restriction with HhaI/MspI are more than 1000 bp/approximately 200 bp, and approximately 200 bp/approximately 200 bp, including Clostridium subcluster XIVa.24 The four bacterial species chosen for real-time PCR were thought to be included in these peaks. We chose C. clostridioforme, E. rectale and D. formicigenerans because of the results of cloning/sequencing and the results of the computer-simulated T-RF length analysis. We chose F. varium and C. clostridioforme because the computer-simulated T-RF length analyses were determined for the 20 bacterial species that we isolated from the colonic mucosa in active UC by the culture method in a previous study.11 The number of bacterial cells of all four species in UC were statistically higher than those of the normal control at entry. Swidsinski et al. reported that the concentration of mucosal bacteria in UC patients was higher than that of a normal control,4 and Fujita et al. reported that the number of B. vulgatus in colon tissue samples in UC patients was higher than those of normal mucosa in colon cancer patients.5 Therefore, the three studies including our study indicate that the concentration of mucosal bacteria in UC patients was higher than those in normal controls. The high concentration of mucosal bacteria in UC patients may be caused by a breakdown in barrier function, such as an abnormal immune response to normal flora. Toll-like receptors (TLR) were found to play an important role in the innate immune system,25 and Cario and Podolsky reported that the expression pattern of TLR homologues in intestinal epithelial cells from patients with UC were distinctly different from those in normal controls.26 Swidsinski et al. hypothesized that the healthy mucosa is capable of holding pathogenic luminal bacteria and that this function is profoundly disturbed in patients with IBD.4 Therefore, the different expression pattern of TLR may cause the breakdown in barrier function to bacteria.

Only the number of F. varium was significantly reduced after the antibiotic treatment, and clinical assessment and colonoscopic and histological scores in the antibiotics group showed improvement. These results support the hypothesis that F. varium may be one of the causative agents in UC. The relationship between the clinical improvement and the number of F. varium should be studied, because all patients in the antibiotics group showed clinical improvement in this study and we were unable to investigate the specific relationship. T-RFLP showed that the peak of the >1000 bp fragment produced with HhaI decreased after treatment, but the numbers of C. clostridioforme and D. formicigenerans (their T-RFs in HhaI are >1000 bp) were not significantly lower after treatment. The reason may be that there are other species including uncultured bacteria and unknown bacteria represented in the >1000 bp HhaI peak, and therefore we were unable to detect the appropriate bacteria. Other causative agents in UC may be included in this peak.

Although there have been several reports of antibiotic therapy for UC,27–31 there has been no studies of three-antibiotics combination therapy in UC in a randomized-controlled clinical trial setting. We chose three antibiotics (amoxicillin, tetracycline and metronidazole), because they were most effective on F. varium. This may be one of the reasons that only F. varium was reduced after the antibiotics combination therapy. Because we determined that Helicobacter pylori could not be effectively eradicated using only a single drug,32 we chose three antibiotics for this study. Because clinical assessment, and colonoscopic and histological scores showed a statistical improvement in the antibiotics treatment group, this treatment was thought to be effective for UC. Although the number of mucosal bacterial cells of all four species were higher in UC than in normal controls, only F. varium was significantly reduced after the combination antibiotic therapy. This pilot study was the first trial with a combination of three antibiotics selected to specifically target F. varium. Because of the relatively small number of patients examined and the lack of a placebo control group, the findings of this study must be viewed with caution.

We have already shown that F. varium was present in the mucosa of a significant number of patients with active UC compared to patients with UC in remission, Crohn's disease, ischaemic colitis and healthy controls.10 We also found that butyric acid in culture supernatants from cultures of F. varium caused UC-like lesions in mice.11 These findings and the results of this study support the hypothesis that F. varium is one of the causative agents of UC.

Acknowledgement

This study was supported in part by Grants-in-Aid for Scientific Research (C) from the Ministry of Education, Science, Sports and Culture, Japan.

Ancillary