Is the abundance of Faecalibacterium prausnitzii relevant to Crohn's disease?

Reports that bacteria within the Firmicutes phylum, especially the species Faecalibacterium prausnitzii, are less abundant in Crohn's disease (CD) patients and supernatants from cultures of this bacterium are anti-inflammatory prompted the investigation of the possible correlations between the abundance of F. prausnitzii and the response to treatment in patients with gut diseases and healthy controls. In a randomized, double-blind trial, faeces were collected from healthy volunteers, and from patients with active CD, ulcerative colitis (UC) and irritable bowel syndrome before and after treatment. The levels of F. prausnitzii DNA in faecal suspensions were determined by PCR. Treatment by an elemental diet was effective, resulting in decreases in both the Harvey and Bradshaw index (P<0.001) and the concentrations of serum C-reactive protein (P<0.05). The total levels of F. prausnitzii in faecal samples from CD patients at presentation were lower than those in the other groups both before and after the treatment. There was no correlation between F. prausnitzii abundance and the severity of CD before treatment. Clinical improvement unexpectedly correlated with a significant decrease in the abundance of F. prausnitzii, especially the A2-165 subgroup (P<0.05). Our data suggest that a paucity of F. prausnitzii in the gastrointestinal microbial communities is likely to be a minor aetiological factor in CD: recovery following elemental diet is attributed to lower levels of gut flora.


Introduction
The cause of Crohn's disease (CD) is unknown and successful, long-term treatment represents considerable clinical and pharmacological challenges. The conventional treatment of disease of mild to medium severity consists of antibiotics to counteract infection and 5-aminosalicylic acid (5-ASA) with or without corticosteroids to counter inflammation, but remission can also be achieved with enteral feeding (King et al., 1997). The faecal bacterial flora in CD is known to be abnormal (Sartor, 2008). CD has been suggested to be an autoimmune disease with an immune attack directed against the resident bacteria of the colon (Macpherson et al., 1996). The coating of colonic bacteria with immunoglobulin approaches 100% in active CD, as compared with 20% in healthy subjects and patients with irritable bowel syndrome (IBS). After enteral feeding, this coating is rapidly reduced (van der Waiij et al., 2004).
There is increasing evidence that gut microorganisms play key roles in gastrointestinal disease (Sartor, 2006;2008). Some are believed to generate a harmful toxin (Pitcher & Cummings, 1996;Loubinoux et al., 2002). Many others are known to yield butyrate, which is not only an essential energy source for colonocytes (Chapman et al., 1994;Costello et al., 1994;Macpherson et al., 1996;Aminov et al., 2006;Mai & Draganov, 2009) but also is both anti-inflammatory and anticarcinogenic (Hamer et al., 2008;Tazoe et al., 2008). Therefore, considerable interest has been generated by reports that a reduction in bacteria within the Firmicutes phylum, especially the species Faecalibacterium prausnitzii, has been observed in the gut of CD patients and that supernatants from cultures of this bacterium are anti-inflammatory (Sokol et al., 2008). Because oxidation of butyrate is the major source of ATP in cells lining the colon (Yin et al., 2001), decreased levels of F. prausnitzii could well reduce ATP within these cells. At the same time, coupled with the loss of anti-inflammatory agents from F. prausnitzii, this would, in a double blow, weaken the ability of the cells to fight infection (Smith et al., 2008). Thus, these findings potentially shed light on the aetiology of CD and offer a novel therapeutic approach involving recolonization of the gut with the bacterium.
To evaluate these findings and to assess the effects of elemental diet therapy (ED) on the population of F. prausnitzii within the gut, faecal samples were collected before and after therapy from patients suffering from CD, IBS and ulcerative colitis (UC) as disease control groups, and also from healthy controls.

Patients and control subjects
Twenty patients diagnosed to be suffering from CD, 21 patients with IBS, 14 patients with UC and 18 healthy subjects as controls were recruited at Addenbrookes Hospital, Cambridge, which functions as a tertiary referral centre for inflammatory bowel disease. A faecal sample and a blood sample were taken at the start of the treatment and again (except for the controls) 2 weeks later. Faecal samples were collected at the homes of patients and control subjects and immediately stored at 4 1C. Samples were taken to Addenbrookes Hospital, assigned undisclosed identification codes and sent by courier at 4 1C to the University of Birmingham for DNA extraction, arriving within 48 h of collection. Sample identifications were decoded only after data collection had been completed. Ethical permission to collect these samples was obtained from Leeds West Ethics Research Committee, ref. 07/Q1205/39, and informed, written consent was obtained from each subject.

Treatment regimes and assessment of the effects of treatment
Most CD patients attending the Addenbrookes Hospital tertiary referral clinic were considered to be relatively complex cases that had proved refractory to previous pharmacological treatments. These included steroids and/ or immuno modulation therapy (azathioprine or methotrexate), or 5-ASA. Treatment was chosen by the individual consultant supervising each case, but all included in the present study were given ED (King et al., 1997). All normal food items were withdrawn and patients were maintained for 2 weeks on ED (E 028 Extra, Scientific Hospital Supplies International, Liverpool, UK), which contains a predigested mixture of amino acids, malto-dextrins, minerals and vitamins, with a single fat source, rape seed oil. Thus, no additional complex carbohydrate or protein was provided for the gut flora to use as energy substrates other than endogenous sources, for example, protein released from sloughed cells. Water ad libitum was the only other item allowed. The nutritional requirements were individually calculated using the Schofield equation (Schofield, 1985). The efficacy of treatment was assessed on symptoms using the Harvey and Bradshaw index (Harvey & Bradshaw, 1980), and by an assay of serum C-reactive protein (CRP) as a measure of inflammatory stress.
None of the UC and IBS patients was given ED. Eleven of the UC patients had received either immuno-modulation or 5-ASA therapy before this study, and this was continued, albeit with changes in the drugs and increases in the doses used. Treatment-naïve patients were placed on similar regimens. IBS patients received conventional treatment.
The Harvey and Bradshaw index and the Walmsley index were used to determine the severity of CD and UC, respectively.

DNA extraction
DNA was extracted from each faecal sample (0.2 g) using a QIAamp DNA stool Mini Kit (Qiagen) by following the manufacturer's instructions. Extracted DNA was stored in aliquots at À 20 1C. DNA quality was assessed by determining the ratio of A 260 nm to A 280 nm .

PCR amplification of bacterial DNA
Two subgroups of F. prausnitzii have been identified on the basis of their published whole-genome DNA sequences. Primers Fp.ID.F2 (GTGACCGGATCGAACGACC) and Fp.ID.R2 (TCCAGGTCATGTGGGCAGC) were designed against the nucleotidyl transferase gene and the butyryl-CoA transferase gene, respectively, using the nucleotide sequences of both F. prausnitzii A2-165 and M21/2. Amplicons from the two F. prausnitzii subgroups could be generated simultaneously within a single PCR reaction and distinguished by the fragment length: approximately 650 bp for the A2-165-related subgroup and 778 bp for the M21/2related subgroup. For each PCR amplification, 50 ng of DNA was used as a template with HotStarTaq Plus Master Mix Kit (Qiagen) plus 4 mL of Q-solution (Qiagen) according to the manufacturer's instructions. The reaction conditions were: initial denaturation (5 min at 95 1C), then 35 cycles of denaturation (1 min at 94 1C), annealing (1 min at 58 1C) and elongation (1 min at 72 1C) and a final extension (10 min at 72 1C).
For samples from which F. prausnitzii was not amplified, the quality of the DNA was confirmed using 50 ng of DNA as template in a second PCR with the primers 27F (5 0 -AG AGTTTGATCATGGCTCAG-3 0 ) and 1492R (5 0 -GGTTACC TTGTTACGACTT-3 0 ) that annealed to bacterial DNA encoding 16S rRNA gene (Lane, 1991). These primers amplify a highly conserved region in the DNA from many bacterial species in human faeces. The reaction conditions were as follows: initial denaturation (5 min at 95 1C), then 35 cycles of denaturation (30 s at 94 1C), annealing (1 min at 55 1C) and elongation (2 min at 72 1C) and a final extension (8 min at 72 1C). The reaction mixture was the same as stated above.

Quantification of amplified DNA
Because the same amount of template DNA, 50 ng, was added to each PCR, the differing brightness of the bands on the gel was used as an indicator of the amount of F. prausnitzii in each patient. The intensity of the bands was measured using the software QUANTITY ONE (Bio-Rad) and quantified by comparison with known amounts of a 1-kb ladder (New England Biolabs). The experiment was repeated once. Samples that failed to yield a PCR product were reassayed a third or a fourth time, and the quality of the DNA was confirmed by amplification of bacterial 16S rRNA gene.

Serum CRP assay
Serum CRP levels were determined by the Clinical Biochemistry Department, Addenbrookes Hospital.

Statistical analyses
The results were assessed using nonparametric methods. The results of the estimated amount of each subgroup of F. prausnitzii DNA generated by PCR amplification for the three patient groups both pre-and post-treatment and the control group were analysed separately using Kruskal-Wallace one-way ANOVA and Dunn's post hoc all comparisons test. To gain an estimate of the effects of illness on the total amount of F. prausnitzii, the results for the controls and the three pretreatment samples were ranked first for the M21/2 subgroup (the larger PCR fragment) and then for the A2-165 subgroup (the smaller PCR product), and the two ranking values for each sample were added together to yield a combined rank value that was regarded as an estimate of the total F. prausnitzii content of the faeces. Post-treatment samples were analysed in the same way. These two new data sets were also analysed using the Kruskal-Wallace one-way ANOVA and Dunn's post hoc all comparisons. Paired data were analysed using the Wilcoxon signed rank-sum (WSRS) test. Kruskal-Wallace analyses were carried out initially using the MINITABS statistical package. Data were then reanalysed using the INSTAT (Graphpad) program, which was also used for Dunn's post hoc all comparisons (but completed only if the two-tailed P value was o 0.05 in the Kruskal-Wallace analysis), Mann-Whitney tests for unpaired data, the WSRS test for paired data and Spearman r to determine the level of correlation between parameters.

Results
The results are summarized in Tables 1 and 2. The original data on which these tables are based are available as Supporting Information, Tables S1-S3 in the online version of this article.

Subject characteristics
The patient groups consisted of 20 patients diagnosed as suffering from CD, 21 patients with IBS and 14 patients with UC. They were recruited at Addenbrookes Hospital, Cambridge, which functions as a tertiary referral centre for inflammatory bowel disease. The patients gave a faecal and a blood sample at the start of the treatment and again 2 weeks later. In addition, blood and faecal samples were obtained from 18 healthy subjects as controls.

Effects of treatment on the clinical state
Eighty-four per cent of the initial group of CD patients returned to follow-up clinics. The effects of treatment were beneficial, as evidenced by decreases in the Harvey and Bradshaw scores of individual patients (P o 0.0001; WSRS test). At presentation, all 15 patients whose serum was assayed had serum CRP levels above the upper limit of the reference range (6 mg L À1 ). Following treatment, there was a significant decline in the CRP levels (P o 0.05, WSRS test).

Quantification of F. prausnitzii-specific PCR products
Bacterial DNA was obtained from each faecal sample. The same amount of faecal DNA was used as a template in PCR reactions with the primer pair Fp.ID.F2 and Fp.ID.R2 to detect the presence of F. prausnitzii. Although the total quantity of F. prausnitzii PCR products and the ratio of the two subgroups varied considerably from sample to sample (Fig. 1), only a minority of samples failed to yield one or both F. prausnitzii PCR products. The sequences of PCR products were 100% identical to those of the published F. prausnitzii genomes A2-165 or M21/2, and to no other. When no F. prausnitzii product was detectable, a second amplification to produce an amplicon from the DNA encoding the 16S rRNA gene was always successful, demonstrating that the bacterial DNA extracted was of good quality.

Comparison of F. prausnitzii abundance in the faecal samples
There were no significant differences in the levels of the A2-165 subgroup of F. prausnitzii in the three treatment groups at initial presentation and in the control group (Kruskal-Wallace P = 0.248, adjusted for ties), although the levels for CD patients were the lowest of the four groups. There was a significant difference among the three post-treatment and control groups (Kruskal-Wallace P = 0.006, adjusted for ties), with the levels for the CD group being significantly lower than those of the control group (Dunn's post hoc test P o 0.05). For the CD group, the levels had decreased significantly post-treatment (WSRS test P = 0.0046). The levels of the M21/2 F. prausnitzii subgroup were analysed similarly. There was a significant difference among the three pretreatment patient groups and the control group (Kruskal-Wallace P = 0.029, adjusted for ties), although the post hoc test revealed no significant difference when individual groups were compared with each other, despite lower values for both the CD and the UC samples. However, for this F. prausnitzii subgroup, the levels in the CD patients were significantly lower than those in the control samples, as judged by the Mann-Whitney test (P = 0.02). Following treatment, there was also a significant difference in the levels of the M21/2 subgroup among the three patient groups and the control group (Kruskal-Wallace P = 0.034, adjusted for ties). The levels in the CD group were significantly lower than those in controls (Dunn's post hoc P o 0.05), although the difference between the pre-and the post-treatment levels (median before treatment 77 compared with 65.5 posttreatment) was not significantly different (WSRS P = 0.61). Thus, the levels of both F. prausnitzii subgroups for CD patients either decreased or remained low rather than increased following successful treatment with an enteral diet.
It is obvious from Fig. 1 that there were wide differences between the ratios of the levels of the two F. prausnitzii subgroups between patient samples. However, the overall total levels of each subgroup were similar, and so it was appropriate to lend equal weight to the two subgroups to compare the total levels of F. prausnitzii among the patient and control populations. The analyses revealed that the total levels of F. prausnitzii in CD patients were significantly lower than those in the other groups both before and after treatment (Kruskal-Wallace test P o 0.005 before treatment and o 0.001 after treatment; Dunn's comparisons for the CD group relative to the control group P o 0.05 before treatment; o 0.001 after treatment).  The same amount of template DNA was added to each PCR reaction, and so these differences indicate the relative amount of F. prausnitzii in each patient. Where a PCR product was present, the brightness of the band was measured using the software QUANTITY ONE (Bio-Rad) and converted to DNA amount by reference to a 1 kb ladder (New England Biolabs: right-hand lane).

Lack of a correlation between the levels of F. prausnitzii and clinical assessment
The levels of each subgroup of F. prausnitzii in CD patients were analysed for correlation with clinical status, as determined by the Harvey and Bradshaw scores, both before and after treatment. The data were assessed using the nonparametric Spearman r coefficient. As the r value for the A2-165 subgroup was 0.018 (P = 0.3) and r = 0.02 (P = 0.92) for the M2/21 group, neither set of F. prausnitzii levels correlated with clinical status.

Further analysis of data from samples from the UC patients
The data for UC patients before treatment appeared to show that the levels of the F. prausnitzii M2/21 subgroup were lower those than in the control group, but the statistical tests used failed to establish that the difference was significant. We therefore used the nonparametric Mann-Whitney test to compare the data for this UC subgroup with the controls. As the resulting P value, 0.087, tended towards, but did not confirm that the difference was significant, this is probably a type 2 error: more data are required for this patient group.

Discussion
Previous studies have established that Firmicutes are dominant among the diverse range of butyrate-producing bacteria in the human gut, with five species being especially abundant: Eubacterium rectal; Roseburia faecis; Eubacterium hallii; an unnamed but cultured species, SS2/1; and F. prausnitzii (Aminov et al., 2006;. The report that the levels of F. prausnitzii in the gut of patients with CD were lower than the levels in healthy subjects was based on analysis of tissue taken from patients undergoing elective surgery (Sokol et al., 2008). The current study used an alternative way of identifying bacteria present in the human gut, which was to investigate the bacterial component of faecal samples. However, analysis of faecal DNA in the present study only partially confirmed these earlier findings. Although F. prausnitzii, especially the M21/2 subgroup, were less abundant in CD patients at presentation than in the control group, contrary to expectation, successful treatment of the CD group resulted in a decrease rather than an increase in F. prausnitzii abundance. It is an open question as to whether the treatment received by CD patients before attendance at the Addenbrookes clinic might have resulted in the statistically significant decline in the faecal levels of F. prausnitzii: a recent study strongly suggests that this might be the case (Benus et al., 2010). However, the subsequent decline following ED treatment clearly excludes the possibility that recovery might be due to an increase in F. prausnitzii levels within the gut and the consequent increase in immuno-suppressive agents secreted by the bacteria (Sokol et al., 2008).
The causes of CD and other gastrointestinal diseases are clearly multifactorial, and it is therefore not surprising that the same is true of mechanisms that protect against disease. The previous report clearly established that the positive effect of F. prausnitzii in deterring the recurrence of CD following surgery was due to the production of an unidentified metabolite other than butyrate (Sokol et al., 2008). Bacteria generate butyrate during fermentation of lactate, but lactate is also the preferred energy source of many other bacteria, including propionibacteria, acetogens and sulphate-reducing bacteria (Macpherson et al., 1996). In the current study, patients were treated nutritionally with a minimal diet that would decrease the availability of lactate and hence also decrease the bacterial flora, including F. prausnitzii (Chapman et al., 1994;Costello et al., 1994;King et al., 1997;Whelan et al., 2005;Johnson et al., 2006). It is known that faecal bacterial dry weight in healthy volunteers fed ED declines drastically (Costello et al., 1994). A positive outcome following treatment might therefore be due to a greater reduction in the production of toxic, inflammatory metabolites by other bacteria relative to any negative effects of a decreased population of butyrate-producing bacteria. For example, decreased availability of lactate would select against sulphate-reducing bacteria such as Desulfovibrio piger, which is abundant in human faeces and the major source of sulphide, which inhibits butyrate oxidation by colonocytes (Pitcher & Cummings, 1996;Roediger et al., 1997;Willis et al., 1997;Levine et al., 1998;Loubinoux et al., 2002;Duncan et al., 2004;Fite et al., 2004;Marquet et al., 2009).
The different results reported in this and the previous study may have arisen from the differences in the material assayed. Normally produced stools, which were used in this study, represent bacterial populations chiefly along the colon, whereas the surgically resected material in the earlier study were from higher up the alimentary canal. Bacterial populations in the two sites may well differ. Furthermore, patients awaiting surgery may have been given enteral or parenteral feeds to improve their state of nutrition or fasted preoperatively. If the organism is dependent on dietary complex carbohydrate, these also could have been factors leading to the low numbers of F. prausnitzii reported.
In conclusion, the clinical improvement in our patients with CD was associated with a decline in the numbers of F. prausnitzii, but this does not mean that the organism can be considered to be a pathogen responsible for producing CD, as it was equally frequent in our healthy volunteers. It is more likely to be an innocent bystander, whose numbers decline with ED feeding, which exerts its effects in CD through its action on the gut microbial communities (Hunter, 1991). We therefore attribute the decrease in the faecal F. prausnitzii levels in CD patients following ED to the fact that this diet reduces the undigested food residues required to sustain normal levels of flora in the lower gut, particularly bacteria such as F. prausnitzii. Such a change in flora was associated with the marked improvement seen in patients' symptoms and the levels of CRP. We suggest, therefore, that this reduction in gut flora was chiefly responsible for the patients' improvement.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Table S1. Quantified yields of the two PCR products from the different groups of Faecalibacterium prausnitzii strains in individual patient samples. Table S2. Clinical assessment of the patient group with Crohn's disease. Table S3. Clinical assessment of the patient group with ulcerative colitis.
Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.