A variety of human and animal data support the hypothesis that Crohn's disease (CD) is a result of immune responses to the fecal microbiota in genetically susceptible individuals. Such immune responses may alter the microbiota, resulting in dysbiosis that may further promote inflammatory responses. Such dysbiosis in Crohn's patients has been better defined in recent years with the advent of nonculturable techniques, such as 16s RNA sequencing for identification of species within the fecal microbiota. The dysbiosis in CD has been characterized by depletion of commensal bacteria including members of Firmicutes and Bacteroidetes and a decreased abundance of class Clostridia including F. prausnitzii, as well as an increase in Proteobacteria, although the actual significance of these species-based alterations is only correlative and therefore is functionally unknown.1,2 Numerous clinical studies have attempted to modulate the fecal microbiome to decrease the inflammatory immune response using prebiotic, probiotic, and antimicrobial therapies. The results of these trials, however, have been mixed.
The possibility of modifying the human microbiome to alter dysbiosis through fecal microbial transplant (FMT) was first reported by Eiseman et al3 in 1958 in the treatment of fulminant pseudomembranous enterocolitis. Since then, many case reports and case series of fecal transplantation were noted in the literature for treatment for Clostridium difficile, constipation, irritable bowel syndrome, and inflammatory bowel disease (IBD), with efficacy only proven for C. difficile toxin-induced recurrent colitis. Fecal transplantation has been described in patients as young as 2 years of age to patients over 90 years of age. Given the efficacy of this therapy in C. difficile infections, the American Gastroenterology Association recently wrote a position paper on the use of fecal transplantation. In this article, preparation, dosage, patient, and donor workups are all reviewed.4 To confirm the safety and potential efficacy of FMT in Crohn's patients, we performed a prospective study of FMT in pediatric CD.
MATERIALS AND METHODS
This is a single-center open-label study designed to determine tolerability, preliminary safety, and potential efficacy in pediatric patients with CD. Nine patients with CD with mild-to-moderate disease symptoms as defined by Pediatric Crohn's Disease Activity Index (PCDAI)5 between 10 and 29 and aged 12 to 21 years were enrolled into this study. Each participant was followed in the study for approximately 12 weeks.
The protocol was approved by the Institutional Review Board of Seattle Children's Hospital. All patients/participants provided written informed consent or assent. Approval from the FDA (investigational new drug number 14942) was obtained. The study was registered with Clinical Trials.gov (number: NCT01757964). Study participants were recruited from Seattle Children's Hospital outpatient gastroenterology clinics.
All patients had a diagnosis of CD made by a primary gastroenterologist based on history, physical examination, laboratory/radiological studies, and gastrointestinal histology. All patients had mild-to-moderate symptoms with a PCDAI score between 10 and 29. Parent/guardian and child consent or assent was obtained. Patient medication for IBD could not have changed for at least 1 month before FMT. Patient exclusion criteria included active or history of intraa-bdominal abscess, intra-abdominal fistula, stricturing CD, or other serious systemic diseases. None of the patients received tumor necrosis factor inhibitors before transplant. Patients were allowed to maintain the use of other IBD medications including immunomodulators during the study.
Study participants had laboratory tests including complete blood count with differential and platelets, C-reactive protein (CRP), albumin, stool studies for C. difficile, bacterial culture, ova, and parasite. The American Association of Blood Banks Donor History Questionnaire was used to evaluate study participant donors. Study participant donor laboratory studies included hepatitis A IgG and IgM, hepatitis B serum antigen, antibody, and core antibody, hepatitis C IgG, human immunodeficiency virus 1 and 2 IgG, rapid plasma reagin and Epstein–Barr viral IgG and IgM, cytomegalovirus IgG and IgM, as well as stool testing for C. difficile, bacterial culture, and examination of stool for ova and parasites. Study participant donors were not allowed to have antibiotics 3 months before procedure. Each patient had a single donor for transplant. The donor for each patient was one of their parents.
Study participant recipients received premedication before fecal transplant, which included rifaximin 200 mg 3 times daily for 3 days until the evening before procedure. Study participant recipients also received omeprazole (1 mg/kg orally) on the day before and morning of the procedure. Transplant recipients also received 1 capful of MiraLAX in 8 oz of water 3 times a day for 2 days. A nasogastric (NG) tube was placed for transplant, and location was confirmed by x-ray. Approximately 30 g of donor stool was mixed with 100 to 200 mL of normal saline and blended with a commercial blender (Hamilton Beach Personal Blender, Southern Pines, NC) at low speed for 2 to 4 minutes until a homogenous texture was achieved. The stool was then filtered twice using 4 × 4 gauze. Infusion was slowly administered through NG tube over a 3-minute period. The NG tube was flushed with 15 mL of normal saline over 1 minute. After 15 minutes, the NG tube was removed.
Study participant FMT recipients were called 2 days after transplantation and had clinical follow-up at 2, 6, and 12 weeks. Standardized questionnaires and the PCDAI were completed during each study visit. The PCDAI uses the patient history, laboratory values, and the physical examination to create a validated score of that individual's disease activity. A PCDAI score of <10 denotes remission, 10 to 29 mild disease, and ≥30 moderate-to-severe disease activity. Study participant recipients/patient's families were provided diary cards to assess possible side effects.
Descriptive statistics were prepared for all data including frequencies and percentages for categorical variables (e.g., gender, disease location) and means, SDs, quartiles, and ranges for quantitative variables (e.g., CRP, PCDAI). When laboratory measurements were above or below the threshold for measurement, the threshold was imputed before computing descriptive statistics.
Method Section for Microbiome
Total genomic DNA was extracted from stool using the PowerSoil DNA Isolation Kit (MoBio, Carlsbad, CA). The protocol was customized to include 2 incubation steps (65°C and 95°C for 10 min each) after the addition of lysis buffer (Solution C1). Additionally, the provided garnet beads for mechanical disruption were substituted for 0.5 g of 0.1-mm diameter Zirconia/Silica beads (BioSpec Products, Bartlesville, OK). Final DNA yield was quantified using the Qubit 2.0 Fluorometer (Life Technologies, Carlsbad, CA) before NGS library construction.
Sequencing was performed on either the Illumina HiSeq 2000 or MiSeq platform. Sequencing libraries were constructed from genomic DNA using Illumina's Nextera technology (Illumina, Inc, San Diego, CA). Briefly, DNA preparations were simultaneously fragmented and tagged with adapter oligomers. A limited-cycle PCR reaction amplified all tagged fragments and added (1) index sequences (the dual indexing strategy uses two 8-base indices) to allow demultiplexing of sequence reads for pooled samples and (2) sequencing primer sequences. After PCR enrichment, libraries were denatured and hybridized through DNA/DNA binding of adaptors to existing features on a glass flow cell compatible with the Illumina sequencers. Sequencing was performed using well-established ultrahigh throughput methods. The HiSeq 2000 produced approximately 200 million pairs of 93 bp reads per lane, and we generated 25 to 30 million raw read pairs per sample using 7-8-plex pools. The single lane of the MiSeq generated 15 to 20 million raw pairs of 150 bp reads per sample.
Human DNA sequence was identified and removed using BMTagger19 (Rotmistrovsky K, Agarwala R. BMTagger: Best Match Tagger for removing human reads from metagenomics datasets, unpublished data, 2011) with the Hg-19 Homo sapiens reference genome. Duplicate reads were marked and removed using EstimateLibraryComplexity, part of the Picard tool package (http://picard.sourceforge.net/index.shtml). Sequence reads with ambiguous bases were trimmed from each end. Reads with Phred quality scores less than 6 over the first 80 (HiSeq) or 120 (MiSeq) base of each read and reads shorter than 80 (HiSeq) or 120 (MiSeq) base after trimming were removed. Similarity scores were calculated using Compareads using 2 km of length 30 nucleotides.6 Sequences for all samples were limited to between 20 and 30 million reads to maintain comparable sample sizes. A minimum of 20 million reads was required obtain sufficient read depth coverage of the samples (Brittnacher, et al., unpublished data, 2014). Relative species abundance was calculated from the sequence reads using MetaPhlAn.7
A total of 14 families were screened for the study. Five families were deemed unsuitable for transplant because of exclusion criteria including no suitable donor. Nine Crohn's patients received fecal transplantation. The stool donor for each patient was their mother except for patient 10 and 13 where fathers were the stool donors. Age of participants was 16.2 ± 2.9 (range, 12–19) years. Five were male. Average disease duration before FMT was 3.9 ± 1.8 (range, 0–7) years. Macroscopic disease at the time of diagnosis was in the stomach/duodenum in 6 patients, colon in 7 patients, and terminal ileum in 6 patients. At the time of transplant, 4 individuals were on methotrexate, 1 individual was on azathioprine, 1 individual was on mercaptopurine, 3 individuals on mesalamine therapy, and 1 newly diagnosed individual was on no medication at the time of FMT (Table 1).
All reported adverse events were graded as mild except for 1 individual who reported moderate abdominal pain after FMT. After the FMT, 5 patients complained of abdominal pain likely related to FMT (4/5 mild abdominal pain), 5 had mild bloating likely related to FMT, and 4 had diarrhea likely related to the pretreatment for the FMT or the FMT. Abdominal pain, bloating, and diarrhea returned to baseline or improved within 48 hours of FMT. One individual had mild post-FMT flatulence lasting 1 day. One individual had a mild stuffy nose after FMT, which was not likely related to the procedure but more likely reflected allergic responses. Three individual had side effects likely related to the NG tube, 1 with rhinorrhea lasting 3 days after FMT, and 2 individuals with sore throats lasting 1 day after FMT.
Two weeks after FMT, 7 of the 9 patients were in clinical remission based on PCDAI scoring. At 6 and 12 weeks, 5 of 9 patients who did not receive additional therapy were still in remission. Two individuals received additional standard medical therapies before the end of the study. Patient 15 began metronidazole before the 6-week follow-up and then infliximab after the 6-week follow-up. Patient 13 began prednisone and methotrexate after the 6-week follow-up (Table 2). Two other individuals began standard medical therapy after the 12-week follow-up because of a flare in symptoms. Mean PCDAI at baseline was 19.7 ± 7.2, at 2 weeks after FMT was 6.4 ± 6.6, and at 6 weeks after FMT was 8.6 ± 4.9. The 6- and 12-week mean PCDAI of individuals not receiving additional medical intervention were 8.8 ± 5.2 and 11.1 ± 7.5, respectively. All but 1 patient had improvement/normalization in their CRP at the week-2 follow-up. The mean CRP levels decreased from 2.4 ± 1.2 mg/dL at baseline to 1.5 ± 0.6 mg/dL at the 2-week post-FMT visit. At 6 weeks and 12 weeks post-FMT in those individuals who did not start on additional medical therapy, the mean CRP still remained below baseline level at 2.0 ± 1.2 and 2.3 ± 2.3 mg/dL, respectively. No significant changes were noted in albumin levels or hematocrit with FMT (Table 2).
Stool calprotectin decreased or remained unchanged for all patients except for one whose level increased slightly at the 2-week follow-up. The mean baseline calprotectin for all patients enrolled was 936 ± 782 mg/L with a mean level of 671 ± 474 mg/L at the 2-week follow-up. Although initial improvement was seen for most patients in calprotectin, the levels rose for most patients by the 12-week follow-up (Table 2).
Fecal microbiome similarity to donor, pre-FMT had a mean of 41.7% ± 16.1% (SD) and ranged from 13% to 69% (Fig. 1A). Similarity to self, pre-FMT (Fig. 1B) was also calculated for all samples. Similarity of donor stool was measured by the relative change in similarity to donor (Fig. 1C; method paper currently submitted). The similarity score ranged from −15% to 46% (mean ± SD, 15.7 ± 18.9,) with individuals following 1 of 3 paths: no similarity (n = 2), gradual similarity (n = 4), and immediate similarity (n = 3). Patients 2 and 18 did not appear to engraft, whereas others (patients 1, 8, and 10) had a gradual increase in donor similarity over the 12-week period (Table 2). The third group (patients 6 and 7) had very quick similarity by the second week analysis. Similarity of donor stool was evident for the entire 12-week period for those who had initially engrafted. It is unclear at this stage if degree of similarity correlates with different clinical responses in the fecal microbiome to FMT, although of the 2 patients with no similarity, one of these patients had no clinical response to FMT, whereas the other had only a modest clinical response (Table 2).
Using MetaPhlAN analysis of the metagenomics data, a total complement of 116 species were identified across all samples of which 107 (92%) were found in the donors. About one-third (11 of 30) of the 30 most abundant species in the donors were found in all 9 donors (see Fig., Supplemental Digital Content 1, http://links.lww.com/IBD/A745). In contrast, only 3 of the 30 most abundant species in the patients, pre-FMT was found in all 9 patients (see Fig., Supplemental Digital Content 2, http://links.lww.com/IBD/A746). Thus, greater homogeneity was observed for the donors in relation to the patients for the most abundant species. Only 1 species, Ruminococcus torques, was found in all donor and patient samples, pre-FMT. To determine which species were potentially transplanted during FMT, we identified 31 species that were not detected in the patient baseline samples but were found after FMT at 2 weeks (Fig. 2). These species were all detected in the patient's respective donor. The species that were found in more than 2 of the 9 patients were Bilophila wadsworthia (5 of 9), Odoribacter splanchnicus and Bacteroides caccae (4 of 9), and Alistipes shahii, A. putredinis, and Parabacteroides merdae (3 of 9). Although there are many factors independent of transplant that could explain the presence of these newly detected species, we found that the percentage of these 31 species detected post-FMT for each patient was correlated with the similarity score at 2 weeks (Spearman's r = 0.9). For example, the percentage of previously undetected species in patients 6, 7, and 15 at 2 weeks was 39%, 45%, and 26%, and their similarity scores at 2 weeks were 41%, 46%, and 22%, respectively. In contrast, no new species were detected for patient 18 whose similarity score was 0%. High relative abundance of a species was not a determinant of which species were possibly transplanted. In patient 6, Ruminococcus bromii contributed toward a large fraction of the increase in similarity to donor at 2 weeks post-FMT (see Fig., Supplemental Digital Content 3, http://links.lww.com/IBD/A747, page 3). However, R. bromii was not detected after FMT in patients 1, 2, 10, and 18 even for donor abundances between 10% and 30% (see Fig., Supplemental Digital Content 4, http://links.lww.com/IBD/A748, page 8).
On evaluation of specific individual's microbial similarity to donor before transplant, 2 patients stand apart from the overall group. The pre-FMT microbiome of patient 1 was the least similar to donor parental microbiome at 13%. His clinical course was also one of the best with a decrease in PCDAI from 27.5 to 7.5 and a decrease in CRP from 3.1 to 1.1 mg/dL. In contrast, patient 18 had the greatest similarity to donor, pre-FMT at 69%. This patient's clinical course did not seem to be significantly altered; clinical remission with FMT was not achieved, and additional medical therapy was required by the end of the study. The patient's PCDAI went from 27.5 to 22.5 with CRP decreasing from 2.6 to 1.5 mg/dL. Interestingly, 2 patients had significant clinical deterioration over the course of the study. The longitudinal analysis of fecal metagenomes during this period allowed us to observe a dramatic increase in the relative abundance of Escheria coli during a clinical disease flare. These data are highly correlative with a bloom of E. coli in response to inflammation, or an expansion of E. coli could contribute to disease outcome. Therefore, an examination of the relative abundance of E. coli in stool samples before and after FMT would suggest that a relative increase in the amount of E. coli in stool samples is associated with increased inflammation. For patient 18 who was a nonresponder to FMT, there was an increase from baseline after FMT. For those who responded initially but then had an increase in disease activity, E. coli abundance appears associated with increased calprotectin, thus worsening inflammation.
The published experience of fecal microbial transplantation for active CD is limited. To date, only 2 retrospective case reports exist in which fecal transplant has been used as a potential treatment for CD. Borody and Grehan each described 1 individual with CD who had clinical benefit from FMT.8,9 The results of this prospective study show that FMT in this small cohort of pediatric patients with CD was safe and well tolerated. Both clinical and laboratory improvements were seen in the majority of patients. This study provides further evidence that the fecal microbiota likely plays an important role in the pathogenesis of CD and indicates that FMT merits further study as a potential therapy for CD.
Current therapies for CD focus on suppressing the immune system. In contrast, FMT focuses on a possible trigger of the immune dysregulation, the fecal microbiota. There have been many clinical observations implicating the fecal microbiota as a contributing agent in CD pathology. This includes the clinical efficacy of antibiotic therapy, clinical improvement with diversion of the fecal stream on distal disease activity, and serum reactivity toward microbial antigens including anti-Saccharomyces cerevisiae antibody and outer membrane protein C of E. coli (Omp-C).10–12 An additional factor known to impact the development of IBD and the fecal microbiota is antibiotic use. The development of pediatric IBD has been closely associated with antibiotic exposure early in life. Kronman et al reported an 84% relative risk increase in development of IBD with antibiotic use. Exposure to antibiotics throughout childhood was associated with developing IBD but decreased with increasing age of exposure.13 This suggests that immune dysregulation may come from a disruption of immune tolerance during development as a result of alteration of the microbiota by antibiotic use.14,15
The goal of FMT in CD is to alter the fecal microbiome by decreasing potential “dysbiotic” bacterium that may be more proinflammatory. With that in mind, rifaximin, a nonabsorbed antibacterial agent was used before FMT to decrease the endogenous dysbiotic bacterial load in the transplant recipient with the goal of allowing a biological niche for the transplanted stool to survive in. Although there is uncertainty regarding the specific clinical implications of dysbiosis and whether the changes seen are a cause or effect of the disease itself, studies have noted differences within the fecal microbiota of patients with CD from healthy controls. This divergence from healthy individual's fecal microbiota is characterized by a decrease in commensal bacteria including members of Firmicutes and Bacteroides as well as a relative increase in proinflammatory bacteria such as Enterobacteriaceae.1,2 In addition, a decrease in butyrate-producing bacterium that is important in intestinal health has been seen in patients with CD.16 Fluorescent in situ hybridization analysis has shown bacteria penetrating the mucus layer in 25% of colonic and 55% of ileal mucosal biopsies of patients with CD as compared with none in controls.17 Although, as a community, we often describe the microbial changes seen in CD as a dysbiosis, it is important to acknowledge our limited understanding of the fecal microbiome and that some of the changes seen may not represent a true functional dysbiosis. Our microbial analysis confirms the changes seen in other studies. One unique aspect of this study is that our longitudinal analysis allowed us to observe a bloom of E. coli as associated with an increase in clinical symptoms and inflammatory markers. This indicates how rapidly a change in clinical status by an increase in intestinal inflammation can be associated with a bloom of bacteria that is both a commensal but well known as a pathogen from horizontally acquired virulence factors and in the setting of compromise of the intestinal barrier function. Though E. coli may not be the ultimate cause of CD, it is interesting to speculate that its expansion and invasion could contribute to symptoms and disease.
There are many intriguing issues related to FMT that arise from this study. There was a significant difference in clinical outcome between the patient with the least and most microbial similarity between recipient and donor. This could indicate that the more divergent a Crohn's patient is from his donor the more the potential benefit of transplantation. The diagnosis of dysbiosis by species analysis is speculative given the huge diversity in human microbiomes and the influence of diet and genetics on the microbiome content. Therefore, the nonspecies-based analysis we performed coupled with the greater similarity in heritable factors and diet between parents and child may suggest who has a greater likelihood of benefit from FMT. Another possible predictor of disease activity and duration of efficacy seems to be the appearance or resurgence of E. coli. We notice a trend of increasing calprotectins with an increase in E. coli abundance. Although this finding may be a helpful predictor of efficacy of therapy, there is no clear casual affect. However, in patients with significant dysbiosis with E. coli, therapy targeted at its suppression followed by FMT could be another potential therapeutic trial in the future.
Although there were clinical and laboratory improvements in the majority of patients, there are number of significant limitations to this study. As an open-labeled study, recruited patients and parents had a strong personal belief that FMT would improve symptoms. It cannot be excluded that participant bias could account for some of the effect seen in the PCDAI. In addition, the relative effect of pretreatment of patients with rifaximin and MiraLAX before FMT could have accounted for some of the benefits seen and the changes observed in the microbiome. Finally, the small sample size for this study limits the precision of estimated effects of the FMT within our Crohn's patients. Despite these limitations, this study suggests a further link between the fecal microbiota and CD and suggests that further study of FMT as a potential therapeutic option using controlled clinical trials with analysis of transplant similarity is warranted.
1. Frank DN, St Amand AL, Feldman RA, et al.. Molecular-phylogenetic characterization of microbial community imbalances in human inflammatory bowel diseases. Proc Natl Acad Sci U S A. 2007;104:13780–13785.
2. Fujimoto T, Imaeda H, Takahashi K, et al.. Decreased abundance of Faecalibacterium prausnitzii in the gut microbiota of Crohn's disease. J Gastroenterol Hepatol. 2013;28:613–619.
3. Eiseman B, Silen W, Bascom GS, et al.. Fecal enema as an adjunct in the treatment of pseudomembranous enterocolitis. Surgery. 1958;44:854–859.
4. Bakken JS, Borody T, Brandt LJ, et al.. Treating Clostridium difficile infection with fecal microbiota transplantation. Clin Gastroenterol Hepatol. 2011;9:1044–1049.
5. Hyams J, Markowitz J, Otley A, et al.. Evaluation of the pediatric crohn disease activity index: a prospective multicenter experience. J Pediatr Gastroenterol Nutr. 2005;41:416–421.
6. Maillet N, Lemaitre C, Chikhi R, et al.. Compareads: comparing huge metagenomic experiments. BMC Bioinformatics. 2012;13(suppl 19):S10.
7. Segata N, Waldron L, Ballarini A, et al.. Metagenomic microbial community profiling using unique clade-specific marker genes. Nat Methods. 2012;9:811–814.
8. Borody TJ, George L, Andrews P, et al.. Bowel-flora alteration: a potential cure for inflammatory bowel disease and irritable bowel syndrome? Med J Aust. 1989;150:604.
9. Grehan MJ, Borody TJ, Leis SM, et al.. Durable alteration of the colonic microbiota by the administration of donor fecal flora. J Clin Gastroenterol. 2010;44:551–561.
10. Zholudev A, Zurakowski D, Young W, et al.. Serologic testing with ANCA, ASCA, and anti-OmpC in children and young adults with Crohn's disease and ulcerative colitis: diagnostic value and correlation with disease phenotype. Am J Gastroenterol. 2004;99:2235–2241.
11. Sartor RB. Therapeutic manipulation of the enteric microflora in inflammatory bowel diseases: antibiotics, probiotics, and prebiotics. Gastroenterology. 2004;126:1620–1633.
12. Winslet MC, Andrews H, Allan RN, et al.. Fecal diversion in the management of Crohn's disease of the colon. Dis Colon Rectum. 1993;36:757–762.
13. Kronman MP, Zaoutis TE, Haynes K, et al.. Antibiotic exposure and IBD development among children: a population-based cohort study. Pediatrics. 2012;130:e794–803.
14. Shaw SY, Blanchard JF, Bernstein CN. Association between the use of antibiotics and new diagnoses of Crohn's disease and ulcerative colitis. Am J Gastroenterol. 2011;106:2133–2142.
15. Virta L, Auvinen A, Helenius H, et al.. Association of repeated exposure to antibiotics with the development of pediatric Crohn's disease–a nationwide, register-based finnish case-control study. Am J Epidemiol. 2012;175:775–784.
16. Wang W, Chen L, Zhou R, et al.. Increased proportions of bifidobacterium and the lactobacillus group and loss of butyrate-producing bacteria in inflammatory bowel disease. J Clin Microbiol. 2014;52:398–406.
17. Kleessen B, Kroesen AJ, Buhr HJ, et al.. Mucosal and invading bacteria in patients with inflammatory bowel disease compared with controls. Scand J Gastroenterol. 2002;37:1034–1041.