Applying fecal microbiota transplantation (FMT) to treat recurrent Clostridium difficile infections (rCDI) in children

Overall study design

This is an observational epidemiology study to look at children who failed multiple courses of standard antibiotic therapy for rCDI and elected to undergo FMT. Prior to sampling, this study was approved by the institutional review board of Children’s Healthcare of Atlanta and Morehouse School of Medicine.

Study participants (n = 15 pediatric patients) were screened and recruited from a pediatric hospital (Children’s Healthcare of Atlanta Hospital at Scottish Rite) and a private pediatric gastroenterology clinic (Children’s Center for Digestive Health Care, LLC) in Atlanta, Georgia. All study participants were children (age range 21 months to 18 years) who presented with symptoms consistent with rCDI, as determined from examination by the same pediatric gastroenterologist. Patients were eligible for the FMT procedure if they fit the study definition of rCDI: ongoing diarrhea, a positive C. difficile fecal specimen test (polymerase chain reaction assay to detect C. difficile toxin genes), and history of failed CDI treatment with antibiotics for a minimum of two prior CDI episodes, of which, at least one episode was treated with a course of oral vancomycin within the previous eight weeks (Cohen et al., 2010). All FMT procedures were performed via either colonoscopy or nasojejunal tube by the same pediatric gastroenterologist. Previous episodes of CDI were recorded if patient had a documented positive C. difficile toxin PCR test and a diagnosis of CDI in his/her medical records. Informed consent (or assent when age appropriate) was obtained from a patient’s parent or legal guardian prior to the FMT procedure. A survey questionnaire was administered in a personal interview from study staff to record participant demographic information (age, gender, race/ethnicity), underlying chronic health conditions, prior antibiotic use within 12 months of the time of enrollment, and CDI related symptoms. The following information was also collected from each participant’s medical records: nature of prior episodes of CDI, dates of positive PCR C. difficile toxin tests, and provider notes documenting history of infections that required antibiotics and specific symptoms consistent with CDI. Five donor samples were ordered from OpenBiome™ (Medford, MA, USA) to be used for the FMT procedure. All donors were non-family related and underwent laboratory screening by Openbiome™ (Osman et al., 2016; Osman et al., 2016) for infectious risk factors (including Multi-drug resistant organisms (MDRO’s), serologic testing, and fecal specimen testing) and potential microbiome-mediated conditions.

FMT procedure

FMT procedure was performed on subjects via colonoscopy (n = 14) and nasojejunal tube (n = 1). Prior to FMT, all patients undergoing a colonoscopy completed a standard bowel prep using Polyethylene Glycol 3350. Vancomycin was discontinued at least 48 h prior to the FMT. Colonoscopies were performed under general anesthesia or under conscious sedation. Up to 240 ml of donor fecal specimen was instilled into the terminal ileum and cecum during colonoscopy in 14 of 15 patients. In one patient, who had undergone colonoscopy due to an underlying disease, 60 cc of donor fecal specimen was instilled via nasojejunal tube into the proximal jejunum.

Follow-up

Follow-up telephone calls were made by research staff after 24 h and one-week post-FMT procedure. Survey questionnaires were administered by research staff at one month and ≥3 months post-FMT time points, alongside scheduled evaluation by single pediatric gastroenterologist. Responses to questions about study participant’s symptoms, including fever, vomiting, abdominal pain and distention, evidence of allergic reaction or bloody fecal specimens were recorded in a relational database. If patients were not available to respond at the three-month follow-up visit, attempts to collect a response were performed for up to 15 months after the procedure. (Although fecal samples were collected outside of the three-month study period, patients were not clinically re-evaluated for evidence of rCDI beyond three months of the date of fecal microbiota transplantation.)

Fecal specimen collection

Three fecal specimen samples were collected for each study participant: (1) from the donor—used for transplant, (2) from the study participant before transplant (pre-FMT), (3) from the study participant following the FMT procedure (at least 3 months after FMT procedure), with the latter used to analyze the effect of FMT on gut microbiome composition. Fecal samples were de-identified and stored in a −80 °C freezer until DNA extraction.

Fecal specimen DNA extraction

DNA was extracted from fecal specimen samples using the MO BIO PowerSoil DNA extraction kit (MOBIO, Carlsbad, CA, USA) following manufacturer instructions, with modifications as recommended by the Human Microbiome Project (HMP) Manual of Procedures (version 12.0). Briefly, before step 2 of the standard protocol, the PowerBead tubes containing samples were heated at 65 °C for 10 min, followed by a second incubation at 95 °C for 10 min to improve cell lysis. The manufacturer protocol was followed for the remainder of the procedure, except for step 12, during which the tubes were centrifuged for 2 min. DNA was extracted from triplicate subsamples for each fecal specimen sample, with the extracted DNA then pooled for quantification and processing. DNA concentrations were measured using the Qubit 2.0 DNA quantification system (ThermoFisher Scientific, Waltham, MA, USA).

16S rRNA gene PCR and amplicon sequencing

Microbiome taxonomic composition in donor, pre-FMT, and post-FMT fecal specimen was characterized by Illumina sequencing of PCR amplicons encompassing the V4 hypervariable region of the 16S rRNA gene. Amplicons were synthesized using Platinum PCR Supermix (ThermoFisher Scientific, Waltham, MA, USA) with V4-specific primers F515 and R806 (Caporaso et al., 2010), and uniform amounts of DNA template per reaction. Both forward and reverse primers were barcoded and appended with Illumina-specific adapters, as per instructions by Kozich et al. (2013). Thermal cycler conditions were as follows: initial denaturation at 94 °C (3 min), followed by 30 cycles of denaturation at 94 °C (45 s), primer annealing at 55 °C (45 s) and primer extension at 72 °C (90 s), and a final extension at 72 °C for 10 min. Amplicons were verified on agarose gel electrophoresis for size (∼400 bp) and purified using Diffinity RapidTip2 pipet tips (Diffinity Genomics, West Chester, PA, USA). Barcoded and Illumina adaptor-appended amplicons for each sample were pooled at equimolar concentrations and sequenced on an Illumina MiSeq using a 500 cycle v2 reagent kit (250 bp paired end) with 15% PhiX genomic library addition to increase read diversity.

Statistical and diversity analyses

Demultiplexed amplicon read pairs were quality trimmed with Trim Galore! (Babraham Bioinformatics, Cambridge, UK), using a base Phred33 score threshold of Q25 and a minimum length cutoff of 100 bp. High quality paired reads were then merged using the software FLASH (Magoč & Salzberg, 2011). Merged reads were analyzed using QIIME v1.9.0 (Caporaso et al., 2010), according to standard protocols. Briefly, reads were screened for chimeras using QIIME’s identify_chimeric_seqs.py script with Usearch61 (Edgar, 2010). Non-chimeric sequences were clustered into Operational Taxonomic Units (OTUs) at 97% sequence similarity using UCLUST (Edgar, 2010) based on open-reference OTU picking with the script pick_open_reference_otus.py. Taxonomy was assigned to representative OTUs from each cluster using the Greengenes database (Aug 2013 release) (DeSantis et al., 2006).

R (R Core Team, 2017) packages Phyloseq (McMurdie & Holmes, 2013) and Vegan (Jari Oksanen et al., 2017) were used to measure alpha and beta diversity, respectively. Alpha diversity was estimated by the number of observed unique OTUs and the Shannon diversity index at an even sampling depth of 27,700 sequences. OTUs observed across at least 70% samples were used for identifying shared OTUs between groups. Microbiome compositional differences among samples were assessed using non-metric multidimensional scaling (NMDS) based on Bray-Curtis distances at OTU level.

The significance of the effect of group status (donor, pre-FMT, post-FMT) on microbiome composition was assessed using the envfit function in Vegan, with 999 permutations. Wilcoxon signed rank test with Benjamini–Hochberg correction for multiple hypothesis testing was used to compare the taxonomic differences between pre- and post- FMT, pre- FMT and donor as well as post-FMT and donor. The differences between IBD and non-IBD were compared using Wilcoxon two sample test for demographic information (age, gender, race/ethnicity), underlying chronic health conditions, prior antibiotic use within 12 months of the time of enrollment, and CDI related symptoms and participant’s medical records: Nature of prior episodes of CDI, dates of positive PCR, C. difficile toxins, and provider notes documenting history of infections which required antibiotics and specific symptoms consistent with CDI. R and SAS 9.4 were used to perform all the statistical tests and adjusted p-value < 0.05 (after multiple hypothesis correction) was reported as statistically significant.

Patient enrollment/population characteristics

Fifteen pediatric study participants (eight female, seven male) were enrolled in this study. Of the potentially 45 specimens, only 40 fecal bio specimens were collected from these 15 patients. (Three post-FMT fecal samples were lost to follow up (LTFU), and two Pre-FMT samples did not yield enough DNA for analysis.) The median age of patients was eight years (range: 21 months to 18 years) (Table 1). Patients had an average of three episodes of CDI prior to FMT (range: two to five CDI episodes.). On average, post-FMT fecal samples were collected 6 months after the FMT procedure (range: three to 14 months) (Fig. 1).

Clinical outcomes

All 15 patients tolerated the FMT procedure without complications immediately post procedure. For the 12 patients who completed their three-month follow-up appointments, all experienced clinical resolution of CDI and had no recurrent episodes within the duration of the follow-up period (three months). Abdominal pain was reported by five patients (none had underlying IBD) at the three-month visit; however, the treating pediatric gastroenterologist attributed the pain to other causes, including functional pain and irritable bowel syndrome not related to CDI. All six patients with underlying ulcerative colitis (n = 5) or Crohn’s disease (n = 1) reported hematochezia and ongoing abdominal distension beyond three months of their FMT date.

Microbiome analysis—overall bacterial community diversity and richness

FMT significantly altered gut microbiome composition in all the participants. Alpha diversity (within sample diversity), measured as both observed OTU richness and Shannon diversity, differed between all compared FMT groups, with diversity in the donor sample higher compared to the matched pre- and post-FMT samples collected from recipients; pre-FMT fecal samples had the lowest diversity. There was no significant difference in observed OTU richness between the five donor samples (Observed OTU p-value = 0.07894; Shannon diversity p-value 0.04306). Post-FMT microbiomes, although more diverse compared to pre-FMT microbiomes, were less diverse than donor microbiomes (Fig. 2A). A significant difference (Kruskal-Wallis test, p-value < 0.001) was also observed in shared OTU’s between all three groups (donor, recipient pre-FMT, and recipient post-FMT), where within each comparison, donor vs. post-FMT profiles showed the smallest variance (calculated using Wilcoxon signed rank test), while pre- vs post-FMT profiles had the greatest variance.

FMT status (donor, recipient pre-FMT, and recipient post-FMT) also significantly influenced microbiome beta diversity (between sample diversity), as measured by Bray-Curtis distance-based NMDS ordinations (r2 = 0.65, p-value < 0.001; envfit function in Vegan), with post-FMT microbiome composition being more like that of donor than that of the pre-FMT group (Fig. 2B). Together, these results indicate a positive shift in microbial community composition in response to FMT, suggesting a restructuring to a non-disease state akin to the donor.

Microbiome analysis—taxon level changes

The relative representation of major bacterial phyla of interest varied substantially in response to FMT (Fig. 3). Notably, the average ratio of Bacteroidetes to Firmicutes (B:F), the predominant phylum-level indicator of a ‘normal/healthy’ gut microbiome, increased over 7-fold from pre-FMT (average B:F of 0.15) to post-FMT microbiomes (average B:F: 1.12), with the recipient post-FMT B:F approaching that of donor communities (average B:F: 1.49) (Table 2). The shift in this ratio was driven largely by changes in the proportional representation of Bacteroidetes, which comprised an average of only 7% relative abundance across all pre-FMT samples (excluding study participant 15, who was an outlier), but showed a relative abundance of 41% and 57% in recipient post-FMT and donor samples, respectively (Table 2). Much of the observed variation within the Bacteroidetes was due to proportional shifts involving a single genus, Bacteroides. In addition to Bacteroidetes, the phyla Proteobacteria and Verrucomicrobia were significantly enriched in donor and post-FMT compared to pre-FMT fecal samples (Fig. 3), whereas these three phyla did not differ in abundance between donor and post-FMT fecal samples, highlighting a shift in the microbiome profile, likely caused by FMT, towards that of the donor.

Microbiome analysis—bacterial profile by genus

The value of examining the proportional representation of microbial phyla with regards to gut microbiome health has been questioned (Jandhyala et al., 2015). When examining the changes that occurred instead at the genus level, the most noticeable shifts were reflective of the changes that occurred at the phylum level (Table 2). Specifically, the genus Bacteroides showed a considerable and significant increase in recipients’ pre- to post-FMT (from average 6.5 to 39.2 %). The average relative abundance of the genus Clostridium in the donor was 0.02% (min = 0.001%, max = 0.08%), consistent with a non-disease state. In response to FMT, the average relative abundance of Clostridium in FMT recipients fell roughly 80%, from an average of 1% pre-FMT (min = 0%, max = 4.9%) to 0.16% post-FMT (min = 0%, max = 1.58%). Three of the FMT recipient participants had Clostridium relative abundance of 0%, four had very low but detectable percentages (≤0.01%), while 6 had levels ≥0.5%. Genera identified to be altered significantly in relative abundance per FMT status are listed in Table S1.

Effect of underlying IBD

Of the 15 patients enrolled, 10 had no underlying inflammatory bowel disease (IBD), whereas five had been diagnosed with IBD (four ulcerative colitis and one Crohn’s disease with one patient having had total colectomy) (Table 3). Community taxonomic profiles clustered according to pre- versus post-FMT status, rather than IBD status (Fig. 4), suggesting that underlying IBD had little to no effect on the microbial shift observed because of FMT. There was a marginal effect in the Proteobacteria abundance on separation of community with and without IBD (r² = 0.54, p = 0.018).

Furthermore, no discernable differences in post-procedural symptoms such as abdominal pain, vomiting, or bloody fecal specimens were observed between patients with or without underlying IBD.

Limitations/future directions

Because this was a prospective observational pilot study, microbiome results and statistical implications are limited by a low sample size (n = 15 patients). Not all follow-up phone calls were completed within the time frame set in the study design, and 5 post-FMT samples were not obtained. Additional variables that could impact the microbiome, such as diet, exercise, environmental factors, puberty, etc., were not collected, although the list of such potential factors could be in practice rather exhaustive. Fecal samples were ultimately analyzed three to 12 months’ post-procedure due to inconsistencies in patient compliance. Our covariate analysis is also limited by the fact that study surveys were self-reported. We recognize the risks associated with FMT, and although this study shows promising results, we suggest that FMT should only be used once current guidelines of vancomycin therapy have failed. Future randomized control trials of vancomycin taper versus FMT would be beneficial in determining which option is most effective for treating recurrence. Trials with larger sample sizes need to be performed to establish the effect that FMT has on the microbiome. Furthermore, longitudinal studies of pediatric patients would establish long-term efficacy of this treatment option. Given the small number of pediatric studies, establishing a national FMT database would be useful in analyzing the impact and side effects of FMT. Since this study was an observational study, we did not collect information on participants’ diet or environmental factors which may play a role in the gut microbiome. However, in future studies these determinants of health are important to consider in patients who undergo FMT and also in characterizing factors which affect the gut microbiome.