High-throughput DNA sequence analysis reveals stable engraftment of gut microbiota following transplantation of previously frozen fecal bacteria

Abstract

Fecal microbiota transplantation (FMT) is becoming a more widely used technology for treatment of recurrent Clostridum difficile infection (CDI). While previous treatments used fresh fecal slurries as a source of microbiota for FMT, we recently reported the successful use of standardized, partially purified and frozen fecal microbiota to treat CDI. Here we report that high-throughput 16S rRNA gene sequencing showed stable engraftment of gut microbiota following FMT using frozen fecal bacteria from a healthy donor. Similar bacterial taxa were found in post-transplantation samples obtained from the recipients and donor samples, but the relative abundance varied considerably between patients and time points. Post FMT samples from patients showed an increase in the abundance of Firmicutes and Bacteroidetes, representing 75–80% of the total sequence reads. Proteobacteria and Actinobacteria were less abundant (< 5%) than that found in patients prior to FMT. Post FMT samples from two patients were very similar to donor samples, with the Bacteroidetes phylum represented by a great abundance of members of the families Bacteroidaceae, Rikenellaceae and Porphyromonadaceae, and were largely comprised of Bacteroides, Alistipes and Parabacteroides genera. Members of the phylum Firmicutes were represented by Ruminococcaceae, Lachnospiraceae, Verrucomicrobiaceae and unclassified Clostridiales and members of the Firmicutes. One patient subsequently received antibiotics for an unrelated infection, resulting in an increase in the number of intestinal Proteobacteria, primarily Enterobacteriaceae. Our results demonstrate that frozen fecal microbiota from a healthy donor can be used to effectively treat recurrent CDI resulting in restoration of the structure of gut microbiota and clearing of Clostridum difficile.

Introduction

The incidence and severity of Clostridium difficile infection (CDI) have risen markedly since the 1990s.^1,2 This is reflected in the increased frequency of diagnosis for both community- and hospital-acquired cases, identification of increasingly hypervirulent strains, greater resistance to antibiotics, higher rates of recurrence and greater prevalence of complications that include toxic megacolon, colectomy and overall mortality.^3-6 Frequent failure of antibiotic therapy to eradicate Clostridium difficile infection, particularly in patients with multiply recurrent CDI, has led to increasing use of fecal microbiota transplantation (FMT) as a last ditch, yet highly effective, therapeutic option. In this procedure, fecal material is taken from a healthy donor and introduced into the gastrointestinal tract of the patient via nasogastric tube, enema or colonoscopy.^7-11

Although FMT was first described as a treatment for pseudomembranous colitis in 1958,¹² and since then reported in over 500 cases worldwide as a treatment for CDI,^8,10,11,13 there has been little mechanistic investigation into how this procedure works to restore gut function. It is presumed that FMT restores the normal microbial community structure in the colon, which protects against colonization by C. difficile and suppresses its growth and production of disease causing toxins.⁸ Patients with multiply recurrent CDI demonstrate marked disruption in the bacterial composition of fecal microbiota, primarily Bacteroidetes and Firmicutes, as compared with control subjects and patients experiencing first infection.⁴ It is likely that repeated rounds of antibiotics used to treat recurrent CDI, C. difficile toxins, as well as antibiotics used concurrently for other infections disrupt the distal gut microbiota such that relapse of the infection becomes inevitable. CDI becomes a chronic, recurrent problem in these patients. Chang and colleagues used a clone library approach and found that three patients with multiply recurrent CDI had distal gut microbiota with reduced dominance of Bacteroidetes and increased numbers of Proteobacteria and Verrucomicrobia.¹⁴ We also previously found support for these ideas in a study of one patient with multiply recurrent CDI, where we characterized the fecal microbiota before and after FMT using terminal-restriction fragment length polymorphism (TRFLP) analyses and limited clone libraries of 16S rRNA genes.¹⁵ Similarly, Tvede and Rask-Madsen, using culture based methods, found that Bacteroides sp strains were largely absent in patients with recurrent CDI.¹⁶ However, all these analyses had only limited taxonomic resolving power compared with current high-throughput sequence-based metagenomic approaches.

Until relatively recently, understanding of the microbiota associated with humans, animals, and the environment, has been largely hampered by our inability to culture the vast majority of the attendant microorganisms. More recently, however, molecular and metagenomic approaches have revealed the complexity and diversity of the microbial constituents of the majority of ecosystems thus far examined.^17-24 The human²³ and earth (http://www.earthmicrobiome.org/) microbiome projects are now attempting to describe the microbiota in two complex systems and this has been largely aided by improved and inexpensive massively parallel DNA sequencing technologies. Over the last several years, metagenomic and molecular approaches have revealed a tremendous amount of new information about the microbiota inhabiting the healthy and diseased human intestinal tract.^25-28 Hattori and Taylor referred to the human intestinal microbiome as a “new frontier in human biology” and several now seminal studies have revealed remarkable information concerning microbial ecology and functioning of the gut ecosystem.^28,29 Perhaps this is most dramatic in the case of Clostridium difficile infection (CDI), a common complication of antibiotic therapy.⁸

Here we report on the bacterial composition of fecal microbiota from samples of three patients that were treated by FMT using standardized donor material taken from the same individual and frozen prior to infusion into patients. We used sequence analysis of the V6 hypervariable region of the 16S rRNA gene before and at various time points after FMT and compared the results to the matched donor samples.

Results

Sequencing metrics

The sequencing run generated 21,700,858 paired-end, raw, sequence reads of the V6 region of 16S rRNA. A total of 8,770,494 high quality sequences remained for analysis, after removing low quality sequence reads, chimeric sequences and sequences that only appeared once in the data set. Sequence numbers ranged from 103,368 to 639,014 sequences per sample. To normalize data sets, a random subsample of 103,368 high quality sequences per fecal sample was used for all analyses.

Taxonomy

Taxonomic assignments were made by aligning the consensus sequences from a given OTU (clustered at > 90%) to the RDP7 database and abundance was determined by the number of sequences belonging to each OTU. The relative abundance of taxa at the phylum and family levels for the donor and recipients over time is shown in Figure 1. More complete taxonomy data are available in Table S1.

Figure 1. Bar charts showing the taxonomic classification and the relative abundance of OTUs. All taxa shown are represented by ≥ 0.5% of the total sequence reads. Phylum and family level classifications are shown in panels (A), (B) and (C) and (D), (E) and (F), respectively. All taxa represented by < 0.5% were added together and shown as Less Abundant Taxa. Samples are listed by patient number and days pre (-) and post (+) FMT.

Fecal bacterial composition of donors

Taxonomy data obtained from the three FMT donor samples were very similar to each other in terms of the taxa present, but the relative abundance of taxa varied between samples (Fig. 1). All of these donor data sets showed that members of the Bacteroidetes and Firmicutes were in relatively greater abundance (> 75% of the total sequence reads), with Proteobacteria, Actinobacteria and Verrucomicrobia being present in low abundance (~5% of total reads). Individually, the FMT donor data sets for samples 1, 2 and 3 were represented by 34, 19 and 46% Bacteroidetes sequences, and 43, 57 and 40% Firmicutes, respectively. Unclassified phyla represented about 19, 18 and 10% of the total sequences for donor samples 1, 2 and 3, respectively. In all samples, the Bacteroidetes phylum was mainly represented by members of the families Bacteroidaceae, Porphyromonadaceae and Rikenellaceae. The dominant genera were represented by Bacteroides, Parabacteroides and Alistipes.

In contrast to what was found with the Bacteriodetes, the phylum Firmicutes in donor samples was represented by a more diverse mix of families and was dominated by Lachnospiraceae, Ruminococcaceae and unclassified Firmicutes. Erysipelotrichaceae and unclassified Clostridiales were present, but in lower abundance than the other families. Most of the Firmicutes sequences from the donor samples were not classifiable to the genus level.

Microbiota of patients prior to FMT

In contrast to what was found with the donor samples, the pre-transplantation samples obtained from all three patients showed a greater abundance of sequence reads belonging to the Proteobacteria and Firmicutes phyla. In samples from patients 1 and 2, > 99% of the sequences were classified as Proteobacteria or Firmicutes, while the sample from patient 3 contained a great abundance of sequences belonging to the Bacteroidetes and unclassified phyla. Relative to that seen with the FMT donor samples, all of the pre-transplantation samples were comprised of very few unclassified sequences at the phylum level. The relative abundance of detected phyla varied considerably among the three pre-transplantation samples from the tested patients. Fecal DNAs from the pre-transplantation sample from patient 1, obtained 3 d before transplantation while taking vancomycin, was comprised of 77% Proteobacteria and 23% Firmicutes. Nearly all of the Proteobacteria sequences in this sample were classified to the Enterobacteriaceae family, with Klebsiella, Salmonella and unclassified taxa as the dominant genera. Firmicutes sequences from the patient 1 pre-transplantation sample were identified as mainly belonging to the Lactobacillaceae and Veillonellaceae families and the genera Lactobacillus and Veillonella.

The pre-transplantation sample from patient 2 (obtained 1 d before transplantation and having just stopped vancomycin therapy) was similarly dominated by Proteobacteria and Firmicutes, representing about 40% and 60% of total sequences, respectively. Similar to what was seen in patient 1, the Proteobacteria were mostly classified as Enterobacteriaceae, and Escherichia/Shigella, Kyuvera and unclassified taxa were the dominant genera present in this sample. Enterococcus and Veillonella were the dominant genera from the Firmicutes phylum in this sample.

The pre-transplantation sample obtained from patient 3 (15 d before transplantation, and not taking vancomycin) contained Proteobacteria, Firmicutes and Bacteroidetes phyla, representing 24, 36 and 33% of the total sequences, respectively. About 6% of sequences from this sample were unclassified at the phylum level and 1% belonged to the phylum Fusobacteria. The Proteobacteria sequences in this sample were again dominated by members of the Enterobacteriaceae family, but with the Escherichia/Shigella, Kyuvera and Parasutterella genera in high relative abundance. Firmicutes taxa were represented by members of Veillonellaceae and Erysipelotrichaceae families, specifically members of the Acidaminococcus and Clostridium type XVIII genera.

FMT results in engraftment of donor fecal microbiota

The taxonomy data from post-transplantation samples obtained from the recipients revealed similar taxa as were found in the donor samples, but the relative abundance varied considerably between patients and time points. All fecal samples obtained from patients post FMT showed an increase in both Firmicutes and Bacteroidetes compared with the pre-transplantation samples from matched patients, except for the later samples obtained from patient 3. The post-transplantation samples from patients 1 and 2 were very similar to each other and to the matched donor sample, in terms of the taxa present, with variability in abundance of specific taxa. All patient 1 samples showed a greater abundance of the Firmicutes and Bacteroidetes, averaging about 75% of the total sequence reads. About 17% of the sequences represented unclassified bacteria. Fecal samples obtained from patient 2 post FMT also showed a relatively greater abundance (about 80%) of sequences consistent with Firmicutes and Bacteroidetes, with an average of 11% unclassified bacteria. Proteobacteria and Actinobacteria were present in lower abundance (< 5%) than found in patients prior to FMT.

The post FMT samples from patients 1 and 2 were very similar to donor samples, with the Bacteroidetes phylum represented by a greater abundance of the Bacteroidaceae, Rikenellaceae and Porphyromonadaceae families. These samples were largely comprised of members of the genera Bacteroides, Alistipes and Parabacteroides. Members of the Firmicutes taxon were represented by the families Ruminococcaceae, Lachnospiraceae, Verrucomicrobiaceae, unclassified Clostridiales and unclassified members of the Firmicutes. Most of these Firmicutes in patients receiving FMT were unable to be classified at the genus level, similar to what was found in the donor samples.

Sequences from fecal samples obtained 20 d post-transplantation from patient 3 contained similar taxa to those seen in donor samples and the other post-transplantation samples from all patients. About 72% of the sequences belonged to the phyla Bacteroidetes and Firmicutes. These sequences were further classified into the same families and genera as the donor samples and post FMT feces obtained from patients 1 and 2. However, feces from patient 3, obtained from the later time point samples, had sequence profiles containing a greater abundance of Proteobacteria (particularly the Enterobacteriaceae), which represented 10, 50, 18 and 28% of the sequences from the 28, 36, 68 and 90 d post-FMT data sets, respectively. At the genus level, the Enterobacteriaceae were almost all classified as Escherichia/Shigella in the 28, 36 and 68 d post-FMT data sets. This shift in microbial composition coincided with the patient developing a brief episode of urinary tract infection that was ultimately treated with a course of antibiotics. Interestingly, we did not see a return of the microbial composition to the early post-FMT state, and the 90 d post-FMT sample from this patient showed a further shift in composition, with Enterobacteriaceae sequences largely classifying as Kyuvera and unclassified ε-proteobacteria.

OTUs and diversity indices

The total number of OTUs (at a 90% cutoff) per sample, Chao richness estimations and Shannon diversity index values for all samples are shown in Table 2. The number of OTUs observed in each sample ranged from 261 to 1233, with the lowest number found in the pre-transplantation samples from the three patients. The mean number of OTUs in the patient samples prior to FMT was 296 ± 33. The greater number of OTUs was detected in the donor samples with a mean of 984 ± 160 per sample. In contrast, patient fecal samples post FMT contained 1072 ± 100, 933 ± 134 and 840 ± 208 OTUs per sample for patients 1, 2 and 3, respectively. The Chao estimate of OTU richness consistently predicted more OTUs than were observed. Shannon diversity indices ranged from 1.9 to 4.6, with the lesser values associated with pre-FMT samples and the day 36 sample obtained from Patient 3. Diversity index values for donor and post FMT samples indicated a high level of OTU richness in these samples. OTUs were shared across all samples. A heatmap of the top 15 OTUs (in the total data set) is shown in Figure 2. A table showing the total number of OTUs for each sample and the number of shared OTUs between samples is shown in Table 3.

Table 2. OTU data and diversity indices of donor and FMT recipients.

Sample	OTUs	Shannon	Chao
Pt 1 Donor	1041	4.24	1395
Pt 1–3	261	1.91	391
Pt 1 +3	977	4.41	1504
Pt 1 +7	896	4.14	1448
Pt 1 +11	1086	4.25	1559
Pt 1 +18	1075	4.23	1640
Pt 1 +25	1100	4.35	1700
Pt 1 +65	1233	4.49	1693
Pt 1 +95	1085	4.55	1496
Pt 1 +128	1127	4.3	1633
Pt 2 Donor	1108	4.63	1579
Pt 2 -1	327	2.07	482
Pt 2 +3	883	4.15	1432
Pt 2 +7	1034	3.59	1626
Pt 2 +27	738	3.23	1152
Pt 2 +41	1100	3.95	1611
Pt 2 +81	849	3.77	1261
Pt 2 +112	994	4.03	1544
Pt 3 Donor	803	3.41	1184
Pt 3 -15	301	2.42	506
Pt 3 +7	995	3.94	1670
Pt 3 +13	1047	4.03	1598
Pt 3 +20	963	4.01	1500
Pt 3 +28	793	3.59	1158
Pt 3 +36	533	2.34	848
Pt 3 +68	963	3.9	1459
Pt 3 +90	585	3.2	952

Figure 2. Heatmap comparison of the abundance of the top 20 OTUs across samples. All sequence data were used to calculate the top 20 most abundant OTUs, rather than the most abundant OTUs in a given sample. Samples are listed by patient number and days pre (-) and post (+) FMT.

Table 3. Shared OTUs between samples obtained from the FMT donor and recipients over time.

1041

261

977

896

1086

1075

1100

1233

1085

1127

1108

327

883

1034

738

1100

849

994

803

301

995

1047

963

793

533

963

585

Pt 1 Donorr	Pt 1–3	Pt 1 +3	Pt 1 +7	Pt 1 +11	Pt 1 +18	Pt 1 +25	Pt 1 +65	Pt 1 +95	Pt 1 +128	Pt 2 Donor	Pt 2 -1	Pt 2 +3	Pt 2 +7	Pt 2 +27	Pt 2 +41	Pt 2 +81	Pt 2 +112	Pt 3 Donor	Pt 3 -15	Pt 3 +7	Pt 3 +13	Pt 3 +20	Pt 3 +28	Pt 3 +36	Pt 3 +68	Pt 3 +90
	140	625	573	634	643	648	680	618	659	696	162	549	626	478	676	559	615	547	147	571	583	579	500	283	619	328	Pt 1 Donor
		155	140	161	152	146	158	159	153	147	152	155	154	140	143	140	156	143	117	142	142	139	147	153	150	144	Pt 1–3
			590	618	632	628	656	625	619	635	177	536	597	494	622	552	593	503	146	544	551	551	468	282	534	322	Pt 1 +3
				583	633	606	613	601	603	582	160	464	560	481	595	527	551	515	137	510	538	531	469	303	480	294	Pt 1 +7
					713	745	762	687	678	662	184	548	595	473	618	528	597	499	146	513	537	516	447	303	540	314	Pt 1 +11
						737	757	704	709	661	173	523	604	505	637	570	603	546	147	557	581	579	503	307	554	335	Pt 1 +18
							801	728	729	700	168	550	611	494	628	559	612	536	141	540	565	553	462	284	557	330	Pt 1 +25
								795	772	712	177	565	654	503	674	581	649	538	148	542	573	556	468	284	583	342	Pt 1 +65
									740	669	167	514	561	496	581	547	577	540	135	504	537	519	447	283	510	300	Pt 1 +95
										690	177	540	633	508	652	555	608	567	148	566	585	566	494	301	558	332	Pt 1 +128
											168	576	613	500	652	583	630	567	158	604	616	618	487	300	596	335	Pt 2 Donor
												188	185	163	174	163	182	162	145	167	166	160	177	174	183	172	Pt 2 -1
													589	426	585	505	597	437	153	501	508	490	430	289	524	298	Pt 2 +3
														536	717	601	673	520	156	573	583	573	500	304	551	342	Pt 2 +7
															558	527	521	441	138	447	475	481	433	270	436	275	Pt 2 +27
																639	710	530	148	587	621	598	518	295	593	352	Pt 2 +41
																	604	473	142	525	536	539	493	285	513	302	Pt 2 +81
																		503	150	548	576	566	482	298	556	323	Pt 2 +112
																			142	505	518	504	451	285	461	294	Pt 3 Donor
																				179	168	172	180	187	181	183	Pt 3 -15
																					684	606	541	336	592	354	Pt 3 +7
																						624	531	338	603	344	Pt 3 +13
																							564	338	572	348	Pt 3 +20
																								342	518	337	Pt 3 +28
																									326	241	Pt 3 +36
																										370	Pt 3 +68

Unifrac analysis

Unifrac analysis was used to compare the community structure of samples. Unifrac trees were examined by principle coordinate analysis (PCoA) to cluster samples from the individual patient data sets. The PCoA plots are shown in Figure 3. Results obtained from patient 1 and 2 data showed a large cluster containing both donor and pre-transplantation samples, while PCoA plots constructed from patient 3 data show a similar effect, except that post FMT samples from days 36 and 95 fell outside of the main cluster. The pre-FMT samples from all patients clustered separately in each of the plots shown in Figure 3.

Figure 3. Principal coordinate plots generated using the Unifrac algorithm for individual patient data sets. Samples were normalized and abundance weighting was used in the analysis. Data for patients 1, 2 and 3 are shown in panels (A), (B) and (C), respectively. Samples are listed by days pre (-) and post (+) FMT. The percent variance explained by each axis is shown in the axis labels.

Discussion

Previous results from our laboratory demonstrated that a standardized frozen preparation of fecal microbiota is a highly effective clinical treatment for recurrent C. difficile infection.¹¹ Results presented in this current study indicate that the standardized frozen preparation of fecal microbiota successfully engrafted into the recipient patients as indicated by a dramatic shift in their gut microbial communities after transplantation. Engraftment occurs rapidly, with patient microbial communities becoming similar in composition to donor samples within three days of the transplant procedure. This was accompanied by a cessation of CDI-related symptoms and eventual resolution of gastrointestinal symptoms. While we previously showed engraftment of fecal microbiota in a single patient using TRFLP analysis and sequencing of a limited number of clones,¹⁵ the present study presents more in depth analysis using a cohort of patients and 8.7 million high quality 16S rRNA sequences of the bacterial V6 region obtained using Illumina technology. While the transplanted microbiota primarily consisted of bacteria and archea, we are cognizant that some viruses and fungi many also be present, albeit in small numbers due to our method of preparation.

Taxonomy results shown in Figure 1 demonstrated a great abundance of Proteobacteria in patient samples obtained prior to transplantation. This is similar to previous reports indicating that the gut microflora of CDI patients is very perturbed relative to that seen in healthy patients.^14,15 Normally, Proteobacteria are a relatively minor constituent of the gut microbiota, with Firmicutes and Bacteroidetes representing the major bacterial phyla found in the feces.³⁰ While Firmicutes strains were present in pre-FMT samples from all three patients, Bacteroidetes taxa were nearly absent in samples from Patient 1 and 2. In contrast, Bacteroides strains were present in pre-FMT samples from patient 3. The most likely explanation for this difference is timing of exposure to antibiotics. Pre-FMT samples from patients 1 and 2 were collected at the end of their pre-FMT vancomycin course. However, the pre-FMT sample from patient 3 was collected just as she was diagnosed with her latest recurrence of CDI and before starting vancomycin therapy. Nevertheless, PCoA of Unifrac data showed that all pre-FMT samples clustered away from the donor and patient’s post-transplantation samples. These results support the idea that exposure to antibiotics, including those used to treat CDI, cause marked disruption of microbial community structure in CDI patients. It is also possible that CDI itself contributes to dysbiosis within the gut although the role that C. difficile alone plays in altering gut microbial populations is difficult to separate because antibiotic induced dysbiosis is typically required to create a suitable environment for C. difficile growth.

In contrast to what was seen in the recipients, the donor and post FMT patient samples showed an increase in the relative abundance of Firmicutes and Bacteroidetes strains. These results are consistent with reports examining the fecal microbiome of healthy individuals.^15,30-33 While the phylum level taxonomy and relative abundance remained fairly consistent across patients, family level taxonomy indicated a shift in the abundances of taxa between samples of matched patient sets and the standard donor samples. Despite the variation in taxonomic abundances in lower level taxonomy, the gut microbial profile of patients more closely resembled that of the donor, compared with the pre-transplantation samples within 3 d after the FMT procedure. Community analysis using Unifrac also showed that post transplantation samples clustered with donor samples and away from pre-FMT samples. This rapid shift was accompanied by recovery from CDI, as measured by a lack of symptoms and a negative clinical PCR test for the C. difficile toxin B gene at one month post-FMT. Similar to what we previously reported using TRFLP analysis of a single patient,¹⁵ this result further confirms the protective effect provided by the normal gut microbiota was restored in these patients followed by clearance of the infection.

The overall profile of gut microbiota in patients 1 and 2 post FMT was similar to that of donor up to almost 4 mo after transplantation. This indicated long-term engraftment of the donor’s microbiota into recipients. It is interesting to note that the proportion of bacteria belonging to the phylum Bacteriodetes and the family Bacteroidaceae in patient 1 decreased over time compared with the donor. It is possible that underlying ulcerative colitis or treatment with immunosuppressive medications were causally linked to this change, although at this stage this remains an anecdotal observation that should be followed-up by studying more recipients with and without underlying UC. Notably, the intestinal gut microbiota in fecal samples from patient 3 showed increased Proteobacteria, similar to that seen at the phylum level in this patient’s pre-transplantation samples. While the abundance of Proteobacteria was less than 5% in all post transplantation samples from patients 1 and 2 and the three donor samples, this phylum in the patient 3 post-FMT samples exceeded 10% of the total sequence reads from 28 to 90 d post transplantation, with days 36 and 90 showing a relative abundance of 52 and 40%, respectively. These samples also clustered outside of the main cluster of donor and early post-FMT samples. The increase in abundance of Proteobacteria in the later patient 3 post-FMT samples coincided with development of urinary tract infection symptoms and may have been perpetuated by a brief course of sulfamethoxazole-trimethoprim, for treatment of a urinary tract infection caused by Proteus mirabilis. The coincidence of the observed shift toward greater abundance of Proteobacteria in the gut microbiota and emergence of a urinary tract infection before antibiotic treatment raises a provocative possibility of a causal link between these events. To help prevent the reactivation of CDI, this patient was simultaneously prescribed low dose oral vancomycin, which also may have contributed to the shift in gut microbiota. Persistence of a high proportion of Proteobacteria even two months after cessation of these antibiotics suggests a long-term perturbation of the microbial fecal composition in this patient and their possible vulnerability to antibiotic insults. Interestingly, about one third of patients successfully treated with FMT for multiply recurrent CDI experienced re-infection with C. difficile after being treated again with antibiotics for an unrelated infection.⁹

The OTU counts and Shannon diversity index values supported the taxonomy results, with greater community diversity in donor and post-FMT samples compared with pre-transplantation samples. The results from this study suggest that FMT performed using a standardized frozen preparation of fecal microbiota is capable of restoring the high level of diversity seen in healthy individuals such as the donor. Additionally, the large proportion of OTUs shared between donor and patient matched post-transplantation samples and post-FMT samples from other patients suggests that the community composition of post-transplantation samples is greatly influenced by the donor’s microbiota itself. Host effects, diet, metabolic state and other variables likely also play a role in shaping the microbial communities in patients after FMT, however these effects were not examined in this study.^34,35

The Chao estimator of species richness indicates that additional sampling would have yielded additional OTUs and rarefaction curves generated using OTU data also suggested that additional sampling would have identified more OTUs (see Fig. S1). While this is likely, approximately 87% of sequences belonged to the top 100 most abundant OTUs, while the remaining 13% of sequences were split between the remaining 3,468 OTUs. This suggests that additional sampling would likely identify additional OTUs also present in very low abundance, possibly arising as a result of sequencing error.¹⁷ The dominant OTUs were represented by tens of thousands of sequences, suggesting our methodology gives an adequate sample data set for the identification of the dominant OTUs and taxa present in each sample, providing adequate resolution to identify the dramatic shifts seen in samples obtained after transplantation.

Despite bi-directional, end-paired sequencing of amplicons using the Illumina system, sequencing errors likely contributed to the presence of sequence reads that were only detected a few times in our data set. Additionally, clustering OTUs at 90% may have lead to the inclusion of sequences that may have classified differently than what was obtained using an OTU consensus sequence. While these effects are unknown, it is important to note that all of our samples were processed and amplified in replicates and biases are assumed to be equal among samples. Taking that into account, the dramatic shift seen in the patient’s fecal microbiota after FMT and the apparent similarity between donor and post-transplantation samples is almost surely due to engraftment of the donor’s fecal microbiota as was previously shown with a single patient and TRFLP and clone library analyses.¹⁵

This study represents the first examination of CDI patients treated by FMT and analysis by using high-throughput sequencing analysis of 16S rRNA genes. Our results suggest that engraftment of a standardized and frozen preparation of donor fecal microbiota restores the protective effect provided by the normal flora similar to that seen when using fresh donor material from a familial donor.¹⁵ While additional experiments are required to determine the key taxa involved in the protective effect against C. difficile, the results presented here represent a significant step toward our understanding of the changes in gut microbial communities of recurrent CDI patients after treatment by FMT. Furthermore, the post-FMT recipients may comprise a unique and interesting group of patients to study various factors that are responsible for the stability of the gut microbial communities in the human host.

Materials and Methods

Patients

All patients suffered from multiply recurrent CDI that was unable to be cleared by standard antibiotic therapies, as defined previously.¹¹ The study of their fecal microbiota before and after FMT was approved by the University of Minnesota Institutional Review Board and all patients provided informed consent to participate in this study.

Patient 1 was a 67 y old man with psoriasis and an approximately 20 y history of ulcerative colitis (UC). His first occurrence of CDI followed a surveillance colonoscopy and a course of antibiotics for unclear indication. He received one, two-week course, of metronidazole, three courses of vancomycin, including one two month-long taper and one vancomycin/rifaximin chaser treatment protocol. However, the patient relapsed with CDI, every time, within 1−2 weeks of stopping antibiotics and had CDI for eight months prior to FMT. The patient received infliximab 10 mg/kg infusions every six weeks and 10 mg methotrexate, intramuscularly once weekly, for his UC. Colonoscopy performed at the time of FMT showed severe colitis using the Mayo Clinic Endoscopic Score system, extending from his rectum to transverse colon. Following FMT, the patient did not experience recurrence of CDI over an 18 mo clinical follow-up period during which time he did not receive any more antibiotics. The activity of his ulcerative colitis improved to mild or moderate by clinical, endoscopic and histologic criteria. We suspect that the clinical improvement in this patient was due to resolution of recurrent CDI, but it is possible that UC activity was also modulated independent of the effects on CDI. The patient’s schedule of infliximab infusions and methotrexate administration continued without change.

Patient 2 was a 76 y old woman who developed CDI during hospitalization following a car accident and subsequent surgeries. She had multiple recurrences of CDI and failed two six-week antibiotic tapers using vancomycin and a course of rifaximin “chaser.” During this period of time she lost 20% of her body weight and suffered a new hip fracture due to a fall. In total, her recurrent CDI lasted 11 mo until she was treated with FMT. The patient did not experience recurrence of CDI after 14 mo of follow-up.

Patient 3 was a 74 y old woman who developed initial CDI, requiring hospitalization, following treatment of diverticulitis with metronidazole and ciprofloxacin. She received oral vancomycin, but the infection relapsed after discontinuation. Ultimately, she had at least six courses of vancomycin for recurrent infection over an 11 mo period, including at least one six-week tapered course of vancomycin. Colonoscopy performed at the time of FMT showed moderate left-sided diverticulosis. The patient reported mild symptoms of dysuria during a routine phone follow-up on day 33 following FMT. These symptoms progressed and on day 46 she was given 800–160 mg sulfamethoxazole-trimethoprim, twice-daily for a week and 125 mg vancomycin, once daily for a week, for prophylaxis against recurrence of CDI. Urinalysis from day 46 was consistent with a urinary tract infection and urine culture showed > 100,000 colonies/mL of Proteus mirabilis. No CDI was detected after one year of clinical follow-up.

Transplantation of fecal microbiota

The FMT was performed using a standardized preparation of concentrated fecal bacteria as previously described.¹¹ Briefly, microbiota were recovered from 50 g aliquots of feces collected from a prescreened donor and were mixed with 250 ml of sterile, nonbacteriostatic phosphate buffered saline (PBS). Feces was blended, sieved and the resulting bacterial preparation was centrifuged and washed in PBS and frozen in 10% (v/v) glycerol to produce the concentrated fecal microbiota preparation for FMT. Preparations were stored frozen at -80°C until used. Frozen lots were thawed on ice for two hours prior to use, at which time they were diluted with room temperature sterile normal saline to a total volume of about 240 mL. The same donor was used for all FMT procedures described here.

FMT was performed via colonoscopy as previously described.¹¹ The patients were treated with 125 mg vancomycin, four times daily by mouth, until two days prior to the procedure. The day before the procedure, the patients were prepped with a split dosage of polyethylene glycol-based purgative (GoLYTELY^®, Braintree Laboratories) to remove residual antibiotic and fecal material. The patients underwent a full colonoscopy under conscious sedation and the donor fecal microbiota material was injected into the terminal ileum and cecum via the biopsy channel of the colonoscope.

Sample collection

Patient fecal samples were collected at home by the patients and stored frozen at approximately -20°C. Samples were transferred to the laboratory within one week of collection and stored at -80°C until used. Donor samples for DNA extraction were collected during processing of material for FMT and stored frozen at -80°C until used.

DNA extraction

DNA was extracted from donor and recipients pre- and post-FMT fecal samples (0.25−0.50 g) using MOBIO PowerSoil DNA extraction kits (MOBIO), according to the manufacturer’s instructions. Samples with high water content (Bristol scale types 5−7) were centrifuged at 10,000× g for three min to pellet solids, which were subsequently extracted for DNA. Triplicate fecal DNA samples were each eluted in 50 µl of 10 mM Tris-Cl buffer, pH 8.0 and pooled. Fecal DNA concentrations were measured using a QuBit DNA quantification system (Invitrogen) with QuBit high sensitivity assay reagents. All fecal DNA samples were stored frozen at -20°C until used.

PCR amplification

Fecal DNA samples were used as template in PCR amplification reactions of the V6 hypervariable region of the 16S rRNA gene. All PCR reactions used 25 ng of fecal DNA as template and were performed in triplicate. Primer sets (Table 1) were designed with a 6 bp ID tag on the 5′ end of either the forward or reverse primer(s), which was specific to each fecal DNA sample. This allowed for multiplexed sequencing. PCR amplicons were visualized using gel electrophoresis to confirm amplification of properly sized products. Triplicate reactions were each purified using the Qiaquick PCR purification kit (Qiagen), eluted in 50 µl of 10 mM Tris-Cl buffer, pH 8.0 and pooled. Purified DNA concentrations were measured using the QuBit DNA quantification system and high sensitivity assay reagents (Invitrogen). Samples were stored frozen at -20°C until pooled for sequencing.

Table 1. Primer sequences used in this study.

Primer Name	Sequence (5′−3′)
Forward Primer 1	CNACGCGAAGAACCTTANC
Forward Primer 2	CAACGCGAAAAACCTTACC
Forward Primer 3	CAACGCGCAGAACCTTACC
Forward Primer 4	ATACGCGARGAACCTTACC
Forward Primer 5	CTAACCGANGAACCTYACC
Reverse Primer 1	[6bp ID tag]CGACAGCCATGCANCACCT
Reverse Primer 2	[6bp ID tag]CGACAACCATGCANCACCT
Reverse Primer 3	[6bp ID tag]CGACGGCCATGCANCACCT
Reverse Primer 4	[6bp ID tag]CGACGACCATGCANCACCT

DNA sequencing

Equimolar aliquots of each product (27 total samples) were pooled to give ~1 µg of DNA in 100 µl total volume. Final pooled DNA concentration was measured by using the Quant-IT PicoGreen quantitation system (Invitrogen). Amplicon size analysis was done using an Agilent DNA 1000 chip and a 2100 BioAnalyzer (Agilent). The pooled sample was shipped on dry ice to the National Center for Genome Resources (NCGR) for library preparation and sequencing using Illumina technology following the manufacturer’s protocols (Illumina). Paired-end sequences were generated on the HiSeq 2000 sequencer (100 nt read length) with 1−3 pooled samples per lane following Illumina multiplexing protocols. Reads in each pair overlapped and paired ends were merged. Briefly, hamming distance (number of substitutions) was calculated for sliding overlaps of the two reads in a pair to find the best overlap (lowest hamming distance with a minimal overlap of 25 nucleotides and 98% identity). Merged sequences (21,700,858) were binned according to barcode sequence and barcode and amplicon primer sequences were trimmed using custom Perl scripts.

Sequence processing and analysis

Sequence data was processed and analyzed using the MOTHUR program.³⁶ To ensure high quality data for analysis, sequence reads containing ambiguous bases, homopolymers > 7 bp, more than one mismatch in the primer sequence, or an average per base quality score below 25 were removed. Sequences that only appeared once in the total set were assumed to be a result of sequencing error and removed from the analysis. Chimeric sequences were also removed from the data set using the UCHIME algorithm within the MOTHUR program.³⁷ A random subset of 103,368 sequences from each sample was used to balance read numbers. This subset of high quality sequence reads was aligned to the RDP7 16S rRNA database and clustered into operational taxonomic units (OTUs) at a cutoff value of > 90%.³⁸ Taxonomy was assigned to OTU consensus sequences by using the Ribosomal Database Project ver. Seven (RDP) taxonomy database, the Bayesian method with a bootstrap algorithm (100 iterations) and a probability cutoff of 0.60. OTU data was used for determinations of Shannon diversity indices (an indicator of community diversity) and Chao richness estimates (an estimate of species number).³⁹

Unifrac analysis

Samples were clustered using the UniFrac algorithm to generate trees and principle coordinate analysis (PCoA) plots.⁴⁰ Unifrac is a method of comparing microbial communities based on phylogenetic data. The UniFrac algorithm was run using the Fast Unifrac program available at www.bmf2.colorado.edu/fastunifrac/. PCoA plots were generated using Sigma Plot ver. 10.0.1 (Systat Software).

Glossary

Abbreviations:

FMT

fecal microbial transplantation

CDI

Clostridum difficile infection

TRFLP

terminal-restriction fragment length polymorphism

UC

ulcerative colitis

OTU

operational taxonomic unit

RDP

ribosomal database project

PCoA

principle coordinate analysis

Disclosure of Potential Conflicts of Interest

A.K. and M.J.S. received funding from CIPAC LLC to carry-out research on FMT using frozen microbiota. A.K. and M.J.S. have provided consulting services for CIPAC and conflicts of interest are being managed by the University of Minnesota Conflicts of Interest Program.