Dynamic changes in short- and long-term bacterial composition following fecal microbiota transplantation for recurrent Clostridium difficile infection

Bacterial composition of fecal samples from patients with recurrent CDI becomes healthy
and donor-like following FMT

Four patients (CD1 to CD4) with recurrent CDI were treated with FMT using material
obtained from a single donor but from different time points, and fecal samples were
collected from these patients before and after the procedure as well as from the donor
at times of donation. Bacterial communities from these fecal samples were characterized
by sequencing the V4 region of the 16S rRNA gene. Following trimming and quality filtering
from a total of 12,536,492 sequences, we randomly subsampled to 5,000 sequences/sample
in order to normalize read depth across all samples. All further analyses were performed
using this rarefied read depth.

To better understand changes in bacterial communities following FMT, we compared the
bacterial composition of patient fecal samples to those of microbial communities from
various body sites from the 252 healthy individuals characterized in the Human Microbiota
Project (HMP) 19] (Figure 1) using unweighted UniFrac 20] followed by principal coordinates analysis (PCoA) 21] (see Additional file 1: Movie supplement). The composition of pre-FMT fecal samples from patients CD1 to
CD4 and 10 additional patients with recurrent CDI was distinct from both fecal samples
from healthy individuals and microbial communities at other body sites, including
mouth, vagina, and skin, demonstrating severe alterations in pre-FMT communities compared
to healthy fecal communities as has been previously shown 4],5]. In contrast, microbial communities from the donor fell within the range of healthy
fecal samples. Using an animated visualization of FMT-associated changes in patients’
fecal microbial communities, we observed rapid and dramatic shifts after FMT towards
the communities found in the feces of healthy individuals and of the original donor
(see Additional file 1: Movie supplement).

Figure 1. Fecal bacterial communities of recurrent CDI patients shift towards HMP fecal bacterial
communities after FMT. Pre-FMT patient samples (red circle); post-FMT patient samples (green circles); trajectory
of patient fecal communities after FMT (blue line).

Fecal microbial communities remain dynamic following FMT

To more closely examine temporal changes in recipient fecal samples following FMT,
we analyzed fecal microbial communities from patients CD1 to CD4 and donor, as well
as from 10 additional donor samples, using weighted and unweighted UniFrac 20] followed by PCoA 21]. This analysis demonstrated that fecal bacterial communities continued to undergo
compositional fluctuation following FMT (Figure 2A and Additional file 2: Figure S1; individuals OTUs listed in Additional file 3: Table S1).

Figure 2. Microbial communities shift following FMT. (A) Unweighted (left) and weighted (right) UniFrac analyses followed by principal component
analysis of bacterial communities of recurrent CDI patient fecal samples before (red)
and after FMT and donor samples (blue). (B) Weighted UniFrac analysis followed by principal component analysis of bacterial communities
of patients before (red) and after FMT versus HMP fecal communities (purple). PC,
principal component. Percentages represent percent variability explained by each principal
component. Se key at right for colors associated with samples before FMT (pre-FMT),
from HMP and donor, and from patients after FMT (CD1 to CD4).

To determine whether this dynamic range of post-FMT microbial composition fits within
the range seen across healthy individuals, we also compared communities in our samples
to those in the HMP via weighted UniFrac and PCoA (Figure 2B). Again, fecal microbial communities prior to FMT were highly distinct from healthy
fecal microbial communities, and following the procedure, these communities more closely
resembled those of healthy individuals. Similar to the comparison with donor communities
above, fecal microbial communities of recurrent CDI patients following FMT shifted
within the cluster of communities from healthy individuals.

Rapid and substantial changes to Enterobacteriales in feces following FMT

While overall fecal microbial communities were dramatically altered following FMT,
we also examined the effects of the procedure on the abundance and dynamics of individual
bacterial taxa within the four original CDI patients. As shown previously 2]-8], the relative abundance of bacterial phyla in patient fecal samples shifted substantially
following FMT, with relative decreases in Proteobacteria and relative increases in Bacteroidetes and Firmicutes (Figure 3). These Proteobacteria are primarily the order Enterobacteriales, which were also substantially decreased in relative abundance following FMT (Figure 4A).

Figure 3. Changes in fecal microbial communities after FMT. Relative abundance of sequences classified to the level of bacterial phyla before
and after FMT in patient fecal samples. Samples after FMT indicated with dashed line.
See key at right.

Figure 4. Changes in the orderEnterobacterialesafter FMT. (A) Relative abundance of Enterobacteriales in donor and patient samples before and after FMT in samples common across all patients.
(B) Control charts of relative abundance of Enterobacteriales in donor (leftmost sample) and patient samples before and after FMT. Patient CD1
(top left), patient CD2 (top right), patient CD3 (bottom left), patient CD4 (bottom
right). LCL, lower control limit; UCL, upper control limit; mean relative abundance
in all samples (center). LCL and UCL represent three standard deviations in relative
abundance below and above the mean, respectively. Dashed lines indicate samples after
FMT.

We focused on these changes by examining the relative abundance of Enterobacteriales alone in each patient before and after FMT. The relative abundance of this taxon
ranged from 44% to 82% in all four patient samples prior to FMT and rapidly dropped
to undetectable levels within 1 week after the procedure. Moreover, abundance of this
taxon remained low at 26 days after FMT, the latest time point shared by all four
patients (Figure 4A), although other members of the Proteobacteria remain detectable if decreased in relative abundance (Figure 3). In addition, we generated individual value control charts based on the average
abundance of this taxon in recurrent CDI patients. Compared to relative abundance,
these control charts displayed the expected variation of the abundance of Enterobacteriales in these fecal samples. In all patients, the abundance of Enterobacteriales was above the expected variation (that is, more than three standard deviations above
the mean relative abundance [the standard upper control limit, or UCL] of this order
across all samples) prior to FMT, and rapidly fell below the upper control limit within
1 to 2 days after the procedure (Figure 4B). These results suggest that the relative abundance of Enterobacteriales significantly decreased in all patients soon after FMT to levels similar to donor
samples and remained within a statistically expected range for the duration of sample
collection (up to 151 days post-FMT).

Post-FMT communities are initially similar to donor samples but can later diverge

Next, we compared fecal microbial communities within each patient over time to that
of the initial donor sample. We generated heat maps based on Pearson correlations
between every sample within a given patient set, including respective donor samples
and samples from 10 additional pre-FMT patients (Figure 5A). This analysis revealed that while microbiota in samples from patients after FMT
quickly became similar to microbiota in donor samples, the similarity of samples taken
at later time points after FMT fluctuated.

Figure 5. Pearson and Spearman correlations between fecal communities before and after FMT.
(A) Heat map of Pearson correlation values between each sample within each patient set,
corresponding donor, and 10 additional pre-FMT patient samples (far right). (B) Pearson correlation values between donor sample and each patient sample. (C) Spearman correlations between donor sample and each patient sample. (D) Heat maps of Pearson (i) and Spearman (ii) correlation values between earliest donor
sample and eleven subsequent samples; days represent collection time of each sample
versus earliest donor sample. CD1 to CD4, patients 1 to 4. Dashed lines indicate samples
after FMT.

To further investigate how fecal microbial communities in these patients correlate
to donor communities, we examined Pearson and Spearman correlations between donor
and patient samples, which were common to each patient (pre-FMT samples and those
up to 26 days post-FMT; Figure 5B,C and Additional file 4: Figure S2). While fecal microbial communities from patients before FMT were highly
distinct from those in the donor, fecal microbial communities from samples 1 day after
the procedure were highly correlated to donor communities via both Pearson and Spearman analyses in all patients. After the initial time point
after FMT, the Pearson correlation values of patient to donor samples were highly
variable within and across patients, although Spearman correlations remained high
for three patients. To examine whether this variation is similar in healthy individuals,
we determined Pearson and Spearman correlations within the four donor samples used
in FMT, as well as eight additional donor samples from the same individual as a control.
Results of this analysis revealed that donor microbiota also changed over time (Figure 5D). These findings suggested that the level of variability seen across patient post-FMT
fecal microbial communities was within the range of normal microbiota behavior in
a healthy individual.

Normalization and dynamic range of post-FMT patient fecal microbial communities are
similar to donor communities

Because of the observed variability in later post-FMT patient fecal communities relative
to single donor communities, we compared the communities of these patient samples
to an expanded set of 17 samples taken from the same donor. We generated two metrics
to evaluate the relationships between these communities: normalization and dynamic
range (stability). Normalization refers to the mean between-sample distance for each
set of patient samples versus the set of donor samples, while dynamic range is the
mean distance between each sample within a single patient set. Effectively, the normality
of a post-FMT patient sample set is a measure of how similar it is to the donor (healthy)
sample set, while dynamic range is a measure of variability within a given patient
sample set. We found that neither the normalization nor the dynamic range of any post-FMT
patient sample set was significantly different than the donor set following analysis
using unweighted UniFrac (Table 1). This suggested that although fecal microbial communities of patients post-FMT do
not remain identical to the donor, they nonetheless fall within expected parameters
relative to the healthy donor. Similar results were obtained when these analyses were
repeated with other parameters, including weighted UniFrac, Jensen-Shannon and root
Jensen-Shannon, and Bray-Curtis (data not shown).

Table 1. Pvalues of normalization and dynamic range of patient samples sets versus donor set