Antibiotic exposure, not alloreactivity, is the major driver of microbiome changes in hematopoietic cell transplantation

Abstract

Background:

Both autologous and allogeneic hematopoietic cell transplantation (auto and allo-HCT) are associated with significant alterations in the intestinal microbiome. However, the relative contributions of antibiotic use and alloreactivity to microbiome dynamics have not been elucidated. There is a lack of data regarding the kinetics of microbiome changes beyond 30 days post-transplant, and how they might differ between different transplant modalities. A direct comparison of differential effects of autologous and allogeneic HCT on the microbiome may shed light on these dynamics.

Objective:

To compare the intestinal microbial diversity between patients undergoing auto and allo-HCT from pre-transplant to 100 days post-transplant, and understand the effect of antibiotics, transplant type (auto vs allo) and conditioning regimens on dynamics of microbiome recovery.

Study Design:

We conducted a longitudinal analysis (starting pre-conditioning and at 14, 28, and 100 days post-transplant) of changes in the intestinal microbiome in 35 patients undergoing HCT (17 auto-HCT, 18 allo-HCT). Granular data regarding antibiotic exposure from day 30 pre-transplant to day 100 post-transplant were collected.

Results:

Pre-transplant, allo-HCT recipients had a lower α-diversity in the intestinal microbiome compared to auto-HCT recipients, which correlated with a higher pre-transplant antibiotic use in allo-HCT patients. The microbiome diversity declined at days 14 and 28 post-transplant in both cohorts, but generally returned to baseline by day 100. Conditioning regimen intensity did not significantly affect post-transplant α-diversity. Through differential abundance analysis, we show that commensal bacterial taxa involved with maintenance of gut epithelial integrity and production of short chain fatty acids were depleted after both auto and allo-HCT.

Conclusions:

In our dataset, antibiotic exposure was the major driver of post-transplant microbiome changes rather than alloreactivity, conditioning intensity or immunosuppression. Our findings also suggest that interventions to limit microbiome injury, such as limiting use of broad-spectrum antibiotics, should target the pre-transplant and not only the peri-transplant period.

Graphical Abstract:

Introduction:

Hematopoietic cell transplantation (HCT) is associated with severe intestinal microbiome injury. Microbiome injury has been associated with increased morbidity and mortality after both autologous (auto-HCT) and allogeneic (allo-HCT) transplant. ^1-7 The microbial dysbiosis is a result of multiple factors including mucositis secondary to conditioning therapy, prolonged depletion of immunocompetent cells, alloreactivity which affects the microbiome through damage to Paneth cells and other mechanisms, use of immunosuppressive regimens for extended duration, use of broad-spectrum antibiotics for prophylaxis and treatment of infections, and dietary changes in the post-transplant period. ^{4, 8-13} In particular, the contribution of alloreactivity to the changes in the microbiome is challenging to assess and a comparison of auto- and allo-HCT may help resolve this question. By design, the two transplant modalities lead to substantially different patterns of post-transplant immune dysregulation and immune reconstitution. There is a complex interplay between the immune system and gut microbiome, as perturbations of the microbiome directly alter host immune phenotype, and conversely deficiencies in host immunity alter the intestinal microbiome. ^13-18 A comparison between auto-HCT and allo-HCT transplant recipients may help determine the relative contribution of alloreactivity, antibiotics and conditioning regimens to microbiome dynamics after HCT.

Few studies have directly compared these two transplant modalities and have yielded mixed results. A large retrospective cohort study by Khan et al. showed a significant decrease in pre-transplant α-diversity in both auto- and allo-HCT. ^{19, 20} In a prospective observational study of 24 patients undergoing HCT, Kusakabe et al. showed no significant decrease in pre-transplant α-diversity compared to healthy controls, though the allo-HCT cohort had a lower abundance of Bifidobacterium and certain butyrate producing species. ²¹ Both studies analyzed the intestinal microbiome starting pre-transplant until 30 days post-transplant. There is limited data regarding the kinetics of microbiome changes post-transplant beyond day 30, and how it might differ between different transplant modalities. Understanding the differences in the pattern of microbiome injury and recovery in between the auto and allo-HCT cohorts, may allow development of tailored interventions (i.e. use of specific probiotics, antibiotic usage protocols) that reduce microbiome injury or accelerate recovery from it.

In the present study, we conducted a longitudinal analysis of stool microbiome changes that occur after auto- and allo-HCT through 100 days post-transplant. We collected highly granular data regarding exposure to antibiotics. We analyzed additional variables to examine their association with decreased microbial diversity including conditioning intensity and colonization with multidrug resistant bacteria, which correlates with greater prior antibiotics exposure and post-transplant mortality. ²² Finally, we performed a differential abundance analysis to examine how the peri-transplant microbiome composition differs in patients with an autologous or allogeneic donor and in patients who developed acute graft versus host disease (GVHD) following transplant.

Methods:

This study was approved by the Columbia University Irving Medical Center Institutional Review Board. All patients provided written informed consent for biospecimen and clinical data collection. This study was conducted in accordance with the principles outlined in the Declaration of Helsinki. We collected serial stool samples from 35 patients undergoing auto or allo-HCT with any conditioning regimen between April 2017 and December 2018. Levofloxacin was universally administered as prophylaxis starting 1 day before transplant. Stool samples were collected within approximately 1 month prior to transplant (range day −1 to day −36); 2 patients had pre-transplant samples collected earlier than day - 30. Post-transplant samples were collected on day 14, day 28 and day 100. We reviewed the electronic medical records of all patients in the study and recorded when they first received a particular antibiotic between day 30 pre-transplant (day −30) and day 100 post-transplant (day +100). All patients were screened for colonization with vancomycin-resistant enterococci (VRE), multidrug resistant Gram-negative bacteria (MDR GNB), and methicillin-resistant Staphylococcus aureus (MRSA) on admission and weekly thereafter while hospitalized as part of their clinical care. Data were also collected regarding Clostridium difficile infection and occurrence of acute GVHD.

DNA extraction, library preparation, and sequencing

We extracted DNA from stool samples by using the MagAttract PowerSoil DNA Kit and amplified the 16S rRNA V3/V4 region by using established primers with Illumina Nextera adaptors. ²³ Libraries were multiplexed by using Illumina Nextera XT Index kits, normalized, and pooled with 10% PhiX prior to sequencing on an Illumina MiSeq.

Demultiplexed FASTQ sequences were quality filtered, trimmed, dereplicated, and filtered for chimeric sequences using pair-ended DADA2 resulting in exact sequence variant (feature) tables. ²⁴ The table was imported into R 3.6.1 to analyze for α-diversity (Shannon), β-diversity (wunifrac) and were performed using a function of the phyloseq v1.28.0 package. ²⁵ Wilcoxon rank-sum test was used to analyze for statistically significant differences between the various cohorts.

On the basis of α-diversity rarefaction, we selected appropriate minimum feature count cutoffs of 10,000. β-diversity was analyzed using PERMANOVA (permutational multivariate analysis of variance) which is a non-parametric method to conduct multivariate ANOVA and determine if the centroids of sample clusters differ. The test statistic is calculated from the comparison of dissimilarities among interclass objects to those among intra-class objects. Differential abundance analysis for bacterial amplicon sequence variant (ASV)s was performed using DESeq2. ²⁶ The p-values were adjusted using the Benjamini-Hochberg method for controlling the false discovery rate.

Sequencing data are publicly available through the NCBI Sequencing Read Archive (SRA) (accession number PRJNA643833) after filtering any human-derived sequences.

Results:

Study population

Eighteen patients received an allo-HCT, and 17 patients received an auto-HCT. Detailed characteristics of patients in both cohorts are summarized in Table 1. Sixty-three percent of transplant recipients were male, and the median age at transplant was 61 years. In the auto-HCT cohort, 71% of patients had a plasma cell dyscrasia. In the allo-HCT cohort, 83% of patients had a myeloid neoplasm. Majority of patients in the allo-HCT cohort received myeloablative conditioning.

Table 1:

Patient characteristics

	Auto-HCT	Allo-HCT
Total number of patients	17	18
Median age at transplant (range)	62 (29-72)	59 (34-68)
Male sex n (%)	8 (47)	14 (78)
Diseases n (%)
Plasma cell dyscrasia	12 (71)
Systemic Sclerosis	1 (6)
NHL	4 (24)	2 (11)
AML		7 (39)
MDS		4 (22)
MF		3 (17)
CMML		1 (6)
T-PLL		1 (6)
Donor Source n (%)
Autologous	17 (100)
Matched Related		3 (17)
Matched Unrelated		8 (44)
Haploidentical		7 (39)
GVHD prophylaxis n (%)
Tac/MTX		10 (56)
Tac/MMF/PTCy		7 (39)
Tac/MMF		1 (6)
Conditioning intensity:
MAC	17 (100)	11 (61)
RIC	0 (0)	7 (39)
Acute GVHD grade 2-4
Within 100 days		6 (33)
Within 180 days		10 (56)
Median time to GVHD onset in patients with acute GVHD grade 2-4		53 (18-375)
Pre-transplant colonization
VRE	3 (18)	7 (39)
MDR GNB	2 (12)	2 (11)
MRSA	1 (6)	2 (11)

Abbreviations: auto-HCT – autologous hematopoietic cell transplantation; allo-HCT – allogeneic hematopoietic cell transplantation; Tac – Tacrolimus; MMF – Mycophenolate Mofetil; PTCy – Post-transplant cyclophosphamide; NHL – Non-Hodgkin Lymphoma; AML – acute myeloid leukemia; MDS – myelodysplastic syndrome; MF – myelofibrosis; CMML – chronic myelomonocytic leukemia; T-PLL – T cell prolymphocytic leukemia; MAC – myeloablative conditioning; RIC – reduced intensity conditioning; GVHD – graft versus host disease; VRE – vancomycin-resistant enterococci; MDR GNB – multidrug-resistant gram-negative bacteria; MRSA – Methicillin-resistant Staphylococcus aureus

Allogeneic transplant recipients have lower pre-transplant microbiome diversity compared to autologous transplant recipients

The pre-transplant intestinal α-diversity, as measured by the Shannon index was significantly higher in patients undergoing auto-HCT compared to allo-HCT (p=0.006). α-diversity decreased further at post-transplant days 14 and 28 in both auto and allo-HCT cohorts, but the auto-HCT cohort continued to have a higher α-diversity post-transplant and recovered faster compared to the allo-HCT cohort (Fig 1A). Both cohorts recovered to pre-transplant baseline by day 100 (Fig 1B). The recovery of α-diversity in the auto-HCT cohort post-engraftment (beyond day +14) was consistent across all patients and was characterized by a much smaller variance compared to the allo-HCT cohort at day 100.

Fig 1.

α-diversity as measured by the Shannon index. A) Except for day 14 time point, α-diversity is significantly higher pre- and post-transplant in the auto-HCT cohort compared to the allo-HCT cohort. B) α-diversity decreases significantly at days 14 and 28 in allo and auto-HCT cohort and recovers to pre-transplant baseline by day 100. α-diversity in the auto-HCT cohort recovers more uniformly with a smaller variance between patients. * signifies p-value ≤ 0.05, ** signifies p-value ≤ 0.01.

To identify potential etiologies for the differences in pre-transplant microbial diversity, we first examined antibiotic exposure (Fig 2A, 2B). Figure 2 shows the cumulative percentage of patients that were exposed to antibiotics between day −30 and day +100 peri-transplant. Excluding levofloxacin prophylaxis (generally started immediately prior to transplant), 18% of all patients in the auto-HCT cohort required antibiotics in the 30 days prior to transplant, compared to 56% of patients in the allo-HCT cohort (p=0.04). Because the pre-transplant sample was collected prior to conditioning, leading to variation in timing between patients receiving different conditioning regimens, we also compared antibiotic use prior to the pre-transplant sample collection. There was still greater use of antibiotics in the allo-HCT cohort (33%) compared to the auto-HCT cohort (12%). When analyzing the combined cohort, receipt of antibiotics prior to conditioning was significantly associated with a lower pre-transplant α-diversity (Fig 2C). The differences in antibiotic exposure between the auto and allo-HCT cohorts largely disappeared by day 100 post-transplant (82% vs 100% exposed to antibiotics excluding levofloxacin, respectively in patients with 100 days of follow up).

Figure 2.

shows the cumulative percentage of patients in auto and allo-HCT cohorts exposed to antibiotics from day −30 to day 100 post transplant. A) Cumulative percentage of patients exposed to any antibiotic excluding prophylactic levofloxacin. B) Cumulative percentage of patients exposed to a particular antibiotic. Anti-anaerobic agents plot denotes patients exposed to either piperacillin/tazobactam, metronidazole or meropenem. X-axis denotes day pre and post-transplant, and Y-axis denotes percentage of patients in all figures. Patients in the allo-HCT cohort had a greater overall antibiotic exposure, spread throughout their transplant course. The antibiotic exposure in the auto-HCT cohort was mainly limited to the 2 week period post transplant. C) In the combined cohort, pre-transplant alpha diversity was significantly lower in patients receiving antibiotics pre-conditioning and within 30 days prior to transplant, compared to patients that did not receive antibiotics pre-conditioning. * signifies p-value ≤ 0.05. x denotes the mean Shannon diversity. TMP-SMX – Trimethoprim-Sulfamethoxazole

We then compared the β-diversity, a measure of community composition, between groups. This measure generates a dissimilarity matrix, with pairwise distances calculated for every pair of samples. In principal component analysis, we observed significant differences between the auto and allo-HCT cohorts pre-transplant and at day 28 timepoint. The differences were no longer significant by day 100 (Fig 3). Overall these findings suggest that the impact of transplant on the early changes in the microbiome is similar in auto- and allo-HCT.

Figure 3.

The microbiome composition between auto and allo-HCT cohorts differed significantly at various time points as analyzed by β-diversity indices. The principal component analysis (PCoA) plots shown here were generated using unifrac generated distance matrix. PERMANOVA was utilized for statistical analysis of β-diversity. The microbiome composition between auto and allo-HCT cohort differed significantly pre-transplant, and at day +28 post-transplant.

Exposure to pre-transplant cytotoxic chemotherapy was not associated with increased pre-transplant antibiotic use or pre-transplant colonization by MDR organisms

Many cytotoxic chemotherapy agents cause gastrointestinal mucositis, which is a major contributor to infections, increased antibiotics use, and colonization by MDR organisms in patients undergoing HCT. ^{7, 27, 28} MDR colonization has previously been associated with increased post-transplant mortality. ²² We therefore analyzed whether treatment with cytotoxic chemotherapy pre-transplant correlated with pre-transplant antibiotic exposure and colonization by MDR organisms at the time of transplant (Supplementary table 1). We defined prior cytotoxic chemotherapy as multi-agent chemotherapy that included an anthracyclines or high dose alkylating agent (i.e. induction or salvage regimens for acute myeloid leukemia, lymphoma, or plasma cell leukemia). Most patients in the auto-HCT cohort did not receive cytotoxic chemotherapy pre-transplant, did not require antibiotics prior to transplant, and were not colonized by MDR organisms at the time of transplant. In the allo-HCT cohort, even patients who did not receive pre-transplant cytotoxic chemotherapy, received pre-transplant antibiotics and were colonized by MDR organisms at a high rate. There were no statistically significant association between the three variables. This may speak to differences in the underlying biology of the disease between the two cohorts, leading to higher antibiotic use prior to allo-HCT outside the context of chemotherapy induced mucositis.

Use of broad-spectrum antibiotics peri-transplant is associated with decreased intestinal microbiome diversity with differences in impact between auto- and allo-HCT

We then examined the effect of specific antibiotics on α-diversity in patients (Table 2). Per institutional protocol, piperacillin-tazobactam is the preferred first line antibiotic for neutropenic fever in auto-HCT patients, and cefepime is the preferred first line antibiotic in allo-HCT patients. Parenteral vancomycin was used in patients who remained febrile after treatment with first line antibiotics, had evidence of cellulitis, or in patients with gram positive bacteremia. This institutional practice is reflected in the antibiotic exposure plots in figure 2. Addition of further anaerobic coverage or gram-positive coverage or escalation to carbapenem was per provider discretion. Trimethoprim-sulfamethoxazole was preferred for prophylaxis against Pneumocystis jirovecii. Exposure to antibiotics beyond prophylaxis was nearly universal in both cohorts and despite our institutional practice regarding first line neutropenic fever management, cumulative exposure to antibiotics that affect anaerobic bacteria was similar in both cohorts. This allowed us to examine the possible impact of different antibiotic regimens on the loss of diversity during transplant by applying univariate analyses. In the univariate analysis of the auto-HCT cohort, use of piperacillin/tazobactam and parenteral vancomycin was associated with decrease in α-diversity at day 28 after HCT. In the allo-HCT cohort use of parenteral vancomycin was associated with decrease in α-diversity at day 14 and day 28 post-transplant. Cefepime was not associated with loss of α-diversity in either cohort.

Table 2:

Associations between antibiotic exposure (from day −30 to the indicated time point) and α-diversity (Shannon index)

Univariate analysis:

Auto-HCT
Antibiotic	Day 14		Day 28		Day 100
	Coefficient [95% CI]	p-value	Coefficient [95% CI]	p-value	Coefficient [95% CI]	p-value
Piperacillin/Tazobactam	−0.49 [−1.63 - 0.65]	0.38	−0.77 [−1.32 - −0.23]	0.01	0.16 [−0.39 - 0.71]	0.51
Vancomycin	−0.70 [−1.86 - 0.47]	0.22	−1.05 [−1.49 - −0.61]	<0.001	0.21 [−0.37 - 0.79]	0.44
Cefepime	0.09 [−1.38 - 1.56]	0.90	−0.39 [−1.66 - 0.89]	0.52	−0.15 [−1.07 - 0.76]	0.71
Meropenem	0.29 [−1.17 - 1.75]	0.68	−0.31 [−1.59 - 0.98]	0.61	0.05 [−0.88 - 0.97]	0.91
Allo-HCT
Vancomycin	−1.11 [−1.86 - −0.35]	0.01	−0.86 [−1.58 - −0.14]	0.02	−0.15 [−1.13 - 0.83]	0.75
Cefepime	−0.68 [−1.51 - 0.16]	0.10	−0.08 [−0.87 - 0.71]	0.84	−0.08 [−1.07 - 0.9]	0.86
Metronidazole	−0.45 [−1.3 - 0.39]	0.27	−0.39 [−1.09 - 0.31]	0.26	−0.11 [−0.95 - 0.73]	0.78
Meropenem	−0.24 [−1.27 - 0.78]	0.62	−0.44 [−1.2 - 0.32]	0.24	−0.04 [−1.02 - 0.95]	0.94
Trimethoprim-Sulfamethoxazole	−0.12 [−1.07 - 0.84]	0.80	0.31 [−0.47 - 1.09]	0.41	−0.25 [−1.05 - 0.55]	0.51

Cells with p≤0.05 are bolded. Coefficient correlates with the strength and direction of the effect on the Shannon index. At least three observations were required for analysis.

We then analyzed how exposure to different antibiotics affected the microbial community composition in each type of transplant. To this end, we analyzed the associations between antibiotic exposure at any time point between day −30 and day 100 and β-diversity using PERMANOVA (Table 3). This test assesses whether the overall microbiome community structure (presence and relatedness of taxa across samples) is significantly different across two groups based on the β-diversity distance matrix. In the allo-HCT cohort, several antibiotics (i.e., parenteral vancomycin, cefepime, metronidazole, meropenem, trimethoprim/sulfamethoxazole, and doxycycline) were associated with a significantly dissimilar β-diversity (PERMANOVA, p<0.05). In the auto-HCT cohort, only exposure to parenteral vancomycin was associated with a significantly dissimilar β-diversity (PERMANOVA, p<0.05). The differences between the cohorts in the impact of antibiotics on both α and β-diversity suggest that it is context specific. Given the small sample size, these findings will need to be confirmed in a larger cohort.

Table 3:

Association of antibiotics exposure with change in β-diversity in allo-HCT and auto-HCT. PERMANOVA test was performed for comparison microbial β-diversity (unifrac) between antibiotic exposure and no exposure at any time between day −30 and day 100. The p values are adjusted by Bonferroni test.

Antibiotic	P value (unifrac)
Auto-HCT
Piperacillin/Tazobactam	0.159
Vancomycin	0.001
Cefepime	0.35
Metronidazole	0.054
Meropenem	0.355
Trimethoprim/Sulfamethoxazole	0.291
Doxycycline	0.652
Allo-HCT
Piperacillin/Tazobactam	0.107
Vancomycin	0.001
Cefepime	0.002
Metronidazole	0.012
Meropenem	0.001
Trimethoprim/Sulfamethoxazole	0.049
Doxycycline	0.015

The bolded cells highlight p ≤0.05

Prior studies have shown that the use of cefepime is not associated with GVHD-related mortality compared to other broad-spectrum antibiotics that have broader activity against anaerobes ⁴, therefore our preferred antibiotic for neutropenic fever in allo-HCT recipients is cefepime. However, our findings demonstrate that it is clinically challenging to maintain patients on cefepime alone and additional antibiotics are frequently added or switched to (Fig. 2B). The impact of this management on the microbiota is not well studied. We found that allo-HCT recipients receiving antibiotics with broad-spectrum anti-anaerobic activity in addition to cefepime had lower α-diversity (coefficient: −0.215, p = 0.40). Therefore, we analyzed the differential abundance of various taxa in patients receiving cefepime with patients receiving broad-spectrum antibiotics with anaerobic coverage in addition to cefepime (Supplementary Fig. 1). We found that addition of other broad-spectrum antibiotics to cefepime was associated with reduction in taxa that have been linked with protection against GVHD, including Bacteroides, Parabacteroides, Enterobacteriaceae and Lachnospiraceae.²⁹ Notably, in our cohort patients receiving cefepime alone had a lower abundance of Blautia, which previously has been linked with increased incidence of acute GVHD.³⁰ In the auto-HCT cohort a minority of patients were exposed to cefepime, therefore a similar analysis was not feasible.

In allo-HCT recipients conditioning intensity was associated with a decrease in α-diversity, without reaching statistical significance

We analyzed how the conditioning regimen affects post-transplant α-diversity of the intestinal microbiome. Allo-HCT patients who received myeloablative conditioning had numerically lower α–diversity at day 14 compared to reduced-intensity conditioned patients without meeting statistical significance (p=0.25, Fig 4A) and both groups had similar recovery at later time points. In the auto-HCT cohort, we compared patients receiving high-dose melphalan conditioning with patients receiving BEAM (carmustine, etoposide, cytarabine, melphalan). Again, no statistically significant difference was seen in post-transplant α-diversity (Fig 4B). Given the small sample sizes, our conclusions should be confirmed in a larger cohort.

Fig 4.

A) Comparison of MAC vs RIC in allo-HCT patients B) Comparison of BEAM vs HDM in auto-HCT patients. C) α-diversity was lower in patients that developed grade 2-4 acute GVHD by day 100. A statistically significant decrease in α-diversity was seen at day 14, with recovery to pre-transplant baseline by day 100. * signifies p-value ≤ 0.05. MAC – Myeloablative Conditioning; RIC – Reduced Intensity Conditioning; BEAM - carmustine, etoposide, cytarabine, melphalan; HDM – High dose melphalan; GVHD – Graft versus Host Disease

In allo-HCT recipients, GVHD patients show an early loss of diversity with a quick recovery to baseline

Allo-HCT patients that developed grade 2-4 acute GVHD by day 100 had a statistically significant reduction in α-diversity at day 14 compared to patients that did not develop grade 2-4 acute GVHD (Fig. 4C). These differences were less prominent at day 28 and no longer present by day 100; patients with and without GVHD had similarly recovered diversity to pre-transplant baseline.

Depletion of bacterial taxa associated with intestinal epithelial wall integrity and production of short chain fatty acids is observed in both auto-HCT and allo-HCT

We then performed a differential abundance analysis to define specific bacterial taxa that are enriched or depleted in patients post-transplant. In the auto- and allo- HCT cohorts (Fig. 5A and 5B, supplementary fig. 2A and 2B, respectively), we compared the abundance of various taxa at days 14, 28 and 100 with their pre-transplant abundance. In both the allo and auto-HCT cohorts we saw a significant post-transplant decrease in the relative abundance of Faecalibacterium prausnitzii. This is a commensal species associated with amelioration of colitis and protection of gut epithelial wall integrity. ^{31, 32} We then examined the relative abundance of taxa involved in metabolism of short chain fatty acids (SCFAs), which have been previously shown to effect immune homeostasis in the gut, gut epithelial barrier integrity and protection against GVHD. ³³ We saw a significant reduction in SCFA metabolizers such as Ruminococcaceae and Blautia, in both auto and allo-HCT cohorts. Veillonela dispar, a commensal organism involved with nitrate metabolism, was enriched after both auto and allo-HCT. ³⁴ High serum nitrate levels have been associated with increased risk of steroid refractory GVHD and post-transplant thrombotic microangiopathy. ^{35, 36} Overall, we saw similar dynamics in auto and allo-HCT regarding taxa that have a potentially pathogenic role post-transplant, including taxa functionally linked to loss of gut barrier integrity, short chain fatty acid production and nitrate metabolism. This suggests that alloreactivity or the immunosuppressive drugs that mitigate it may not be the major factors driving post-transplant microbiome changes.

Fig 5.

Effect of donor source (autologous vs allogeneic) and graft versus host disease (GVHD) on relative abundance of various taxa after transplant. A) Change in relative abundance after autologous HCT. B) Change in relative abundance after allogeneic HCT. C) and D) Differences in relative abundance of various taxa at day 14 and day 28 respectively, in patients that developed grade 2-4 acute GVHD by day 180

In the allo-HCT cohort, we compared the microbiome of patients that developed acute GVHD grade 2-4 by day 180 with the microbiome of patients that had grade 0-1 acute GVHD. Comparisons were performed at each time point separately - pre-transplant, day 14, 28 and 100 (Fig. 5C and 5D, supplementary fig. 2C). The relative abundance of SCFA metabolizers was preferentially decreased in patients with grade 2-4 GVHD. These included Ruminococcaceae, E. dolichum, Bifidobacterium and Blautia. B. ovatus was increased in patients with GVHD. B. ovatus is important in production of succinate, which can act as a substrate for conversion to propionate by commensal taxa.³⁷ Succinate induces IL-1β during inflammation, and has been associated with worsening of acute GVHD.^38-40

Discussion:

The gut microbiome is profoundly injured during a hematopoietic cell transplant. Observational studies have shown that this injury is associated with adverse outcomes post-transplant. We hypothesized that auto and allo-HCT cohorts may have different patterns of microbiome injury and recovery kinetics. These differences are important to investigate in humans as they may shed light on the relative contribution of alloreactivity to the microbiome injury and allow development of interventions such as probiotics or antibiotics use guidelines that are adapted to the deleterious microbiome changes expected with a particular transplant modality. Prior studies have reported GVHD as a major contributor to changes in intestinal microbiome post-transplant. In pre-clinical studies, mice receiving an allo-HCT with a major MHC mismatch showed loss of Paneth cells and decreased expression of α-defensins. This was not seen in mice receiving a T-cell depleted allo-HCT. ^{9, 10} α-defensins are known to closely regulate the intestinal microbiome by selectively targeting non-commensal bacteria. ^{41, 42} Surprisingly, in our auto and allo-HCT cohorts, we saw significant overlap in the pattern of bacterial taxa that were enriched or depleted post-transplant (Supplementary Figure 2). Both cohorts showed a decrease in F. prausnitzii, which is a commensal organism in the intestinal microbiome with an important role in maintenance of the gut epithelial barrier. ⁴³ F. prausnitzii has been linked with upregulation of regulatory T cells and immunosuppressive cytokines. ^{44, 45} It is also a major producer of butyrate, which is an important energy source for colonocytes, has anti-inflammatory functions in the intestinal tract and promotes gut epithelial integrity. ^46-53 We also saw a decrease in post-transplant abundance of multiple taxa involved in the production of SCFAs in both the auto and allo-HCT cohort. Reduction of some SCFA producing taxa, such as Blautia, has previously been associated with increased GVHD related mortality. ² Though loss of SCFAs producing bacteria may exacerbate the impact of alloreactivity, our data suggests that alloreactivity is not the main contributor to their depletion.

Our data suggest that antibiotic use may be a bigger driver of changes in peri-transplant α-diversity than differential immune modulation due to differences between autologous and allogeneic donors. Allo-HCT patients had greater antibiotic exposure compared to auto-HCT patients, and it correlated with a lower pre-transplant α-diversity. Patients in the auto-HCT cohort showed faster recovery kinetics post-transplant and had a much smaller variance compared to the allo-HCT cohort at day 100, but ultimately both cohorts recovered baseline diversity. The different kinetics may again reflect cumulative antibiotic exposure in the allo-HCT cohort post-transplant. After allogeneic transplant, patients typically remain on immunosuppression beyond day 100 and are at an increased risk of infection for an extended period. In the auto-HCT cohort, the antibiotics exposure was concentrated between day 0 and 14, whereas patients in the allo-HCT cohort received antibiotics consistently from day 30 pre-transplant to day 100 post-transplant. This suggests that alloreactivity itself is not the primary cause for loss of diversity and it is overshadowed by increased antibiotic use in transplant patients, regardless of whether they receive immunosuppressive agents or not.

Recent studies have implicated loss of α-diversity with increased transplant related mortality in allo-HCT. ^{2, 3, 6} However, these studies typically track intestinal diversity up to the peri-engraftment period (less than 30 days). An important implication of our longitudinal analysis is that the damage to the intestinal microbiome happens early, and interventions to mitigate it should likewise start early, even during the pre-transplant period. In our analysis, addition of antibiotics with anaerobic coverage depleted bacterial taxa previously associated with GVHD protection (Supplementary Figure 1). In addition, parenteral vancomycin emerged as a variable associated with loss of α-diversity in both auto and allo-HCT. Broad clinical experience has shown that parenteral vancomycin does not have a therapeutic effect in treating intraluminal gastrointestinal (GI) infections. It has a narrow therapeutic index, and likely does not reach therapeutic intraluminal concentrations and physiologically relevant serum concentration. Parenteral vancomycin has been previously associated with increased risk of Clostridium difficile infections, independent of other broad spectrum antibiotics. ⁵⁴ In murine models, use of subcutaneous vancomycin caused substantial alteration of the gut microbiome. ⁵⁵ In humans, vancomycin was shown to be excreted via bile into stool after intravenous administration. ⁵⁶ A confounding factor may be that most patients received empiric vancomycin for gram positive coverage, but a variety of other antibiotics for gram-negative and/or anaerobic coverage. Despite this caveat, our analysis reinforces the need for better antibiotic stewardship. Meta-analyses have repeatedly shown that routine addition of glycopeptide antibiotics for empiric gram-positive coverage does not improve outcomes of febrile neutropenia. ^{57, 58} At our institution patients are routinely screened for methicillin-resistant Staphylococcus aureus colonization, which may aid in limiting empiric use of vancomycin.

Our findings conflict with a previous retrospective study that showed similar pre-transplant loss of diversity between the auto and allo-HCT cohorts. ^{19, 20} This may be explained by the different composition of the study groups. 76% of all patients in our cohort had plasma cell dyscrasia or systemic sclerosis, compared to 57% of such patients in the study by Khan et al. ¹⁹ Myeloma typically requires a lower intensity of traditional cytotoxic chemotherapy compared to lymphoma patients. Proteasome inhibitors and immunomodulators, frequently used in the treatment of plasma cell dyscrasias, attenuate experimentally induced colitis in murine models. ^59-61

In summary, our study shows that patients undergoing auto and allo-HCT have strikingly similar patterns of microbiome injury post-transplant. Antibiotic use is the major driver of decrease in microbiome diversity, rather than conditioning intensity or alloreactivity. The loss of diversity is most acute in the early peri-transplant period, suggesting that interventions to preserve the microbiome may have a narrow therapeutic window.

The major strengths of our study include collection of detailed data regarding peri-transplant antibiotic exposure and the longitudinal follow up till 100 days post-transplant. We also acknowledge the limitations inherent in the retrospective nature and small sample size of our study. The small sample size also limits our ability to do robust subgroup and multivariate analyses. There are also limitations attached to manual annotation of bacterial taxa. The vast majority of bacterial species have not been studied in detail, and there is paucity of data regarding their physiological role in the human microbiome. 16s rRNA gene sequencing provides a basic survey of the intestinal microbial communities. Metagenomics and metatranscriptomics may provide more physiologically relevant and actionable data. There may be institutional differences in antibiotic usage protocols based on local resistance patterns. As our data suggests that antibiotic exposure rather than alloreactivity, conditioning intensity or immunosuppression is the primary factor driving post-transplant microbiome changes may have broader implications to guide selection of antibiotic regimens in the pre-transplant and early peri-transplant periods.

Highlights:

• Pre- and peri-transplant antibiotic exposure is a stronger driver of loss of intestinal microbiome diversity in hematopoietic cell transplantation than donor source.

• There is a significant overlap in bacterial taxa that are altered post-transplant between autologous and allogeneic transplant recipients.

• Intestinal microbiome diversity is already low at baseline and recovers to pre-transplant by day 100 post-transplant. This suggests that clinical trials investigating intervention to prevent microbiome loss should target the pre or early post-transplant period.