Trending ileal microbiome dysbiosis over time as an early assessment of intestinal transplant rejection risk

Abstract

Rejection is a barrier to intestinal transplantation (ITx). ITx rejection may be associated with changes in the ileal microbiome. We sought to analyze whether shifts in the microbiome were associated with intestinal transplant rejection. Ileal effluent samples were collected from ITx patients (n=8) with multiple samples taken from each patient at times of no (n=83), mild (n=39) or moderate (n=3) rejection, Crohn’s disease (CD, n=20), and non-inflamed control patients (NC, n=25). Ileal microbiota were quantified using 16S rRNA gene sequencing. Compared to non-transplant samples (NC, CD), ITx samples had lower alpha diversity (Shannon and Chao1, p<0.001) and different beta diversity (Bray-Curtis, p<0.005). Beta diversity differed between samples with and without rejection (p=0.002). Differential abundance analyses showed enrichment of pathogenic taxa and depletion of commensals in ITx rejection samples. ITx rejection is associated with ileal microbiome dysbiosis, which is a potential target for diagnostic and therapeutic interventions.

1). Introduction:

Intestinal transplantation (ITx) remains the definitive treatment for irreversible intestinal failure but is associated with the highest rejection rates among solid organ transplants¹. Increased rejection risk requires elevated levels of immunosuppression, which predisposes recipients to infections, malignancies, and graft-versus-host disease^2,3. Moreover, intestinal rejection often presents with minimal or non-specific symptoms that can make a clinical diagnosis difficult, and confirmation by histopathological examination is non-specific and at risk of false positives⁴. To make ITx a more viable therapy, better diagnostic and therapeutic modalities are needed. Research from our group and others suggests that the intestinal microbiome could be a target for emerging intervention strategies.

Dysbiosis of intestinal microbiome impacts many health conditions, including obesity, inflammatory bowel disease, and arthritis⁵. Studies have documented the importance of microbiome dysbiosis in Crohn’s disease (CD)^6–8, which shares similar pathophysiology with ITx rejection^9–12. Characteristic changes in the gut microbiome have been described in the setting of both bone marrow transplantation (BMT)^13,14 and solid organ transplantation^15–18, including a few small studies in ITx^19,20. It is speculated that the relative abundance of specific bacterial species may be associated with rejection through the depletion of commensal bacteria that, under normal conditions, either outcompete pathogenic bacteria or are directly beneficial to the mucosa, such as those that secrete short chain fatty acids (SCFAs)^14,21–25. Alternatively, an increase in pathogenic bacteria, such as multi-drug resistant (MDR) strains, invasive strains, and strains that secrete toxic metabolites may be associated with rejection^15,19,26,27. Besides changes in the relative abundance of microbial species, data from kidney¹⁷, liver¹⁵, and intestinal²⁰ transplantation have associated solid organ rejection with an overall decrease in bacterial richness (alpha diversity). Therefore, we aimed to combine immunological analysis from our human ITx cohort with novel microbial and mucosal analysis to investigate the effect of the graft microbiome on rejection after ITx.

2). Materials and Methods

2.1. Study design

This is a prospective cohort study investigating samples from patients treated at a single academic center between 2020 and 2023. The transplant cohort included recipients of deceased donor isolated intestine or intestine-containing multivisceral allografts. For comparison, we also collected samples from patients with and without defined ileal inflammation. These consisted of patients with active ileal Crohn’s disease (CD) and non-inflamed control (NC) patients undergoing ileocolic resection for cancer.

2.2. Sample collection

Approval from an internal ethics committee and institutional review board was obtained (IRB AAAJ5056, AAAT7647, AAAU1954). For intestinal transplantation, which included deceased donor transplantation of isolated intestinal allografts or intestine-containing multivisceral grafts, preoperative informed patient consent for adults and parental consent for pediatric patients were obtained. All patients received 7 days of piperacillin/tazobactam in the immediate postoperative period as a prophylactic antimicrobial course. Ileal effluent samples were collected from the ileostomy at predetermined timepoints (post-operative days [POD] 0, 3, 7, 14, 21, 28, 42, 50, 60, 90, 120, 183, 365, 730) and at times of for-cause endoscopic mucosal biopsies. One to two mucosal biopsies were collected using cold forceps during screening endoscopies and placed into sterile normal saline or Hanks’ balanced saline solution (HBSS, Corning, USA, Ref 21-022-CV). Rejection status for each sample was determined by histopathological review of mucosa collected at a similar time using commonly accepted criteria²⁸. Of note, all ITx patients in this cohort had at least one rejection episode, and there were no true non-rejecting individuals in this cohort. Samples that were indeterminate for rejection or between categories were retrospectively reviewed to determine clinical diagnosis.

Informed patient consent was obtained for a consecutive series of patients undergoing ileocolic resection for severe Crohn’s disease (CD) or for non-inflammatory diseases, which was usually ascending colon cancer (these were considered “non-inflamed control” (NC) subjects since they lacked inflammation in the ileum). All potential subjects during this time were approached by the study team if the clinical team approved, and no patient was intentionally excluded for any reason. Surgically resected specimens were received in a sterile container. Ileal effluent from the proximal end of the specimen was collected into a new sterile container. Using sterile tools, one full-thickness (1 cm²) sample and nine mucosal samples (5 mm²) were collected from the terminal ileum into a sterile container containing normal saline. Samples were then transported to the laboratory.

2.3. Ileal effluent storage and processing

Ileal effluent samples were transported immediately to the laboratory at room temperature. Each sample was portioned into at least three 0.5 mL aliquots in cryovials. At least one sample was aliquoted for the Microbiome Core facility at Columbia University Irving Medical Center and stored at −80°C prior to sequencing. The others were immediately transferred into liquid nitrogen (−180°C) for long-term storage in our laboratory.

2.4. DNA extraction, 16S rRNA microbiome sequencing, and culturing

Ileal effluent aliquots that had been stored at −80°C were thawed and processed using automated extraction platforms (epMotion 5075, Eppendorf; QIAcube HT, Qiagen) using the DNeasy^® 96 PowerSoil^® Pro QIAcube^® HT Kit (Qiagen, Germany). Nine samples were extracted manually using the DNeasy PowerSoil Pro Kit. A positive control (ZymoBIOMICS Microbial Community Standard, D6300; Zymo Research, USA) was included in each extraction batch. The ileal microbiota were profiled by targeting the 16S rRNA gene (V3-V4) using the 16S metagenomic sequencing library preparation kit (Illumina, USA) on the Illumina MiSeq platform (2 x 300 bp paired-end). Samples were sequenced in two batches (May 2022 and July 2023). Demultiplexed FASTQ sequences were quality-filtered, de-replicated, denoised and merged, followed by removal of chimeric sequences using DADA2 (v1.12.1) in R (v4.3.2) to generate amplicon sequence variants (ASVs). Sequences were aligned using the MAFFT and FastTree modules in QIIME2 to generate a phylogenetic tree, and taxonomic classification was conducted using a native naïve RDP Bayesian classifier aligned against the SILVA database (v138).

One hundred samples were additionally cultured on extended-spectrum beta-lactamases (ESBL) and vancomycin-resistant enterococci (VRE) CHROMager^™ plates (CHROMagar, France). Samples were then sub-cultured on MacConkey and BD BBL^™ (BD Biosciences, USA) to evaluate for the presence of positive colonies.

2.5. Statistical analysis

Basic descriptive statistics were performed using the Kruskal-Wallis test and logistic regression in R (v4.4), SAS Academics, and GraphPad prism (v10). Pre-processed sequencing data were integrated into R. To ensure sufficient sequencing depth for diversity analyses, samples that did not reach a plateau in rarefaction curve analysis were excluded. This corresponded approximately to a threshold of 2,000 reads. Seven samples included in analysis reached rarefaction but did not exceed 2,000 reads. Rarefaction was prioritized over a threshold of 2,000 reads to preserve the ability to analyze longitudinal data. Alpha diversity (i.e., within-sample microbial diversity) was assessed using the Chao1 (richness) and Shannon index (richness/evenness) indices and compared statistically between treatment groups (pooling all available data) and within each patient (trending values in longitudinal samples) using the Kruskal-Wallis test. Beta diversity (i.e., between-sample microbial dissimilarity) was assessed using Bray-Curtis dissimilarity and visualized by treatment group using Principal Coordinates Analysis (PCoA), controlling for patient ID in a subset of comparisons. Furthermore, pairwise PERMANOVAs adjusted for multiple comparisons using Bonferroni correction. Differential abundance analyses were performed using DESeq2²⁹ on all treatment groups and ZicoSeq³⁰ for ITx patients only, the latter additionally adjusting for patient ID. Statistical analyses were performed in R, SAS, and GraphPad Prism. Several important bacteria were explored further. Top bacteria that were most predictive of mild-to-moderate rejection were identified using random forest machine learning models based on Gini decrease and cross-referenced with top bacteria that were most associated with mild-to-moderate rejection using linear mixed models based on p values <0.25 in covariate-adjusted and unadjusted models. Bacteria identified by machine learning and linear mixed models were contextualized with bacteria of importance in relevant literature as well as with bacteria with highest relative abundance in our data ^18,31,32.

3). Results

3.1. Patient Demographics and Sample Characteristics

From 2020-2023, this study included eight consecutive intestinal transplant (ITx) recipients, 25 non-inflamed controls (NC), and 20 patients with Crohn’s disease (CD). Longitudinal sampling of ileal effluent for a total of 125 samples was carried out in ITx recipients at defined timepoints and also in the setting of suspected rejection. ITx samples were subdivided by rejection status: Negative (Neg) n=83, Mild n=39, and Moderate (Mod) n=3. Of the eight ITx patients, seven patients had mild rejection episodes, while two patients had moderate rejection episodes during the study period. There were no ITx patients in our cohort who did not have any rejection episode during the study period. All “negative” rejection samples in this cohort either precede or follow a rejection sample in a patient who had at least one rejection episode in this cohort. Subsequent analyses included fewer NC (n=24), CD (n=19) and ITx Negative (n=74; total ITx n=116) samples due to excluding samples that did not reach rarefaction plateau, as described. ITx patients had a median age of 12.1 years (IQR 5.4–29.2) and were predominantly female (63%). CD patients were older (median age 36.9 years, IQR 31.9–56.2) and 54% female, and NC patients were the oldest (median age 70.3 years, IQR 62.0–78.5) and 62% female, with significant differences in age among the groups (Kruskal-Wallis, p<0.001) (Supplemental Table 1).

3.2. Alpha Diversity: Decreased Species Richness in Transplant Samples

Alpha diversity differed significantly between ITx recipients compared to non-transplant patients (NC, CD), as demonstrated by both Chao1 and Shannon indices (Kruskal-Wallis test, p<0.001 and p<0.001; unadjusted p-values shown in Figure 1). Excluding samples immediately after broad spectrum antibiotic administration showed consistent findings (Supplemental Figure 1). Pairwise comparisons using Dunn’s test indicated significantly lower Chao1 in all ITx recipients (Neg, Mild, Mod) compared to both NC and CD patients (FDR-adjusted p<0.05). For the Shannon index, all ITx recipients (Neg, Mild, Mod) had lower diversity compared to NC patients, and significantly lower diversity in Neg ITx recipients compared to CD patients. Linear mixed effects models covarying for age, gender, and patient ID replicated these results, except for significantly lower Chao1 in CD patients relative to NC patients, and a non-significant difference in Shannon diversity between NC and Neg ITx patients. Pairwise analysis indicated nonsignificant trends toward lower alpha diversity with increasing inflammation severity (rejection status and CD; Figure 1). Logistic regression analysis adjusting for postoperative day (POD) and total bacterial count revealed nonsignificant trends toward reduced alpha diversity associated with rejection episodes (Shannon OR=0.59, p=0.18; Chao1 OR=0.60, p=0.16; Figure 2).

Figure 1.

Alpha diversity as approximated by Chao1 and the Shannon index. Overall group comparisons analyzed using Kruskal-Wallis test and Wilcoxon rank-sum test. Intestinal transplant samples: Neg, negative rejection; Mild, mild rejection; Mod, moderate rejection. CD, Crohn’s Disease. NC, Noninflamed control.

Asterisks indicate p-values such that * (p<0.05), ** (p<0.01), *** (p<0.001). NS = nonsignificant (p>0.05).

Figure 2.

Alpha diversity represented by Chao1 (left) or Shannon index (right) graphed over time by patient. Rejection episodes are marked by orange (mild rejection) or red (moderate rejection) markers. Sample collections that coincided with administration of broad-spectrum antibiotics were marked with a dashed line. Green vertical line indicates instances of viral infection. Light blue horizontal line indicates average Chao1 and Shannon alpha diversity for negative rejection samples. Green vertical line indicates instances of viral infection. Light blue horizontal line indicates average Chao1 and Shannon alpha diversity for negative rejection samples.

Blue asterisk indicates donor samples. Blue line indicates the median Chao1 and Shannon scores for negative rejection samples.

Intestinal transplant samples: Neg, negative rejection; Mild, mild rejection; Mod, moderate rejection.

To assess whether there was a difference between non-rejection ITx samples taken prior to (n=32) versus subsequent to (n=42) rejection events, subset analysis was performed comparing these samples. We found no significant differences between negative rejection samples taken before and after rejection episodes (Supplemental Figure 2, Supplemental Table 2).

3.3. Beta Diversity: Distinct Microbiota in Intestinal Rejection Compared with Non-Rejection and Crohn’s Disease

Beta diversity was assessed visually using Principal Coordinates Analysis (PCoA) as approximated by Bray-Curtis dissimilarity. Pairwise PERMANOVAs revealed significant differences in the intestinal microbiome between non-rejecting (Neg) ITx samples and those with mild rejection (R²=0.022, p=0.002, Figure 3, Supplemental Figure 3). Differences between negative and mild rejection remained significant after adjusting for patient ID (p=0.018; Figure 3g). In both cases, while the negative and mild rejection groups differed significantly, the moderate rejection group (Mod) was not significantly different from either, likely reflecting its limited sample size (n=3).Whereas comparison between NC and CD samples showed compositional microbiota differences in the setting of known inflammation (R²=0.05, p=0.008), the changes in the CD group were different from those in the mild rejection group (R²=0.087, p=0.001) and the moderate rejection group (R²=0.1, p=0.004), suggesting that the pathophysiology of ITx rejection is distinct from that of Crohn’s disease, at least from a microbial compositional standpoint (Figure 3). Longitudinal analysis of beta diversity did not show significantly increased dissimilarity during rejection episodes compared to baseline microbiota composition (Bray-Curtis dissimilarity: OR=11.9, p=0.66; Figure 4).

Figure 3.

Principal Coordinate Analysis (PCoA) of beta diversity approximated by Bray-Curtis dissimilarity. P-values are based on PERMANOVA analysis.

After adjusting for patient ID (g), negative rejection had different microbiota compared to mild rejection (p=0.002, padj=0.018). Comparisons including the moderate rejection group were not analyzed due to low sample size.

Intestinal transplant samples: Neg, negative rejection; Mild, mild rejection; Mod, moderate rejection.

CD, Crohn’s Disease. NC, Noninflamed control.

Figure 4.

Beta diversity as represented by Bray-Curtis dissimilarity graphed over time by patient. Rejection episodes are marked by orange (mild rejection) or red (moderate rejection) markers. Sample collections that coincided with administration of broad-spectrum antibiotics were marked with a dashed line. Subheadings indicate the reference sample from which beta diversity was calculated (relative to microbiome of donor, recipient at POD0, or recipient at first available POD).

Intestinal transplant samples: Neg, negative rejection; Mild, mild rejection; Mod, moderate rejection.

3.4. Relative Abundance and Differential Abundance

Compared to NC and CD, ITx samples demonstrated increased relative abundance of pathogenic genera (notably Klebsiella species and Escherichia-Shigella) and reduced abundance of beneficial commensals (e.g., Bifidobacterium; Figure 5). Differences by group were noted for comparisons of relative abundance of Enterococcus (p=0.002), Klebsiella (p<0.001), Streptococcus (p=0.020), Bifidobacterium (p<0.001), Bacteroides (p=0.039), and Veillonella (p=0.0132), which was confirmed upon analysis of absolute abundance (Supplemental Figure 4). Using Dunn’s pairwise analyses, group differences primarily were detected between negative rejection ITx samples compared to NC, such that negative rejection ITx samples had decreased relative abundance of Bacteroides (p=0.001), Bifidobacteria (p<0.001), and increased relative abundance of Enterococcus (p=0.019).

Figure 5.

Relative abundance of bacteria (genus level) by group.

Intestinal transplant samples: Neg, negative rejection; Mild, mild rejection; Mod, moderate rejection.

CD, Crohn’s Disease. NC, Noninflamed control.

Differential abundance analyses confirmed baseline microbiota differences between transplant and non-transplant groups, with significant enrichment of beneficial species such as Bacteroides dorei, Ruminococcus gnavus, and Roseburia spp. in NC compared to ITx samples (Figure 6). While CD was associated with increased abundance of some pathogenic genera, such as Granulicatella, it also had decreased abundance of other pathogenic species belonging to Klebsiella and Enterococcus, relative to ITx samples. Differential abundance analyses of Neg vs Mild alone (Mod was excluded due to small sample size), further adjusting for patient ID, revealed that mild rejection samples were significantly enriched for pathogenic taxa, such as Klebsiella, compared to negative rejection samples, while beneficial taxa (Lachnoclostridium) were significantly depleted (Supplemental Figure 5).

Figure 6.

Differential abundance analysis of bacteria by group compared to a baseline of intestinal transplant samples that were negative for rejection (ITx neg). Fold change increase (red) and decrease (blue) are represented by colors. Statistical significance of difference is represented by asterisks.

Intestinal transplant samples: Neg, negative rejection; Mild, mild rejection; Mod, moderate rejection.

CD, Crohn’s Disease. NC, Noninflamed control.

Exploratory linear mixed model analyses revealed elevated counts of pathogenic bacterial genera (Enterococcus, Escherichia, Klebsiella) associated with rejection episodes and ITx samples in general. Conversely, commensal bacteria (Bifidobacterium, Lactobacillus, Veillonella) were significantly associated with negative rejection and non-transplant samples (Supplemental Figure 6).

These effects appeared to compound over time for each patient. Linear regression analysis showed that after controlling for patient ID, with increasing POD, there was an increase in Enterococcus (p<0.001) and Klebsiella (p<0.001), and a decrease in Lactobacillus (p=0.002) and Veillonella (p=0.019). Other important bacteria, such as Escherichia-Shigella and Streptococcus, did not appear to change over time but differed by patient (both p<0.001).

3.5. Analysis of Microbial Resistance

Incidence analyses of bacteria harboring extended-spectrum beta-lactamases (ESBL) approached statistical significance (Fisher’s exact test, p=0.052), suggesting increased presence with higher rejection grades, while vancomycin-resistant enterococci (VRE) did not show significant differences (Chi-square, p=0.59; Supplemental Figure 7). Details of antibiotics administered over time by patient are shown in Supplemental Figure 8.

3.6. Association Between Microbes and Cytokine Axes

The ileal microbiome of NC patients tended to be characterized by higher abundance of taxa associated with anti-inflammatory pathways, whereas the microbiota of CD patients tended to feature fewer of these taxa and instead higher abundances of bacteria associated with the Th17 pathway than the Th1 pathway. Furthermore, ITx patients with rejection also featured fewer anti-inflammatory bacteria, and showed a higher abundance of taxa associated with the Th1 pathway than the Th17 pathway (Figure 6 and Table 1).

Table 1:

Relative abundance of bacteria in different states as organized by cytokine axis

		Th1	Th2	Th17/Tmem recruitment	Undefined/Variable pathways	Rejection	Anti-inflammatory/Commensal	Direct tissue injury
Non-transplanted intestine	NC		Bacteroides 1 Bifidobacterium longum 2	Citrobacter 3–5 Streptococcus gallolyticus 6 , 7	Alphaproteobacteria Granulicatella Staphylococcus caprae 8		Bacteroides dorei 9 Ruminococcus gnavus 10,11 Alistipes 12 Bacteroides vulgates 13 Roseburia 14 Bifidobacterium longum 2 Bacteroides thetaiotaomicron 15,16 Subdoligranulum 17 Bifidobacterium longum 2 Blautia 18 Faecalibacterium prausnitzii 19 Streptococcus salivarius 20 Lactobacillus 21 Lacticaseibacillus 22 Veillonella 23,24	Ruminococcus torques 22,25
	CD	E coli 26		Streptococcus 27–29 Actinomyces 30,31 Citrobacter3–5	Parasutterella Actinomyces graevenitzii Gemella Alphaproteobacteria Streptococcus cristatus Granulicatella Enterococcus faecium, non-VRE 32,33		Ruminococcus gnavus 10,11 Lacticaseibacillus 22 Veillonella 23,24
Intestinal transplant	Negative	E coli 26	Lactobacillus iners
	Mild	Klebsiella variicola 34 Klebsiella other species 21,35–41 Klebsiella pneumoniae 36,39,41,42		Actinomyces 30,31,43 Citrobacter 3–5	Lachnoanaerobaculum Lactobacillus iners		Parabacteroides distasonis 15 Lactobacillus 21,35,44
	Moderate	E coli 26 Klebsiella other species 21,35–41 Klebsiella pneumoniae 36,39,41,42		Streptococcus 27–29	Streptococcus oralis Enterococcus faecalis 45,46 Enterococcus faecium, non-VRE 32,33		Lactobacillus 21,35,44 Veillonella 23,24 Streptococcus parasanguinis 47,48 Lacticaseibacillus 22 Streptococcus salivarius 20

Red = Increased

Blue = Decreased

3.7. Impact of Stoma and Exposure to Oxygen

To assess whether the presence of a stoma affected the relative abundance of aerobes versus anaerobes, we compared the relative abundance of the top 13 most abundant bacteria in samples collected from a patients with a stoma (ITx samples, n=89) against samples collected from a surgical specimen in the absence of a stoma (CD and NC samples, n=53) (Supplemental Table 3). Pseudomonas was the only aerobic genus and did not differ between those with or without a stoma (p=0.089). Abundance of anaerobic bacteria showed no obvious association with a stoma, with some species more abundant in non-stoma samples (Bacteroides p<0.001, Bifidobacterium p<0.001), stoma samples (Veillonella p<0.001), or neither (Actinomyces p=0.43). Facultative anaerobic bacteria abundance was either elevated in stoma samples (Enterococcus p<0.001, Klebsiella p<0.001, Citrobacter p=0.049, Staphylococcus p=0.011) or not significantly different (Escherichia-Shigella, Streptococcus, Lacticaseibacillus, Lactobacillus).

4). Discussion

Overall, this study reveals characteristic differences in the microbial richness and composition of the ileal microbiome in patients after ITx, and specifically between those with and without rejection. Some of these results are consistent with findings previously published by other groups, which serves to validate our findings. Most recently, Zhang et al describe similar changes in the ileal microbiome during intestinal transplant rejection with enrichment by pathogenic bacteria, such as Enterobacteriaceae, and decreased bacterial diversity.³³ Together, our two independent cohorts, both using longitudinal 16S sequencing of microbiota in ileal effluent, support that intestinal microbial dysbiosis may occur around rejection episodes. Zhang et al sampled at predetermined intervals and showed that microbial dysbiosis can precede rejection and analyzed bacterial metabolites and mucosal damage to show the impact of dysbiosis. This study complements these findings by showing that these results are observable through a longer period (up to two years), are distinct from inflamed (CD) and noninflamed (NC) non-transplant comparators, and link microbial shifts to mucosal immune responses using mucosal mRNA analysis. Differences in the specific commensals depleted in both studies may reflect differences in study cohorts and methodology. When these additional comparison groups are incorporated, notable findings include: 1) lower alpha diversity in ITx (but not CD) relative to NC, with a trend towards decreased alpha diversity within individual ITx recipients at rejection timepoints. 2) Significant differences in overall microbiota composition between patients with and without ITx, including those with CD, as well as significant differences between ITx patients with and without rejection. 3) The predominance of specific bacterial species in ITx versus non-transplant samples, as well as ITx samples with and without rejection, but also notable differences between CD and ITx rejection samples. Intriguingly, but somewhat speculatively, our data also suggest that differences in relative abundance between groups might be associated with different underlying inflammatory processes. In a separate manuscript (currently under revision), we analyzed cytokine expression in the ileal mucosa of patients in this study. Relevant to the current study, we found that the Th17 axis appeared to be more strongly associated with CD, while the Th1 pathway was more strongly associated with ITx rejection. Microbial analysis, while not conclusive, is consistent with that finding. Our findings are somewhat unexpected since recent work has led to the consideration of CD and ITx rejection as similar pathophysiological processes^9–12. However, prior work has not directly compared these groups in side-by-side analysis.

The results of this study have the potential to lead to novel diagnostic and therapeutic strategies desperately needed in ITx. For example, decreased alpha diversity in ITx patients, including data suggesting that alpha diversity might specifically decrease further at rejection timepoints based on longitudinal samples from individual patients, raises the consideration of restoring microbial richness as a novel therapy at the time of ITx rejection. In that case, transplantation of ileal microbiota, similar to the existing fecal microbiota transplantation (FMT), could be a relatively low risk and high reward intervention. Beyond decreased alpha diversity, ITx rejection in our cohort is further characterized by other forms of microbial dysbiosis that could be utilized for diagnostic purposes or targeted by novel therapeutics, such as an overrepresentation of pathogenic taxa (Klebsiella, Escherichia-Shigella) in addition to a simultaneous depletion of beneficial commensals (Bifidobacterium, Faecalibacterium). Our findings that microbial dysbiosis is associated with rejection in ITx patients align with emerging data in other inflammatory gastrointestinal conditions, such as Crohn’s disease^34,35. These findings confirm that the gut microbiome may be a helpful, non-invasive adjunct to evaluating gastrointestinal health, if not a biomarker or target of modulation using oral probiotics³⁶, diet³⁷, and transplantation of intestinal microbiota, as previously suggested³⁸.

Limitations of this study include a relatively small sample size (consistent with the rarity of ITx), especially within the moderate rejection subgroup. There could also have been a confounding impact of antibiotic administration. While we cannot definitively assess the effect of antibiotics on the experimental results, we have thus far observed no significant difference in microbial richness (alpha diversity) between the CD and NC groups, despite recent antibiotic use in a large percentage of the CD (but not NC) patients (Supplemental Table 4). On the other hand, our preliminary data show significant differences in the microbial alpha diversity between the CD and ITx groups, despite large percentages of patients in both groups receiving antibiotics. These data suggest that antibiotics are unlikely to have been responsible for significant differences in our results. Additionally, effluent samples obtained from the ileostomy are uniquely exposed to air, which may have altered the gut microbiome³⁹, ultimately changing the alpha diversity and differential abundance^11,40,41. However, because all ITx patients underwent stoma formation at this institution during the study period, presence of a stoma was not considered a dependent variable. False positives for mild rejection could have decreased our ability to distinguish true differences in the microbiome between negative and mild rejection, however, this remains a gap in knowledge that exceeded the scope of this study. Notably, there were no patients without any rejection episode, and there was therefore no true nonrejector control group for comparison. This limits the generalizability of this study’s findings to other institutions and other settings. We emphasize that this remains an exploratory study of a nascent topic that must be investigated further before any attempts for practical use. Finally, our data regarding the association of taxa and cytokine pathways are tempered by two factors. First, while bacterial associations with a particular cytokine pathway were based on previously published associations, bacterial taxa can be involved in multiple immune pathways⁴². Those clearly associated with multiple pathways were categorized accordingly, however, there is likely plasticity for almost all bacteria. Second, not all taxa that were increased in each treatment group mapped uniformly to a particular cytokine pathway. Therefore, these data remain inconclusive. Future multicenter studies with larger cohorts, standardized sampling methodologies, and novel approaches to account for variations in antibiotic and immunosuppression exposure, and variations in presence of stoma would help support or clarify the shifts in the ileal microbiome observed in this study.

Abbreviations:

CD

Crohn’s disease

IBD

Inflammatory bowel disease

IRB

Institutional review board

ITx

Intestinal transplant

NC

Noninflamed control

POD

Post-operative day

Data availability:

All data (raw and analyzed) will be available upon request.