Dynamic establishment of recipient resident memory T cell repertoire after human intestinal transplantation

Summary

Background

Understanding formation of the human tissue resident memory T cell (TRM) repertoire requires longitudinal access to human non-lymphoid tissues.

Methods

By applying flow cytometry and next generation sequencing to serial blood, lymphoid tissue, and gut samples from 16 intestinal transplantation (ITx) patients, we assessed the origin, distribution, and specificity of human TRMs at phenotypic and clonal levels.

Findings

Donor age ≥1 year and blood T cell macrochimerism (peak level ≥4%) were associated with delayed establishment of stable recipient TRM repertoires in the transplanted ileum. T cell receptor (TCR) overlap between paired gut and blood repertoires from ITx patients was significantly greater than that in healthy controls, demonstrating increased gut-blood crosstalk after ITx. Crosstalk with the circulating pool remained high for years of follow-up. TCR sequences identifiable in pre-Tx recipient gut but not those in lymphoid tissues alone were more likely to populate post-Tx ileal allografts. Clones detected in both pre-Tx gut and lymphoid tissue had distinct transcriptional profiles from those identifiable in only one tissue. Recipient T cells were distributed widely throughout the gut, including allograft and native colon, which had substantial repertoire overlap. Both alloreactive and microbe-reactive recipient T cells persisted in transplanted ileum, contributing to the TRM repertoire.

Interpretation

Our studies reveal human intestinal TRM repertoire establishment from the circulation, preferentially involving lymphoid tissue counterparts of recipient intestinal T cell clones, including TRMs. We have described the temporal and spatial dynamics of this active crosstalk between the circulating pool and the intestinal TRM pool.

Funding

This study was funded by the National Institute of Allergy and Infectious Diseases (NIAID) P01 grant AI106697.

Research in context.

Evidence before this study

Tissue resident memory T cells (TRMs) are conventionally characterized by their persistence and dominance in non-lymphoid tissues (NLTs), providing frontline defense against infection upon exposure to recurrent or persisting antigens. Recent studies in mice and limited human studies provided increasing evidence for a recirculating counterpart of TRMs. Previous investigation of human TRMs has been limited by a lack of access to longitudinal NLT samples. Intestinal transplantation (ITx) provides an outstanding opportunity to study human TRM repertoire establishment, as serial surveillance biopsies are taken from ileal allografts post-Tx for clinical monitoring. Our previous studies have demonstrated that T cell macrochimerism (peak level ≥4% donor T cells) in post-Tx peripheral blood is associated with slower replacement of intestinal mucosal T cells by the recipient and less rejection. We also found enriched host-versus-graft (HvG)-reactive clones, defined by pre-Tx mixed lymphocyte reaction (MLR) and high throughput T cell receptor (TCR) beta chain sequencing, in rejecting biopsies. However, the dynamic establishment of a recipient TRM repertoire in intestinal allografts and the origin, distribution, and specificity of intestinal allograft-infiltrating T cells has been largely unexplored.

Added value of this study

We performed multicolor flow cytometry and next generation TCR sequencing on blood, lymphoid tissues, and gut specimens from ITx patients, and found that: 1) Donor age ≥1 year and blood T cell macrochimerism were associated with delayed establishment of stable recipient TRM repertoires in the transplanted ileum. 2) TCR sequences identifiable in pre-Tx recipient intestine but not those detected only in pre-Tx lymphoid tissues were overrepresented in post-Tx allograft ileum. 3) TCR sequences identifiable in both pre-Tx recipient intestine and pre-Tx recipient lymphoid tissues had distinct transcriptional profiles compared to those identifiable in one tissue. 4) Paired gut and blood specimens from ITx patients showed higher TCR repertoire overlap than those in healthy controls, demonstrating increased gut-blood crosstalk after ITx. 5) Wide distribution of the same recipient TCRs was observed throughout the ileal and colon allografts and the recipients’ native colon. Some of these clones were identifiable as either alloreactive or microbe-reactive. 6) A higher detection rate of HvG alloreactive CD8 compared to non-HvG CD8 clones was observed in ileal allografts during quiescent periods in patients without macrochimerism, suggesting that these expanded HvG T cells contributing to the TRM pool may pose a constant risk of rejection.

Implications of all the available evidence

Our results provide novel spatial, temporal, and functional insights into the human TRM repertoire during quiescent periods after ITx. The understanding provided by our study of the mechanisms of human TRM repertoire establishment after ITx suggests approaches to improve immunosuppressive treatments following this lifesaving procedure. Furthermore, our study provides a unique and novel window into human TRM repertoire establishment, advancing our understanding of human adaptive immunity.

Introduction

Tissue resident memory T cells (TRMs) are characterized by their persistence and dominance in non-lymphoid tissues (NLTs), such as gut, lungs and skin,¹ and provide frontline defense against infection upon exposure to recurring or persisting antigens.2, 3, 4, 5, 6, 7 TRM can be distinguished by the high expression of retention markers such as CD69 and CD103 and low expression of lymph node homing markers such as CD62L and CCR7.⁸ T cells in NLTs have been divided into transiting, temporary and permanent subsets based on their tissue persistence,⁹ in line with emerging concepts of recirculating features of TRMs in both murine and human studies.10, 11, 12, 13, 14, 15 While strategies such as parabiosis surgery and in vivo intravascular antibody staining can reveal the migration patterns and resident features of TRMs in animal models,¹⁰ analysis of human TRM dynamics has been limited by a lack of access to longitudinal human NLT samples and lack of integration of phenotypes with T cell clonotypes.^16,17 Analysis of T cell receptor (TCR) repertoire establishment in human NLTs over time and space is needed to unravel TRM origin, tissue compartmentalization and persistence, and to interpret antigen recognition patterns of human TRMs in health and disease.

Intestinal transplantation (ITx) provides a unique opportunity to study human TRM repertoire establishment, as serial surveillance biopsies are taken from ileal allografts post-Tx for clinical monitoring and the gastrointestinal tract, unlike many other solid organ transplants, is richly endowed with immune cells under physiologic conditions.18, 19, 20 By utilizing human leukocyte antigen (HLA) allele group-specific monoclonal antibodies, we are able to distinguish donor- and recipient-derived T cells and track their chimerism and phenotypes post-Tx in the graft and in the blood. By combining pre-Tx 5,6-carboxyfluorescein diacetate succinimidyl ester-labeled mixed lymphocyte reaction (CFSE-MLR) with high throughput TCR-β complementarity-determining region 3 (CDR3) sequencing,²¹ we have defined and tracked alloreactive and non-alloreactive T cell sequences across time and space to address biologically relevant events at clonal levels in multiple types of human organ transplantation, including conventional kidney transplantation,²¹ combined kidney and bone marrow transplantation,²² liver transplantation,²³ and ITx.24, 25, 26

The dynamic landscape of the TCR repertoire in intestinal allografts from early to late periods post–Tx has not previously been explored. We previously demonstrated that high levels of donor T cell chimerism in peripheral blood of ITx recipients (peak level ≥4% donor T cells, defined as “macrochimerism”) is associated with slower replacement of intestinal mucosal T cells by the recipient, reduced rejection rates and reduced de novo development of donor-specific antibodies (DSAs).25, 26, 27 Our current study investigated the dynamics of establishment of recipient TRM repertoires in the allograft of ITx patients with and without blood T cell macrochimerism. We also explored the clonal distribution and potential antigen recognition patterns of recipient T cell repertoires, as well as the crosstalk with their circulating counterparts. Besides providing novel insights into human TRM establishment and dynamics, our study demonstrates that recipient intestinal T cells with a circulating counterpart preferentially contribute to the TRM pool in human intestinal allografts. Differential mRNA expression between cells expressed in both gut and lymphoid tissue compared to those detected in only one tissue identifies distinct features of the migratory subsets. We also demonstrated a significant contribution of host-vs-graft alloreactive T cells and of microbe-reactive T cells to this repertoire.

Methods

Study design

Sixteen ITx patients (Table S1) with both pre- and post-Tx specimens available to perform alloreactive TCR clonal tracking up to 1849 days post-Tx in ileal allografts were prospectively enrolled into this study, including one patient who underwent a re-transplantation (Pt16’’: second Tx of Pt16. The first Tx of Pt16 (Pt16′) was not included because post-Tx ileal specimens were unavailable for research.). Nine of the patients (Pts7, 10, 15, 16″, 18, 19, 21, 22, and 23) underwent multivisceral transplantation. Seven of the patients (Pts4, 9, 13, 14, 17, 20, and 24) received isolated ITx. Our cohort represents 16.7–19.6% of the total number of intestinal transplants performed in the United States each year and includes both pediatric and adult intestinal recipients.^28,29

Human subject recruitment and clinical protocols

Protocol graft biopsies were taken during the initial post-Tx period for clinical monitoring (ideally twice weekly for the first month, weekly from 1 to 3 months, biweekly from 3 to 6 months, monthly from 6 to 12 months until ileostomy closure), and additional biopsies were conducted for cause.²⁶ When available, blood samples were taken up to four times during the first month post-Tx and at least once per month thereafter. Multiple types of recipient tissue were collected up to July 2021, including pre-Tx spleen, lymph nodes (LNs), explanted gut tissues, and pre- and post-Tx peripheral blood mononuclear cells (PBMCs), ileal and colon biopsies (transplanted), and native colon biopsies (Fig. S1). A total dose of 6–10 mg/kg anti-thymocyte globulin was administered to all the patients, then a maintenance regimen including long-term tacrolimus and steroids for approximately 3 years was given. Tacrolimus was started on the first day post-Tx, adjusted to aim for a target trough level of 15–20 ng/mL in blood during the first two months, and then tapered down to a maintenance level of 5–15 ng/mL. Patients were given two methylprednisolone boluses on day 0 and, beginning on day 1, a dosage of 10 mg/kg/day. After that, the dose was tapered to a maintenance dose of 3–5 mg/day by 6–9 months, and then it was discontinued by 24–36 months. Depending on the degree of rejection determined by histology and in consideration of clinical symptoms and endoscopic findings,³⁰ increased immunosuppression was used. Typically, there was no treatment when biopsies were indeterminate for rejection. For histologically mild rejection with symptoms, methylprednisolone (adults: 0.5–1 g, children: 20 mg/kg) was used for 2 days, combined with a 33–50% increase in tacrolimus above prior levels. For histologically moderate rejection, 1.5 mg/kg thymoglobulin with a 33–50% increase in tacrolimus above prior levels was applied. Epidemiological and clinical characteristics of patients can be found in Table S1. All the available quiescent samples, namely those lacking a pathological diagnosis of moderate or severe rejection,³⁰ graft versus host disease, infection or other clinical or pathological immunological events, were included in this study without further selection.

IEL and LPL isolations

In cases with endoscopic biopsies or stoma closures/revisions, intraepithelial lymphocyte (IEL) and lamina propria lymphocyte (LPL) were isolated from specimens using a methodology previously described for flow cytometric analysis.24, 25, 26^,31 Briefly, the specimens were stirred constantly for 20 min with 2 mmol/L dithiothreitol in 1x HBSS buffer at 37 °C, followed by a 30-min incubation with 0.5 mmol/L EDTA in a 37 °C water bath to isolate IEL. LPLs were separated from the residual tissue in collagenase-containing media (RPMI 1640, 1 mg/mL Collagenase D, 100 IU/mL penicillin-streptomycin) in a 37 °C water bath for 1 h. DNAse (0.1 mg/mL) was added to the EDTA and collagenase medium when IEL and LPL isolation was performed on large amounts of mucosal tissue available from stoma closures/revisions.

Flow cytometric staining

Based on clinically accessible molecular HLA typing information, candidate monoclonal HLA class I allele-specific antibodies were tested for their capacity to distinguish donor and pre-Tx recipient cells. Each HLA-specific monoclonal antibody was evaluated for specificity in combination with a pan–HLA Class I antibody (HLA-ABC). Those that distinguished donor from pre-Tx recipient PBMCs were included in lineage-specific antibody panels, as described previously.^24,27 Flow cytometry antibodies used in this study are summarized in Table S2. Data were collected using the BD Biosciences LSR II flow cytometer and DIVA software. FlowJo v9.9.6 and v10.8.0 were used for data analysis. Representative gating strategies for recipient chimerism and tissue resident memory marker expression proportions are shown in Fig. S2.

CFSE-MLR, cell sorting and TCR-β CDR3 DNA sequencing

CFSE-MLR and cell sorting were performed as described.^21,24, 25, 26 In brief, using thawed pre-Tx recipient and donor cells from spleen or lymph nodes, a host-versus-graft (HvG) MLR was performed. A quantity of 200,000 violet-dye-labeled irradiated (35 Gy) stimulators from pre-Tx donor cells and 200,000 CFSE-labeled responder cells from pre-Tx recipient specimens in MLR medium (AIM-V supplemented with 50 μm 2-mercaptoethanol, 0.01 M HEPES, and 5% heat-inactivated human serum) were plated in each well of a round bottom 96-well plate. Before fluorescence activated cell sorting (FACS) using a BD Influx cell sorter, cells were labeled with anti-CD3, CD8, and CD4 antibodies to identify two distinct violet dye-negative DAPI⁻ CD45⁺ cell populations (CD3⁺CD8⁺CFSE^low, CD3⁺CD4⁺CFSE^low), indicating CD8⁺ and CD4⁺ recipient-anti-donor-reactive T cells (stimulated populations: stim). For unstimulated cell populations (unstim), pre-Tx recipient and donor cells were thawed and stained with anti-CD45, CD3, CD8, and CD4 antibodies and DAPI before being FACS sorted into DAPI⁻CD45⁺CD3⁺CD8⁺ and DAPI⁻CD45⁺CD3⁺CD4⁺ populations. For Pts4, 7, 9, 10, 13, 14, 15, 16″, 17, 18, 19, 20, and 21, genomic DNA was extracted using the Qiagen blood and tissue kit from sorted PBMC and intact endoscopic tissue pieces, which included both IEL and LPL populations, as described previously,²⁵ and frozen at −20 °C and shipped to Adaptive Biotechnologies (Seattle, WA) on dry ice for high-throughput TCR-β sequencing. The TCR sequencing data were retrieved from Adaptive's ImmunoSEQ software. For Pts22, 23, and 24, targeted cells were directly sorted into cell lysis buffer (Qiagen, catalog 158906) and transported to the University of Pennsylvania at room temperature for DNA extraction and high-throughput TCR-β sequencing as described.²⁵

TCR-β CDR3 data processing and analysis

Adaptive's ImmunoSEQ platform was used to extract TCR sequencing data for Pts4, 7, 9, 10, 13, 14, 15, 16″, 17, 18, 19, 20, and 21. Raw sequences for Pts22, 23, and 24 were quality filtered and their clone assemblies were processed with MiXCR (v. 3.0.7) and VDJtools (v1.2.1) as described previously.32, 33, 34, 35 Unique TCR-β sequences are defined by the combination of CDR3 sequence plus V and J genes at the nucleotide level. CD8 versus CD4 sorting error was resolved and donor- and recipient- shared CDR3s were removed as described previously.^25,36 Alloreactive clones were defined as having a 2-fold or greater expansion in stimulated versus unstimulated CD4 or CD8 pre-Tx repertoires, as well as a minimum frequency of 0.001% in CFSE^low populations when using read counts, or 0.002% in CFSE^low populations when using template counts, which ensures 85% repeatability as determined by power analysis.^21,36

Single cell RNA sequencing (scRNA-seq) data processing and analysis

The 10x Genomics (Pleasanton, CA) platform was used for mRNA expression and paired V(D)J TCR sequence measurement at the single cell level. For scRNA-seq analysis, frozen cells isolated from gut or spleen samples were thawed, washed, and sorted separately for T cells with the BD Influx sorter. Manufacturer's protocols were used for single cell libraries. For 5′ gene expression sequencing (5′ GEX-seq), libraries were sequenced on an Ilumina NovaSeq 6000 platform. TCR-seq were enriched by the V(D)J enrichment kit from 10x Genomics and sequenced on an Ilumina 550 Sequencer. The FASTQ files were processed using the 10x Genomics cloud-based pipeline Cell Ranger Count v7.0.0 with GRCh38 2020-A transcriptome as the reference. Seurat pipelines were used for scRNA-seq data quality control.37, 38, 39 Cells with less than 15% mitochondrial counts and unique feature counts ranging between the first quartile − 1.5 ∗ IQR and third quartile + 1.5 ∗ IQR (IQR: interquartile range, third quartile − first quartile) were selected. Feature expression was normalized by the total expression in each cell. Up to 20000 variable features were selected before data integration. Data integration was performed to generate heatmap and Uniform Manifold Approximation and Projection (UMAP) plots, following Seurat pipelines. Volcano plots were generated using R package EnhancedVolcano (v1.18.0) to visualize the differential expressed genes.

Ethics

The collection and use of patients’ samples were approved by the Columbia University Institutional Review Board under the following protocols: AAAJ5056, AAAS7927 and AAAF2395. Written informed consent was obtained from all participants or their legal guardians.

Statistics

R (R-4.1.2) and Rstudio (2022.02.0) were used to analyze TCR-seq data and generate circle plots. GraphPad Prism (v 9.3.1, GraphPad Software) was used to produce figures. Log-rank test was performed for the Kaplan–Meier plot showing proportion of patients with >50% recipient CD69⁺/CD103⁺ replacement in groups with or without blood macrochimerism. A Shapiro–Wilk test was performed to determine normality of data distributions. For data sets that met the assumption of a normal distribution, one-way ANOVA test followed by Tukey's multiple comparisons test was performed for comparisons of means among three or more groups. Ordinary one-way ANOVA was performed in most cases, except when comparing related subgroups collected from the same patient on the same day, such as “gut only”, “lymphoid only”, “gut-lymphoid shared” subgroups, and “non-shared”, “double-shared”, “triple-shared” subgroups, repeated measures (RM) one-way ANOVA was performed. For the two paired data sets that didn't meet the assumption of a normal distribution, a non-parametric Wilcoxon test was performed to compare the median ranks. A non-parametric Kruskal–Wallis test followed by Dunn's multiple comparisons test was performed for comparisons among three or more unpaired independent groups that didn't meet the assumption of a normal distribution, to compare the median ranks. For three or more independent groups with paired data that didn't follow the assumption of a normal distribution, a non-parametric Friedman test followed by Dunn's multiple comparisons test was performed. To reduce sampling limitation and dispersion, a stringent threshold was applied to all samples by only including samples with TCR template counts (read counts for Pts 22–24) >800 and only samples lacking evidence for active rejection or infection are considered here.

Role of the funders

The funders had no role in the conduct, analysis, interpretation, or publication of the study.

Results

Graft-infiltrating recipient T cells gradually establish a stable repertoire with TRM phenotypes in the ileum allograft post–Tx

We previously showed that repopulation of recipient T cells in intestinal allografts in patients with macrochimerism was slower than that in patients without macrochimerism when donor age was ≥1 year (1Y).²⁵ Donor-derived graft T cells maintained high expression levels of TRM markers (CD69⁺ CD103^+/−).²⁶ In contrast, graft-infiltrating recipient T cells initially lacked TRM markers but gradually acquired typical TRM phenotypes after the transplant.²⁶ The dynamics of this phenotypic transition, which was most complete among CD8 IELs, were comparable between patients with and without macrochimerism (CD69, p = 0.8715; CD103, p = 0.9089) (Fig. 1a and b). Dynamics of acquisition of TRM phenotypes by recipient CD4 IELs and CD4 and CD8 LPLs showed an overall similar pattern to that of CD8 IELs (Fig. S3).

Fig. 1.

Phenotypic and clonal analysis of graft-infiltrating recipient T cells in the ileum allograft post-Tx. (a) Representative flow cytometric (FCM) data from graft-infiltrating recipient CD8 IELs in patients without (−) (Pt20) and with (+) (Pt13) macrochimerism illustrate the gradual acquisition of TRM phenotypes (CD69⁺CD103⁺) post–Tx. (b) Percentages of patients with >50% of recipient CD8 IELs expressing CD69 and CD103 over time post-Tx among patients without (Pts9, 10, 14, 20) and with (Pts13, 18, 23) macrochimerism. Pts18, 20, 23 are newly enrolled patients in addition to Pts9, 10, 13, 14, whose results have been published previously.²⁶ Patients with ≥ six post-Tx time points by POD250 are included. Log-rank test was performed to determine statistical significance. (c) Dynamic establishment of TCR repertoires in ileal allografts in patients without (Pt20) and with (Pt13) macrochimerism. Constitution from pre-Tx donor- or recipient- mappable clones and clones first identified at designated PODs is shown. Pre-Tx donor and pre-Tx recipient sequences were defined from the donor or recipient spleen, LN, and gut samples (for Pts16″, 18, 19, 20, 21 only), which were collected pre-Tx whenever each type of tissue was available. “Putative de novo sequences” are defined as sequences first identified at the designated time point but not previous time points and are located on the top section of each bar, and recurring sequences are arranged below the “putative de novo sequences”. (d) Proportion of recurring sequences in ileal allografts post-Tx by cumulative frequency (unique sequences weighted by copy numbers). Open symbols indicate that >50% of T cells in the sampled allograft IELs had been replaced by the recipient according to FCM data. (e) Stability of dynamically established TCR repertoires in ileal allografts post-Tx was measured by changes of Jensen-Shannon Divergence (JSD) values between adjacent time points.⁴⁰ JSD values range between 0 and 1, where 0 indicates identical repertoires and 1 indicates complete divergence. Open symbols indicate that >50% of T cells in the two adjacent time points' ileal IELs had been replaced by recipient as assessed by FCM. For all plots in D and E, left dotted line: POD400, right dotted line: POD1200. Comparison of recurring sequence proportions (f) and JSD values (g) in ileal allografts with <20% and >80% recipient replacement in IELs measured by FCM. For plots in f, solid horizontal bars indicate the mean of cumulative frequency of recurring TCR sequences, while the dotted horizontal bars indicate the median. For plots in g, dotted horizontal bars indicate the median JSD values. (Ordinary one-way ANOVA followed by Tukey's multiple comparisons was performed for the left plots of panel f. Kruskal–Wallis test followed by Dunn's multiple comparisons was performed for the left plot of panel g and the right plots of panel f and panel g. ∗p < 0.05, ∗∗p < 0.01).

To address the relationship between TRM phenotypes and TCR repertoire stability and diversity, we stratified post-Tx T cell repertoires in quiescent ileal allografts (sequenced from intact endoscopic tissue pieces including both IEL and LPL populations) into different subsets according to the first time point at which each TCR was detected (Fig. 1c and Fig. S4), including pre-Tx (mappable to pre-Tx donor and recipient repertoires) and multiple post-Tx time points for each patient. We found that newly detected TCR sequences were continuously added to graft repertoires up to 5 years post-Tx. However, a significant fraction of TCR sequences identifiable at earlier time points were repeatedly detectable at later time point(s), indicating long-term persistence and distribution within the grafts.

When cumulative percentages of TCRs identified at earlier time points in the graft (referred to as “recurring” sequences) were summed, patients without macrochimerism showed a rapid increase around POD100–400 (Fig. 1d, left panel), a period when greater than 50% of donor T cells in the graft were replaced by recipient T cells, as shown by flow cytometric (FCM) data on IELs (Fig. S5). This accumulation of recurring sequences was reflected in a rapid decline in Jensen-Shannon Divergence (JSD) values (from 1.0 to 0.4) between adjacent timepoints from POD100–400 in this group of patients, followed by a relatively constant level between 0.5 and 0.7 at later follow-up times when available, indicating the establishment of relatively stable recipient repertoires (Fig. 1e, left panel).

Patients with macrochimerism showed different kinetics of intestinal mucosal repertoire turnover depending on the age of the intestinal donor. In patients with macrochimerism whose donor age was ≥ 1Y, stable TRM repertoires detected early post-Tx reflected the persistence of donor T cells and the delay in their replacement by recipient T cells (Fig. 1d–g). When donor replacement was <20%, high proportions (mean: 50.14% ± 19.55%) of recurring (donor) TCR sequences were seen compared to proportions in patients with macrochimerism and donor age <1Y (p = 0.0041) (Fig. 1f, left panel) and JSD values were relatively low (mean: 0.66 ± 0.16), though not statistically different from those in patients with macrochimerism and donor age <1Y (p = 0.1297) (Fig. 1g, left panel). As recipient T cells slowly replaced those of the donor (approximately days 400–1200, Fig. S5), repertoire stability declined and JSD values tended to increase (Fig. 1d and e, middle panels). “Recurring” recipient sequences started to accumulate in association with decreasing JSD values after POD1200 (Pt16″ and Pt15) when >50% of donor IELs had been replaced by the recipient (Fig. 1d and e, middle panels). At later timepoints, when recipient T cells made up >80% of cells in the ileum, the TCR repertoire of patients with macrochimerism and donor age ≥ 1Y showed relatively high frequencies of recurring clones (mean: 48.20% ± 24.00%) and relatively low JSD values (mean: 0.69 ± 0.14) compared to patients with macrochimerism and donor age <1Y (p = 0.0259) between adjacent timepoints (Fig. 1f and g, right panels).

For patients with macrochimerism whose donor age was <1Y, rapid population of the graft by recipient T cells (Fig. S5) was associated with marked repertoire instability and high JSD levels (mean: 0.88 ± 0.08) between adjacent timepoints in the first 400 days post-transplant (Fig. 1d and e, right panels). Despite this rapid replacement by the recipient, “recurring” sequences did not increase to levels seen in patients without macrochimerism within the first >400 days (Fig. 1d) and JSD values remained >0.8 for all patients in the group until a decrease occurred between POD800–1200 in Pt21 (Fig. 1e, right panel). When considering only timepoints at which >80% of IEL T cells were recipient-derived, this group of patients still showed greater repertoire instability compared to those with macrochimerism and donor age ≥1Y (cumulative frequency of recurring TCR sequences, p = 0.0558; JSD values, p = 0.0259) and showed higher JSDs between adjacent time points than patients lacking macrochimerism (p = 0.0081) (Fig. 1f and g, right panels).

Although longer inter-sample intervals may theoretically diminish the similarity of TCR repertoires, our observation of increased stability as time post-transplant increased, when sampling was less frequent, argues against this being a significant confounder (Fig. 1d and e). To further illustrate the minimal influence of interval variation on TCR repertoire similarities compared with chimerism status and donor age, we assessed the proportions of recurring TCR sequences and JSD values in relation to time interval between two adjacent sampling points of each patient (Fig. S6) and did not observe increased JSD values or lower recurring TCR sequence proportions associated with longer time intervals (Fig. S6).

Taken together, these studies show that patients without macrochimerism experienced rapid recipient T cell infiltration and established a stable recipient T cell repertoire, at approximately one-year post-Tx. In contrast, high proportions of newly detected T cell clones were observed much longer in patients with macrochimerism and donor age <1Y, delaying stable recipient repertoire establishment to over 3 years post-Tx. The patients with macrochimerism and donor age ≥ 1Y, however, maintained a stable donor repertoire before donor T cells had been replaced by the recipient and eventually (up to 4–5 years post-Tx) established a more stable recipient repertoire than that in patients with donor age <1Y. These data suggest that establishment of a stable TRM repertoire in grafts from infant donors (<1Y) may still take many years in patients with macrochimerism, despite early entry of recipient T cells into the relatively empty allograft.

TCR sequences identifiable in pre-Tx recipient intestine are more likely to populate ileal allografts than those detected only in lymphoid tissues and those detected in both pre-Tx recipient intestine and lymphoid tissues have distinct transcriptional profiles

In order to investigate the origin of recipient T cells that contribute to the establishment of post-Tx TRM repertoires in ileal allografts, we categorized the TCR sequences by their detection in either pre-Tx recipient lymphoid tissues (LN or spleen) or NLT (gut) alone or both, whenever pre-Tx gut specimens were available. Pre-Tx recipient gut tissue was available for only 5 of the 16 patients (Pts16″, 18, 19, 20, and 21). Regardless of the status of macrochimerism, we found that the post-Tx allograft ileum contained a variable range of pre-Tx recipient-mappable TCRs (Fig. S7a, one patient without macrochimerism: 8.77% ± 3.07%, four patients with macrochimerism: 7.73% ± 6.37%). In all groups, recipient-mappable TCRs included those defined as “gut only”, “lymphoid only” and “gut-lymphoid shared” (Fig. 2a and Fig. S7a). Despite the fact that the number of TCR templates and unique sequences detected in the recipient pre-Tx lymphoid tissues was markedly greater than that in the gut in all cases (Fig. S7b), the sequences detected in the recipient pre-Tx gut tissues, especially the “gut-lymphoid shared” sequences, were highly represented in the post-Tx ileal graft tissue (Fig. 2a). This result suggests the existence of recirculating counterparts of clones detected in pre-Tx intestine that have a propensity to populate the intestinal graft.

Fig. 2.

Composition, relative detection rate and persistence of intestinal T cell subsets according to their origin. (a) Cumulative frequencies of recipient TCRs identifiable as pre-Tx “gut only”, “lymphoid only” or “gut-lymphoid shared” sequences in post-Tx ileal allografts from patients without (Pt20) and with (Pts16″, 18, 19, 21) macrochimerism. LN: lymph nodes. (b) Relative detection rates in post-Tx allografts of sequences identifiable as pre-Tx “gut only”, “lymphoid only”, and “gut-lymphoid shared”. Relative detection rate is the odds ratio of the detection rate of two designated sequence subsets. A value > 1 indicates greater rate of detection of the first sequence subset versus the second sequence subset. Unique sequence number of pre-Tx identified and post–Tx detected “gut only”, “lymphoid only”, and “gut-lymphoid shared” sequences are shown in Table S4. (c) Representative circle plots of three different recipient repertoires (pre-Tx “gut only”, “lymphoid only” or “gut-lymphoid shared”) in Pt19 post-Tx ileal samples collected on POD127, 356, and 729. The length of the colored lines indicates the % overlap by cumulative frequency, ranging from 0 to 100, between the sample represented by each colored line and the sample labeled on the outside circle of each pie, the latter of which serves as the denominator. (d) For each colored line as illustrated in panel c, % TCR overlap by cumulative frequency values was summarized for each patient and subgrouped by their origin in pre-Tx recipient tissues: “gut only”, “lymphoid only”, and “gut-lymphoid shared”. Solid horizontal bars indicate the mean of cumulative frequencies of overlapping TCR sequences, while the dotted horizontal bars indicate the median. For a to d, post-Tx ileal samples were included if recipient T cell chimerism in ileal IEL was >50% as assessed by FCM. For a to d, TCR sequences from both unstimulated and CFSE^low CD4/CD8 T cells after MLR were included to map comprehensive lymphoid repertoires to define “gut only”, “lymphoid only”, and “gut-lymphoid shared” sequences. However, only “lymphoid only” and “gut-lymphoid shared” sequences that appeared in pre-Tx unstimulated CD4/CD8 populations were used to compare the detection rates in b. (In panel d, Friedman test followed by Dunn's multiple comparisons test was performed for Pt20 and Pt21, RM one-way ANOVA followed by Tukey's multiple comparisons test was performed for Pts16″, 18, and 19. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001).

We compared post-transplant detection rates of pre-transplant-detected “gut only”, “lymphoid only”, and “gut-lymphoid shared” T cell clones in ileal allograft samples. All “gut only” and “gut-lymphoid shared” clonal detection rates post-Tx were substantially greater than those of “lymphoid only” clones (Fig. 2b, yellow and green curves, respectively), regardless of whether recipient pre-transplant lymph node or splenic T cells were the source of lymphoid tissue. Furthermore, “gut-lymphoid shared” clones had higher detection rates than “gut-only” clones (Fig. 2b). To correct for any differences in copy number distributions of “gut only”, “lymphoid only”, and “gut-lymphoid shared” sequences (Fig. S8a) that could bias our analysis, we also compared detection rates of sequences whose copy number was one in pre-Tx samples (Fig. S8b). Higher detection rates were still observed for “gut-lymphoid shared” sequences compared to “gut only” and “lymphoid only” sequences and higher detection rates of “gut only” sequences compared to “lymphoid only” sequences persisted in all the patients. Statistically significant differences between “gut-lymphoid shared” versus “lymphoid only” were consistent in four out of five patients (Fig. S8b).

We observed significantly higher degrees of overlap of sequences among post-Tx ileal samples that were identifiable as pre-Tx “gut-lymphoid shared” TCRs compared to “lymphoid only” TCRs, in four out of five patients (Pt20, p < 0.0001; Pt18, p = 0.0005; Pt19, p = 0.0046; Pt21, p = 0.0047) (Fig. 2c and d). Additionally, significantly higher degrees of sequence overlap among post-Tx ileal samples that were identifiable as pre-Tx “gut only” TCRs compared to “lymphoid only” TCRs were seen in three out of five patients (Pt20, p < 0.0001; Pt16″, p = 0.0001; Pt21, p = 0.0342) (Fig. 2c and d). Thus, “gut-lymphoid shared” T cells, “gut only” T cells and “lymphoid only” T cells, in descending order, can enter and persist in intestinal allografts. These data suggest that recipient intestinal TRM clones have circulating counterparts that home to, as well as distribute and persist within, the intestinal allograft following transplantation.

To further investigate biological differences among these T cell subsets, we performed 5′GEX scRNA sequencing paired with TCR sequencing for pre-Tx recipient spleen samples, IELs, and LPLs from Pt21. After integrating these three samples, these cells produced 16 clusters shown in the UMAP (Fig. S9a). TRM and non-TRM-defining genes expressed by cells in each cluster are shown in the heatmap (Fig. S9b).⁸ Cells detected in the pre-Tx recipient spleen sample whose TCRs were also detected in pre-Tx gut (including IEL and LPL) are defined as “pre-Tx spleen shared”, while those whose TCR appeared only in pre-Tx spleen are defined as “pre-Tx spleen only”. Cells detected in pre-Tx recipient gut samples whose TCR were also detected in pre-Tx spleen are defined as “pre-Tx gut shared” while those whose TCR appeared only in pre-Tx gut were defined as “pre-Tx gut only”. The clearest TRM (cluster 8) and non-TRM (clusters 1 and 2) cells, defined by gene expression in Fig S9b, are circled in the split UMAP plots of “pre-Tx gut shared” and “pre-Tx spleen shared” (Fig. S9c) and their cell numbers are summarized in Fig. S9d and presented in Table S6. The similar proportion of TRMs detected in “pre-Tx gut shared” (36.96%) and “pre-Tx spleen shared” (37.84%) in Fig. S9d and the similar gene expression of both cell sets shown in the left panel of Fig. S9e indicate strong similarity of these pre-Tx recipient T cells, regardless of whether they were detected in spleen or gut. In contrast, the transcriptional profiles of these “shared” cells differed markedly from the cells detected in only the gut or the spleen. The different proportion of cells in the “non-TRM” (clusters 1 and 2) (pre-Tx gut shared: 6.60%, pre-Tx spleen shared: 21.62%) (Fig. S9d) and higher CCR7 expression among those detected in spleen (Fig. S9e, left panel) (Table S7) demonstrates localized differences among these cells belonging to the same clones. Similarly, when compared with the “pre-Tx gut only” cells, the “pre-Tx gut shared” cells demonstrated a distinct gene expression profile, with increased expression of effector genes, such as GZMA, GZMB, KLRB1, NKG7, GNLY, IFNG, and PRF1 (Fig. S9e, middle panel) (Table S8). The right panel of Fig. S9e (Table S9) shows that the “pre-Tx spleen shared” cells expressed higher levels of the TRM-associated gene CXCR6 and had lower expression of the non-TRM genes KLF2, SELL and CCR7 compared to the “pre-Tx spleen only” cells, illustrating the TRM-like differentiation tendency of “pre-Tx spleen shared” cells.

High level of crosstalk between graft-infiltrating and circulating recipient T cells post-ITx

The data presented above suggests that recipient T cells enter intestinal allografts from the circulation. We therefore examined the crosstalk between TCRs in the intestinal graft and the circulation at matched time points (collected on the same day or adjacent 2 days). We analyzed the percentages of sequence overlap (Fig. 3a and b) and compared cosine indices (Fig. 3c) of paired ileum-PBMC samples (n = 16) with those of paired rectum-PBMC samples from published ulcerative colitis (UC) patients (n = 7) and their healthy controls (HCs, n = 4).⁴¹

Fig. 3.

Similarity analysis of graft-associated and circulating TCR repertoires post-Tx. (a) Representative correlation plots of clonal frequencies between paired gut (ileal biopsy) and blood samples from ITx patients, compared to paired rectal biopsies and PBMCs for ulcerative colitis (UC) patients and their healthy controls (HC) retrieved from recent published work.⁴¹ (b) Percentages of shared sequences in gut by cumulative frequency (unique sequences weighted by copy numbers) and clone fraction (unique sequences unweighted by copy numbers) are summarized for each gut-PBMC pair from ITx patients, UC patients, and HCs. Solid horizontal bars indicate the mean of gut-PBMC shared sequences' proportions in gut by cumulative frequency, while the dotted horizontal bars indicate the median of gut-PBMC shared sequences' proportions in gut by clone fraction. (c) Cosine indices among paired samples from ITx patients are presented up to POD1500. Cosine index values range between 0 and 1, where 1 indicates identical repertoires and 0 indicates complete divergence. Horizontal dotted line: cosine index = 0.1. Vertical dotted line: POD600. For a to c, samples were included if recipient T cell chimerism in ileal IEL biopsies were >50% during that period as measured by FCM. Paired ileal samples and PBMC were collected on the same day or within two days. (In panel b, ordinary one-way ANOVA followed by Tukey's multiple comparisons test was performed for the left plot and Kruskal–Wallis test followed by Dunn's multiple comparisons test was performed for the right plot. ∗p < 0.05).

Fig. 3a shows representative correlation plots of clonal frequencies between paired ileal biopsy and blood samples for Pt21 POD262 (approximately 98–99% of T cells in ileal IEL and LPL were recipient HLA⁺, as detected by FCM), along with paired rectum-PBMC samples from representative UC patient #03 (UC03) and HC02. The percentages of overlap between paired gut and blood repertoires from ITx patients were significantly higher than those in the UC and HC groups. The UC group showed an intermediate level of overlap between the ITx patients and the HC group by cumulative frequency (ITx compared to UC patients, p = 0.0343; ITx compared to HC, p = 0.0111) (Fig. 3b, left panel) and clone fraction (ITx compared to UC patients, p = 0.0325; ITx compared to HC, p = 0.0506) (Fig. 3b, right panel). These data indicate increased gut-blood crosstalk after ITx, even without clinically apparent rejection or infection. Although a clear temporal pattern was not apparent, five out of seven gut-PBMC pairs obtained later than POD600 had relatively low cosine indices (0.1 or less), suggesting that steady state levels of gut-blood crosstalk may be reached late post–Tx (Fig. 3c).

Recipient T cells tend to form similar repertoires in the intestinal allograft and native colon

We further assessed the spatial distribution of recipient T cells throughout the gut. Late post-Tx (POD357–1847) pan-scope samples, including ileal and colon graft biopsies and native colon tissues, allowed examination of the recipient T cell distribution among different regions of the gut. First, TCR sequences from pan-scope samples were divided into “non-shared”, “double-shared”, and “triple-shared” among these sites according to their appearance (Fig. 4a), when T cell replacement rates by the recipient were >50% in ileal IEL. Representative contributions of “non-shared”, “double-shared”, and “triple-shared” sequences are shown in Pt14 POD717 and Pt15 POD1847 by cumulative frequency (Fig. 4b). A significant proportion of double-shared (3.37–29.10%) and triple-shared (1.82–52.58%) sequences was detected among ileum, colon, and native colon samples (Fig. 4b, c and Fig. S10), suggesting the establishment of similar repertoires in multiple regions of the gut, including both the allograft and native intestine. Cosine indices (0.4–0.6) from representative pan-scopes (Pt15 POD1847) showed broad similarities of TCR repertoires in pan-scope specimens (Fig. 4d). The levels of similarity reflected by cosine index values in ileum vs colon, ileum vs native colon and colon vs native colon were significantly higher than those in pre-Tx gut vs lymphoid tissue from the same set of patients (“gut vs lymphoid tissue” compared to “ileum vs colon”, p = 0.0002; “gut vs lymphoid tissue” compared to “ileum vs native colon”, p = 0.1440; “gut vs lymphoid tissue” compared to “colon vs native colon”, p = 0.0002) (Fig. 4e, left panel). Similar results were observed when we used Jaccard Index to compare overlap of unique sequences, avoiding any possible influence from varying copy numbers of clones (“gut vs lymphoid tissue” compared to “ileum vs colon”, p = 0.0021; “gut vs lymphoid tissue” compared to “ileum vs native colon”, p = 0.0090; “gut vs lymphoid tissue” compared to “colon vs native colon”, p = 0.0008) (Fig. 4e, right panel). The considerable proportion of double-shared (without macrochimerism: 11.75%–22.05%, with macrochimerism: 3.37%–29.10%) and triple-shared sequences (without macrochimerism: 1.83%–26.15%, with macrochimerism: 1.82%–52.58%) (Fig. S11a) and higher similarity in different parts of the gut compared to gut vs lymphoid tissue (left panel “gut vs lymphoid tissue” compared to “ileum vs colon”, p = 0.0219; left panel “gut vs lymphoid tissue” compared to “ileum vs native colon”, p = 0.4946; left panel “gut vs lymphoid tissue” compared to “colon vs native colon”, p = 0.1077; right panel “gut vs lymphoid tissue” compared to “ileum vs colon”, p = 0.0028; right panel “gut vs lymphoid tissue” compared to “ileum vs native colon”, p = 0.2312; right panel “gut vs lymphoid tissue” compared to “colon vs native colon”, p = 0.0004) (Fig. S11b) persist after subgrouping the ITx patients by macrochimerism status.

Fig. 4.

Clonal distribution and similarity across different regions of the gut late post-Tx. (a) Venn diagrams illustrating subsets of sequences appearing in different regions of the intestine in a patient at a time point when transplanted ileum, transplanted colon and native colon were simultaneously biopsied through a pan-endoscopic procedure. Color codes in panels b and c match each subset defined in panel a. (b) Representative pan-scope samples collected from Pt14 POD717 and Pt15 POD1847 illustrate clonal distribution of “non-shared”, “double-shared” and “triple-shared” sequences among recipient T cell repertoires in gut. Total template counts within samples are shown in the center of each pie. (c) Percentage of sequence subsets in all pan-scope samples from patients without (n = 3) and with (n = 7) macrochimerism. Solid horizontal bars indicate the mean of cumulative frequency of each TCR sequence subsets. (d) Representative circle plot from pan-scope samples collected from Pt15 POD1847. The length of the colored lines indicates the cosine index between the sample represented by each colored line and the sample labeled on the outside circle of each pie. (e) Comparison of cosine indices and Jaccard indices of unique sequence number from all pan-scope samples. Cosine indices and Jaccard indices of 10 pairs of pre-Tx gut and lymphoid tissues shown in Fig. 2 are presented as gut vs lymphoid tissue negative controls (different tissue from the same person). Dotted horizontal bars indicate the median cosine index or Jaccard index among paired samples. Jaccard index values range between 0 and 1, where 1 indicates identical repertoires and 0 indicates complete divergence. For b to e, samples were included if recipient T cell chimerism in ileal IEL was >50% measured by FCM. (In panel c, RM one-way ANOVA followed by Tukey's multiple comparisons test was performed. In panel e, Kruskal–Wallis test followed by Dunn's multiple comparisons test was performed. ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001).

Alloreactivity and microbe reactivity of recipient TCR sequences in ileal allografts and circulation post-Tx

Ileal-infiltrating recipient T cells are constantly exposed to alloantigens and microbes, even in the absence of clinically apparent rejection and infection. In a previous study, by combining pre-Tx CFSE-MLR and high throughput TCR-β sequencing, we were able to define and compare HvG and non-HvG sequences in rejecting intestinal biopsy specimens.²⁶ In the current study, we focused on the detection of HvG and non-HvG sequences within CD4 and CD8 populations among ileal and PBMC samples from patients with and without macrochimerism during quiescent periods (Fig. 5a and b). The detection rates of CD8, but not CD4, HvG sequences tended to be higher than those of non-HvG sequences in ileal samples, achieving statistical significance in patients without macrochimerism (p = 0.0155) (Fig. 5b). In contrast, CD8 HvG clone detection rates were significantly lower than those of non-HvG CD8 clones in the PBMCs, regardless of macrochimerism status (without macrochimerism, p = 0.0005; with macrochimerism, p < 0.0001) (Fig. 5b).

Fig. 5.

HvG, non-HvG and microbe-reactive potential of T cells in post-Tx ileal and PBMC samples. (a) Representative calculation of detection rates of CD4 HvG, CD4 non-HvG, CD8 HvG, and CD8 non-HvG TCRs in Pt14 POD527 ileum and POD456 PBMC samples. Detection rate is the number of unique HvG or non-HvG sequences detected post-Tx normalized by the number of unique sequences identified pre-Tx among CD4 or CD8 HvG or non-HvG subsets. (b) Comparison of detection rate between HvG and non-HvG sequences in post-Tx ileal (without macrochimerism: Pt4 POD524, 674, Pt9 POD254, Pt10 POD204, 338, Pt14 POD156, 226, 527, 717, Pt20 POD68, 104, 250, 306, 523, 796, Pt24 POD39, with macrochimerism: Pt13 POD1013, Pt15 POD1018, 1336, 1847, Pt16″ POD494, 662, 1004, 1329, Pt17 POD472, 584, 662, Pt18 POD105, 307, 357, Pt19 POD127, 356, 729, Pt21 POD109, 262, 626, 915, 1145, Pt23 POD37, 49, 149, 346, 892) and PBMC (without macrochimerism: Pt4 POD324, 541, Pt9 POD156, Pt14 POD456, Pt20 POD68, 110, 242, 306, 524, 635, 796, Pt24 POD20, 36, with macrochimerism: Pt13 POD1032, Pt15 POD1070, 1336, 1849, Pt16″ POD494, 565, 663, 1004, Pt17 POD662, Pt18 POD98, 314, 357, Pt19 POD355, 734, Pt21 POD109, 262, 626, 916, Pt23 POD46, 119, 346) samples. Unique sequence number of pre-Tx identified and post–Tx detected CD4 HvG, CD4 non-HvG, CD8 HvG, and CD8 non-HvG sequences of ileal and PBMC samples are shown in Table S10. (c) The proportion of microbe-reactive sequences, defined by sequence overlap with published microbe-reactive TCR databases42, 43, 44, 45, 46 at the amino acid level of CDR3+V+J of TCR beta chain, in HvG, non-HvG, and unmappable subsets from post-Tx ileum and PBMC samples. Dotted horizontal bars indicate the median proportion of microbe-reactive sequences within each subset. For a to c, samples were included if recipient T cell chimerism in ileal IEL was >50% during that period as measured by FCM. For c, subsets were included if their template counts (read counts for Pts 22 to 24) >100. (In panel b, Wilcoxon test was performed to determine statistical significance. In panel c, Kruskal–Wallis test followed by Dunn's multiple comparisons test was performed to determine statistical significance. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001).

In addition to detection rates, we also compared the overall contribution of HvG and non-HvG sequences to ileal TCR repertoires post-Tx. After excluding pre-Tx donor TCRs and pre-Tx recipient “gut only” TCRs, which could not be classified for alloreactivity (these populations were not tested in MLR assays), TCR sequences detected in post-Tx ileal and PBMC samples were divided into three categories: HvG-reactive, non-HvG-reactive, and pre-Tx unmappable (Fig. S12a). Smaller proportions of HvG sequences (ileum: 0–6.48%, PBMC: 0–1.40%) than non-HvG sequences (ileum: 0.37–31.07%, PBMC: 0–61.61%) were detected in post–Tx ileum (Fig. S12a, upper panel) and PBMC (Fig. S12a, lower panel) samples. The majority of post–Tx intestinal and PBMC sequences (ileum: 63.93–97.27%, PBMC: 37.81–99.88%) were unmappable (Fig. S12b).

To assess potential microbe reactivity of the sequences detected in post-Tx samples, we interrogated published TCR databases to look for clonal overlap at the amino acid level of CDR3+V+J of TCR beta chain.42, 43, 44, 45, 46 Non-HvG sequences in both ileum and PBMC (Fig. 5c) included greater percentages associated with recognition of microbial antigens compared to HvG and unmappable sequences (ileal samples of patients without macrochimerism, non-HvG sequences compared to HvG sequences, p = 0.0005, non-HvG sequences compared to unmappable sequences, p = 0.0024; ileal samples of patients with macrochimerism, non-HvG sequences compared to HvG sequences, p < 0.0001, non-HvG sequences compared to unmappable sequences, p = 0.0087; PBMC samples of patients without macrochimerism, non-HvG sequences compared to HvG sequences, p = 0.0993, non-HvG sequences compared to unmappable sequences, p = 0.0062; PBMC samples of patients with macrochimerism, non-HvG sequences compared to HvG sequences, p = 0.0002, non-HvG sequences compared to unmappable sequences, p < 0.0001). These data suggest the preferential migration of non-alloreactive TCRs with anti-microbial reactivity from circulation to the ileal allograft and persistence as TRMs in the gut to participate in host immune defense.

Discussion

Our study provides an unprecedented longitudinal analysis of the dynamic landscape of the TCR repertoire in human intestines. This was achieved by analysis of pre-transplant specimens, quiescent ileal allograft biopsies and PBMCs from early to late periods post-Tx and transplanted colon and native colon at certain late time points post-Tx. These studies provide new insights into the biology underlying human TRM repertoire establishment and its relationship to T cell alloresponses and microbial recognition. Our study not only assesses TRM repertoire stability and persistence in ileal allografts post-Tx, but also addresses the tissue origin, distribution, and antigen recognition patterns of graft-infiltrating recipient T cells.

By performing longitudinal assessments of bulk TCR repertoire in allograft mucosae, we were able to document the establishment of a stable intestinal mucosal TRM repertoire post-Tx, as reflected by increased cumulative frequencies of recurring TCRs over time and decreased JSD values between adjacent timepoints. We confirmed that ITx patients with blood T cell macrochimerism and donor age ≥ 1Y demonstrate very slow replacement of donor T cells by those of the recipient, as previously reported.²⁵ By adding bulk TCR sequencing to this analysis, we now demonstrate the very slow (over a period of years) stabilization of the newly-established recipient TRM repertoire in the allograft. The recipient T cells entering the graft quickly acquire the TRM phenotype, but recruitment of new recipient clones may continue for several years following the transplant, as reflected in the continued detection of new TCRs, the persistently high JSDs and high level of TCR cross-talk between gut and PBMC. However, patients with macrochimerism and donor age <1Y, who experience much more rapid reconstitution of the mucosal T cell compartment by the recipient, show an even less stable recipient TCR repertoire at late timepoints. Like the group with macrochimerism and older donor ages, this group shows low rates of allograft rejection or DSA development.²⁵ The rapid recipient reconstitution in the grafts from infant donors most likely reflects the filling by the recipient of a relatively “empty” TRM compartment in these infant-derived allografts, as studies of deceased donors of different ages have shown that the intestinal TRM compartment is not fully established until later in childhood.⁴⁷ The long-term instability of the recipient repertoires in grafts from infant donors with quiescent allografts and macrochimersim suggests that de novo establishment of an intestinal TRM repertoire from circulating T cells can take several years.

Patients without macrochimerism have increased rejection rates and de novo DSA development compared to those with macrochimerism.25, 26, 27 Our current study demonstrates relatively greater and earlier stability of the recipient TRM repertoire in patients without macrochimerism, regardless of donor age, compared to that in patients with macrochimerism and early recipient T cell infiltration due to young donor age (<1Y). Increased detection rates of CD8 HvG clones in graft mucosa compared to those in PBMCs during quiescence only in the group lacking macrochimerism suggests that these HvG clones take up residence as mucosal TRMs that pose a constant risk of rejection. A more profound predominance of HvG clones is observed during active rejection episodes early post–Tx, as we reported previously.²⁶ In studies that integrate TCR clonal analysis with transcriptomic profiles at the single cell level during quiescence and rejection, we have demonstrated a predominant TRM transcriptional profile of pre-transplant-identified HvG clones and obtained evidence that these cells may be tolerized. In contrast, a more effector-like profile is seen for HvG T cells in rejecting allografts.⁴⁸

An emerging concept in TRM biology is that TRMs can exit the tissues and become “ex-TRMs” and even translocate to distant NLTs.13, 14, 15 The existence of circulating counterparts to TRMs has been demonstrated through animal models and studies of human skin.11, 12, 13^,49, 50, 51, 52 Our analysis of the origin of recipient TRMs in ileal allografts reveals several dynamic processes, including repertoire establishment from the circulation, preferential origin from the pool of recipient intestinal TRM with circulating counterparts, high levels of TCR sharing between the post-Tx graft and recipient intestinal mucosae, and active crosstalk between the circulating pool and the intestinal TRM pool.

T cell clones that we detected in both gut and lymphoid tissues or in pre-Tx recipient intestine only were found to enter and persist in donor intestinal allografts for up to 4 years post-Tx. Clones detected only in pre-Tx recipient lymphoid tissues were least likely to be detected in intestinal allografts. The considerable proportion of gut-lymphoid shared TCRs in the allograft, which were detected at higher rates compared to gut-only TCRs, together with the similar TRM-like gene expression profiles of gut-lymphoid shared cells regardless of whether they were detected in gut or spleen, strongly supports the existence of a lymphoid or circulating counterpart to human intestinal T cells and suggest that it is the circulating counterparts that are most likely to enter an allograft. The enrichment of effector genes in gut-lymphoid shared T cells compared to gut-only T cells among those detected in pre-transplant recipient gut suggests that this subset of T cells with a counterpart in lymphoid tissues and circulation has specialized function in immune protection. To our knowledge, these data are the first to demonstrate transcriptional differences that define a migratory T cell population in a non-lymphoid tissue. In another study, we used single cell RNA-seq to compare the contribution of HvG T cells to TRM and effector T cell populations during rejection and quiescence and directly demonstrated a TRM profile in quiescent allografts and an effector phenotype for HvG T cells in rejecting allografts. We identified a mixed transcriptional profile between these two states and demonstrated trajectories between them in association with rejection vs quiescence. We also showed directly that individual HvG clones could take on TRM-like or effector-like transcriptional phenotypes.⁴⁸ We believe that knowing the origin of TRMs that populate an intestinal allograft may lead to new approaches to preventing rejection. For example, blockade of T cell trafficking to the gut with anti-α4β7 integrin mAb might be predicted to prevent the migration of circulating recipient T cells, which we show here are likely to be an ongoing source of graft HvG T cells, into the graft mucosa and thereby prevent rejection.

Similar to the increased clonal overlap between rectum and blood reported in UC patients compared to HCs,⁴¹ there was increased clonal overlap between paired ileal graft biopsy and PBMC specimens collected at similar times late post-Tx. This result may suggest that transplantation creates a highly inflamed environment even during rejection- and infection-free periods, when all of the samples in the present analysis were taken, triggering active translocation between graft-infiltrating recipient T cells and their circulating counterparts for years. However, we favor the alternative explanation that this crosstalk reflects the ongoing population of the intestinal TRM compartment by relatively rare circulating counterparts of recipient intestinal TRM cells, resulting in an inherently slow process. Consistent with this interpretation, the increased crosstalk between circulating and intestinal TCRs at late time points in recipients of grafts from donors <1 year old may reflect a long period required to fill an “empty” TRM compartment.

Since longitudinal biopsy specimens in our study were collected not only at different times but also from different locations, our repertoire overlap studies demonstrate considerable spatial in addition to temporal overlap of TCR clones. When we were able to collect pan-scope specimens from late post-Tx time points, a wide distribution of TCRs among transplanted ileum, transplanted colon and native colon was detected. Significantly higher cosine indices were detected between simultaneous pan-scope specimens—between both allografts and native tissues—compared to pre-Tx gut vs lymphoid tissues collected from the same set of patients. These data suggest that potential recognition of microbial antigens may be a major driver of T cell clones detected in these multiple locations. Anti-microbial activity can result in a shared memory T cell repertoire in liver and gut in patients with primary biliary cirrhosis and inflammatory bowel disease.⁵³ We were also able to demonstrate the presence of TCR sequences associated with microbial antigen recognition among gut-infiltrating non-HvG sequences, consistent with a role for such reactivity in driving the intestinal TRM repertoire. Rodent studies demonstrated that responses to commensal organisms can participate in and promote graft rejection.^54,55 Further investigation at the TCR repertoire and transcriptomic levels will promote a deeper understanding of the effect of anti-microbial T cells on human allograft outcomes.

Although TCR sequences retrieved from small endoscopic biopsy specimens cannot fully represent the entire gut, the application of a requirement for at least 800 TCR template counts (read counts for Pts 22 to 24) in each sample for inclusion in our analysis helps reduce the effect of sample size limitations and captures TCR overlap sufficiently to allow assessment of repertoire diversity, stability, and persistence. In fact, the significant level of clonal overlap demonstrated among small endoscopic biopsies taken at different sites over periods of years post-Tx strongly argues for the establishment of a stable recipient TRM repertoire throughout the entire gut. Our analysis of the dynamic establishment of a recipient TRM repertoire post-ITx during quiescence is likely to be analogous to and informative regarding TRM repertoire establishment during normal ontogeny. Our study mandates future investigations to add transcriptional and epigenomic profiling of intestinal T cells with TRM features to TCR repertoire analysis,^51,56 which will allow an even broader and deeper understanding of human TRM biology.

Limitations of our study include the small ITx patient cohort, which makes it unreasonable to conduct multivariate analyses. Instead, we elected to focus on the impact of several specific factors that we expected to have biological significance, such as chimerism and donor age.^25,26 The strong association between these two factors and rejection occurrence, which is associated with accelerated replacement of donor TRMs in allograft mucosa,²⁶ may obscure the influence of other factors on TRM repertoire establishment, including recipient age, the presence or absence of the native spleen, and clinical events such as rejection and infection. Due to the lack of access to other NLTs such as skin and lung, we were not able to compare clonal overlap between gut TRMs and such tissues. Although it was not possible to perform phenotypic or transcriptional analysis on each sequenced specimen to confirm the TRM phenotype of recipient T cells, the transition of recipient T cells to the CD69+/CD103+ phenotype described here and single cell transcriptional analyses we have performed on quiescent allograft T cells in our new published paper have clearly demonstrated the prominent TRM profile of these recipient T cells.⁴⁸ The new understanding of TRM dynamics obtained from this study and future studies should help to refine the modalities used to prevent rejection and optimize outcomes following intestinal transplantation.

Contributors

W.J., J.F., and M.S. designed the study. W.J., J.F., J.Z., K.L., E.W., K.F., R.J., A.G., and B.S. performed the experiments. J.F., M.M., K.F., R.J., C.B.M., W.J., K.R., S.R., M.V., T.K., and J.W. coordinated the clinical sample collection. C.B.M., W.J., J.F., M.M., and K.R. coordinated the clinical data collection. T.K. and J.W. performed the intestinal transplantation, stoma revision/closure and graft removal surgeries. M.M. and S.R. performed the routine endoscopy and patient care. W.J., A.O., G.T., Z.W., J.F., W.M., A.M.R., J.Z., G.L., Y.S. and M.S. participated in data analysis. W.J., C.P., A.O. J.F. and Y.S. wrote the codes to identify and track alloreactive clones. W.J., J.F., and M.S. wrote the final report. W.J. and J.F. have verified the underlying data. All authors contributed to the editing of the final report. All authors have read and approved the final version of the manuscript.

Data sharing statement

Pts 4–21's raw TCR -seq data are available at https://clients.adaptivebiotech.com/pub/jiao-2024-ebm. For Pts 22–24, raw TCR-seq data in FASTA format is accessible at Sequence Read Archive (SRA: https://www.ncbi.nlm.nih.gov/sra) with the BioProject accession number PRJNA578087 and PRJNA926497. The single cell sequencing data from Pt 21 is accessible at PRJNA1012302. The R code used to analyze TCR-seq data and single cell sequencing data is available in the GitHub repository at https://github.com/drjiaowy/TRM_manuscript. TCR sequences reactive to certain microbials in post–Tx gut specimens of ITx patients were identified by sequence overlap (V gene + J gene + CDR3 at amino acid level) with previously published papers and publicly available databases: TCRdb, VDJdb, and PIRD.42, 43, 44, 45, 46 Commensal bacteria, opportunistic bacteria, and viruses included: Escherichia coli, staphylococcus aureus, mycobacterium tuberculosis, shigella, klebsiella pneumoniae, salmonella typhimurium, Enterobacter aerogenes, yellow fever virus, human immunodeficiency virus, hepatitis C virus, Epstein–Barr virus, and cytomegalovirus.

Declaration of interests

J.F. serves as a Scientific Consultant for Adaptive Biotechnologies Corp. since June 2022. A.O. serves as a Scientific Consultant for Janssen Pharmaceuticals and served as a Scientific Consultant for Enable Medicine.

Appendix A. Supplementary data

Supplementary Tables

Table S1. Epidemiological and clinical characteristics of patients.

Table S2. Flow cytometry antibodies and other reagents used in this study.

Table S3. (related to Fig. S7a and Fig. S12) Number of total template counts, unique sequence number in post-Tx ileal and PBMC samples and pathological diagnosis of ileal samples are shown with immunosuppressive treatments.

Table S4. (related to Fig. 2b) Unique sequence number of pre-Tx identified and post-Tx detected “gut only”, “lymphoid only”, and “gut-lymphoid shared” sequences. “Lymphoid only” and “gut-lymphoid shared” sequences that appeared in pre-Tx unstimultaed CD4/CD8 cells were included at here.

Table S5. (related to Fig. S8b) Unique sequence number of pre-Tx identified “gut only”, “lymphoid only”, and “gut-lymphoid shared” sequences with copy number = 1. “Lymphoid only” and “gut-lymphoid shared” sequences that appeared in pre-Tx unstimultaed CD4/CD8 cells were included at here.

Table S6. (related to Fig. S9c and d) Cell numbers of different clusters in “pre-Tx gut shared” subset and “pre-Tx spleen shared” subset.

Table S7. (related to the left panel of Fig. S9e) All the different express genes between “pre-Tx gut shared” subset and “pre-Tx spleen shared” subset.

Table S8. (related to the middle panel of Fig. S9e) All the different express genes between “pre-Tx gut shared” subset and “pre-Tx gut only” subset.

Table S9. (related to the right panel of Fig. S9e) All the different express genes between “pre-Tx spleen shared” subset and “pre-Tx spleen only” subset.

Table S10. (related to Fig. 5b) Unique sequence number of pre-Tx identified and post-Tx detected “CD4 HvG”, “CD4 non-HvG”, “CD8 HvG”, and “CD8 non-HvG” sequences.

Fig. S1.

Collection of pre-Tx and post-Tx samples with TCR-β sequencing data. D: donor, R: recipient.

Fig. S2.

Representative contour plots demonstrating gating strategy for recipient chimerism and tissue-resident memory (TRM) marker expression in Pt20 POD208 peripheral blood mononuclear cell (PBMC), ileal intraepithelial lymphocyte (IEL), and ileal lamina propria (LPL). (Pt20 donor: HLA-A3^- HLA-A2⁺, Pt20 recipient: HLA-A3⁺ HLA-A2⁺) (Pan-HLA marker, HLA-ABC, is replaced by HLA-A2 in this case.)

Fig. S3.

(a) Percentages of patients with >50% of recipient ileal CD4 IEL expressing CD69 (left panel) and CD103 (right panel) over time post-Tx in patients with (+) (Pts13, 18, 23) and without (-) (Pts9, 10, 14, 20) macrochimerism. (b) Percentages of patients with >50% of recipient ileal CD4 LPL expressing CD69 (left panel) and CD103 (right panel) over time post-Tx in patients with (+) and without (-) macrochimerism. (c) Percentages of patients with >50% of recipient ileal CD8 LPL expressing CD69 (left panel) and CD103 (right panel) over time post-Tx in patients with (+) and without (-) macrochimerism. For a to c, patients with ≥ six post-Tx time points by POD250 are included. (For a to c, log-rank test was performed to determine statistical significance.)

Fig. S4.

Dynamic establishment of TCR repertoires in ileal allografts post-Tx in patients without macrochimerism (a), patients with macrochimerism and donor age ≥ 1Y (b), and patients with macrochimerism and donor age <1Y (c).

Fig. S5.

Percentage of recipient CD3+ T cell chimerism in PBMCs, ileal IEL, and LPL of patients without macrochimerism (a), patients with macrochimersm and donor age ≥ 1Y (b), and patients with macrochimerism and donor age <1Y (c).

Fig. S6.

(a) Cumulative frequency of recurring TCR sequences in relation to time intervals between two adjacent sampling points. (b) JSD values in relation to time intervals between two adjacent sampling points. (△POD: time interval between two adjacent sampling points).

Fig. S7.

(a) Percentages of cumulative frequency (unique sequences weighted by copy numbers) and clone fraction (unique sequences unweighted by copy numbers) of “unmappable”, “donor”, “gut only”, “lymphoid only”, and “gut-lymphoid shared” sequences in post-Tx ileal samples from patients without (Pt20) and with (Pts16’’, 18, 19, 21) macrochimerism. Total template counts and unique sequence number of these post-Tx ileal samples from Pts16’’, 18, 19, 20, 21 are shown in Table S3. (b) Total template counts and unique sequence number of pre-Tx recipient lymphoid tissues (open symbols, including both unstimulated and CFSE^low CD4/CD8 T cells after MLR) and gut tissues (solid symbols) from patients without (Pt20) and with (Pts16’’, 18, 19, 21) macrochimerism.

Fig. S8.

(a) Copy number distribution of pre-Tx defined “gut only”, “lymphoid only”, and “gut-lymphoid shared” sequences in post-Tx ileal samples. (b) The detection rate of pre-Tx identified “gut only”, “lymphoid only”, and “gut-lymphoid shared” sequences in post-Tx ileal samples. Sequences were included if their copy number in pre-Tx samples was equal to one. Post-Tx samples were included if recipient T cell chimerism in ileal IEL was >50% during that period as measured by FCM and all three types of sequences were detected. Dotted horizontal bars indicate the median detection rate among samples. Unique sequence numbers are shown in Table S5. For a and b, TCR sequences from both unstimulated and CFSE^low CD4/CD8 T cells after MLR were included to map comprehensive lymphoid repertoires during the defination of “gut only”, “lymphoid only”, and “gut-lymphoid shared” sequences, while “lymphoid only” and “gut-lymphoid shared” sequences which appeared in pre-Tx unstimulated CD4/CD8 populations were used in the comparison. (In panel b, Friedman test followed by Dunn’s multiple comparisons test was performed to determine statistical significance. ∗p<0.05, ∗∗p<0.01).

Fig. S9.

(a) Combined UMAP plots show cell clusters in all pre-Tx samples from Pt21 (pre-Tx recipient spleen, colon IEL and colon LPL). (b) Heatmap of TRM and non-TRM-defining genes in each cell cluster.⁸ (c) Split UMAP plots of “gut-spleen shared” cells in pre-Tx recipient gut (either IEL or LPL) and pre-Tx recipient spleen respectively, with major clusters expressing canonical TRM (cluster 8) and non-TRM (clusters 1 and 2) genes indicated. (d) Proportional constitution of cells in TRM, non-TRM and all other clusters in “gut-spleen shared” cells in pre-Tx recipient gut and pre-Tx recipient spleen respectively. The number in the center of each pie plot represents the total number of single cells in that subset. (e) Left panel: volcano plot of all differentially expressed (DE) genes between “gut-spleen shared” cells in pre-Tx recipient gut and pre-Tx recipient spleen. Middle panel: volcano plot of DE effector genes between “gut-spleen shared” cells and “gut-only” cells in the pre-Tx recipient gut. Right panel: volcano plot of DE TRM/non-TRM genes between “gut-spleen shared” cells and “spleen-only” cells detected in the pre-Tx recipient spleen. Log₂ fold change >0.5 and -Log₁₀P >1.301 (P <0.05) were applied at each direction to identify DE genes.

Fig. S10.

Clonal distribution of “non-shared”, “double-shared” and “triple-shared” sequences among recipient T cells repertoires from pan-scope samples (transplanted ileum, colon and native colon) collected from patients without (-) and with (+) macrochimerism. Total template counts within samples are shown in the center of each pie. Samples were included if recipient T cell chimerism in ileal IEL was >50% during that period as measured by FCM.

Fig. S11.

(a) Percentage of each class of sequences in pan-scope samples of patients without (Pt4 POD674, Pt14 POD527, 717) and with (Pt13 POD1013, Pt15 POD1018, 1336, 1847, Pt17 POD472, 584, Pt18 POD357) macrochimerism. Solid horizontal bars indicate the mean proportion of non-shared, double-shared, or triple-shared sequences in ileum, colon, or native-colon sample, while the dotted horizontal bars indicate the median proportions. (b) Comparison of cosine indices from patients without and with macrochimerism. Cosine indices of 10 pairs of pre-Tx intestinal and lymphoid tissues shown in Fig. 2 are persented as gut vs lymphoid tissue negative controls. Dotted horizontal bars indicate the median cosine indices among paired samples. For a and b, samples were included if recipient T cell chimerism in ileal IEL was >50% as measured by FCM. (In panel a, Freidman test followed by Dunn’s multiple comparisons test was performed for the plot of colon with macrochimerism and RM one-way ANOVA test followed by Tukey’s multiple comparisons test was performed for the other plots. In panel b, Kruskal-Wallis test followed by Dunn’s multiple comparisons test was performed. ∗p<0.05, ∗∗p<0.01, ∗∗∗p<0.001).

Fig. S12.

(a) Percentages of cumulative frequency of HvG, non-HvG, and pre-Tx unmappable TCR sequences defined by pre-Tx MLR in post-Tx ileal and PBMC samples. (b) Comparison of the percentages of cumulative frequency of HvG, non-HvG, and pre-Tx unmappable TCRs from panel a. Dotted horizontal bars indicate the median of cumulative frequency percentages of these categories. Total template counts and unique sequence number of post-Tx ileal and PBMC samples are shown in Table S3. (In panel b, Friedman test followed by Dunn’s multiple comparisons test was performed. ∗p<0.05, ∗∗p<0.01, ∗∗∗p<0.001, ∗∗∗∗p<0.0001).