Article Text

ScRNA-seq defines dynamic T-cell subsets in longitudinal colon and peripheral blood samples in immune checkpoint inhibitor-induced colitis
  1. Jacqueline E Mann1,
  2. Liliana Lucca2,
  3. Matthew R Austin1,
  4. Ross D Merkin1,
  5. Marie E Robert3,
  6. Badr Al Bawardy4,
  7. Khadir Raddassi2,
  8. Lilach Aizenbud1,
  9. Nikhil S Joshi5,
  10. David A Hafler2,5,
  11. Clara Abraham4,
  12. Kevan C Herold5,6 and
  13. Harriet M Kluger1
  1. 1Department of Internal Medicine (Medical Oncology), Yale School of Medicine, New Haven, Connecticut, USA
  2. 2Department of Neurology, Yale School of Medicine, New Haven, Connecticut, USA
  3. 3Department of Pathology, Yale School of Medicine, New Haven, Connecticut, USA
  4. 4Department of Internal Medicine (Digestive Diseases), Yale University, New Haven, Connecticut, USA
  5. 5Department of Immunobiology, Yale School of Medicine, New Haven, Connecticut, USA
  6. 6Department of Internal Medicine (Endocrinology), Yale School of Medicine, New Haven, Connecticut, USA
  1. Correspondence to Dr Harriet M Kluger; Harriet.Kluger{at}


Immune checkpoint inhibitors (ICIs) are increasingly being used to manage multiple tumor types. Unfortunately, immune-related adverse events affect up to 60% of recipients, often leading to treatment discontinuation in settings where few alternative cancer therapies may be available. Checkpoint inhibitor induced colitis (ICI-colitis) is a common toxicity for which the underlying mechanisms are poorly defined. To better understand the changing colon-specific and peripheral immune environments over the course of progression and treatment of colitis, we collected blood and colon tissue from a patient with Merkel cell carcinoma who developed colitis on treatment with pembrolizumab. We performed single-cell RNA sequencing and T-cell receptor sequencing on samples collected before, during and after pembrolizumab and after various interventions to mitigate toxicity. We report T-cells populations defined by cytotoxicity, memory, and proliferation markers at various stages of colitis. We show preferential depletion of CD8+ T cells with biologic therapy and nominate both circulating and colon-resident T-cell subsets as potential drivers of inflammation and response to immune suppression. Our findings highlight the need for further exploration of the colon immune environment and rationalize future studies evaluating biologics for ICI-colitis, including in the context of ICI re-challenge.

  • CD8-Positive T-Lymphocytes
  • Immune Checkpoint Inhibitors
  • Immunotherapy
  • Programmed Cell Death 1 Receptor
  • T-Lymphocytes

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


This study suggests that distinct colon-resident and circulating T-cell populations, defined by cytotoxicity markers and granzyme K expression, respectively, may contribute to ICI-colitis pathogenesis and treatment response.


Immune-related adverse events (irAEs) impede long-term disease control in many immune checkpoint inhibitor (ICI)-treated patients. ICI-colitis, a common irAE, often necessitates ICI discontinuation and use of immune suppressants which can diminish tumor rejection.1 While mechanisms underlying ICI-induced colitis (ICI-colitis) are incompletely defined, immune cell subsets enriched in ICI-colitis colon tissue have been identified, including expanded CD8+ effector and memory T-cells.2 Understanding the dynamics of these populations in ICI-colitis is crucial for clinical interventions, and circulating markers could enable non-invasive irAE monitoring. We, therefore, present longitudinal single-cell RNA sequencing (scRNA-seq) analyses of peripheral blood and colon samples from an ICI-colitis patient treated with corticosteroids and antibodies against tumor necrosis factor (TNF)-α and integrin-α4β7.


A man in his 70s presented with a growing lesion on his left buttock. Biopsy revealed Merkel cell carcinoma (MCC). Wide local excision revealed a 6 cm wide (pT3) tumor with clear margins. Sentinel nodes were uninvolved. After 6 months he developed in-transit and lymph node metastases. He received three cycles of ICI (pembrolizumab) and developed ICI-colitis (figure 1A). Oral budesonide was started. Five weeks later, sigmoidoscopy revealed severe, diffuse inflammation (figure 1B,C). He received high-dose corticosteroids without improvement, and after 3 days received anti-TNF-α (infliximab) with clinical improvement. He completed a 34-day corticosteroid taper and repeat sigmoidoscopy revealed moderately persistent colitis. After another infliximab dose and three doses of anti-integrin-α4β7 (vedolizumab) his symptoms resolved. A third sigmoidoscopy revealed mildly persistent colitis. After a 5-month delay and symptom resolution, he received a fourth cycle of pembrolizumab resulting in recurrent, severe colitis. He was treated with additional corticosteroids and vedolizumab with colitis resolution. Further immunotherapy was withheld. He was given chemotherapy without benefit and died from progressive MCC.

Figure 1

(A) Timeline of patient’s clinical signs and interventions (upper) and sample collections (lower). (B) Multiple random colon biopsies revealed diffuse involvement by chronic active colitis, with increased lamina propria plasma cells, and active crypt epithelial injury with apoptosis (C; arrows). (D) Subclustering of colon biopsy-derived CD3+ cells. (E) Proportion of total CD3+ cells in each subcluster for each biopsy sample and dot plot representing expression of select markers in each subcluster. (F) Subclustering of peripheral blood-derived CD3+ cells. (G) Proportion of cells in each subcluster for each blood sample. G. Proportion of total CD3+ cells in each subcluster for each blood sample and dot plot representing expression of select markers in each subcluster. ICI, immune checkpoint inhibitor; IFX, infliximab; IMDC, immune-mediated diarrhea and colitis; MAIT, mucosal-associated invariant T cell; Pred, prednisone; Tcm, central memory T cell; Teff, T effector cell; Teff-GZMK, GZMK-positive Teff; Tregs, regulatory T cells; Trm, resident memory T cell; tSNE, t-distributed stochastic neighbor embedding; VDZ, vedolizumab.

Using 10X Genomics scRNA-seq, transcriptomic and T-cell receptor (TCR) libraries were analyzed from four peripheral blood and three colon samples collected at five time points: ICI-naïve (‘Baseline’), during severe initial ICI-colitis (‘Tox-1’), after infliximab (‘Tox-2’), after infliximab and vedolizumab (‘Post-immune-suppression’), and after pembrolizumab re-challenge (‘ICI re-challenge’; figure 1A). Specimens were processed and sequenced as described (online supplemental materials).3

Supplemental material

High quality transcriptomes were obtained from 17,725 colonic and 11,842 circulating cells. Among CD3+ cells, we identified 9 subclusters with productive TCR-β chains in the colon and 10 in blood (figure 1D-G; online supplemental figures S1–S3). Clusters were annotated by reported cluster-identifying gene sets2 4 including markers for cytotoxicity and activation (GZMB/A, PRF1, NKG7, IFNG, HLA-DRA), exhaustion and dysfunction (PDCD1, LAG3, CTLA4), and memory (CD103/ITGAE; figure 1E,G and online supplemental figures S4 and S5; online supplemental table S1).

Supplemental material

Supplemental material

Transcriptional programs and differential depletion of colonic T-cells

Colonic CD8+ clusters included a effector T-lymphocyte (Teff) cluster with enhanced GZMK expression (Teff-GZMK) and two clusters expressing GZMA, ITGAE, and IL7R, designated resident memory T-cells (‘Trm-1’, ‘Trm-2’; figure 1F). TRAV1-2, KLRB1, and SLC4A10 expression defined mucosal-associated invariant T-cells (MAITs).2 5 A ‘Cycling’ cluster showed proliferation markers (MKI67, STMN1). CD8+ T cells, especially Teff-GZMK, Trm-1 and cycling populations, diminished following immune suppression and reduction in ICI-colitis severity. In contrast, CD4+ T cells, including CD4-1, CD4-2, CD4-3, regulatory T cells (Tregs), and a cluster highly expressing Interleukin-17A (IL17A) and markers of dysfunction, increased or remained unchanged over these same conditions. CD8+ T cells, especially Teff-GZMK, Trm-1 and cycling populations, diminished following immune suppression and reduction in ICI-colitis severity. In contrast, CD4+ T cells, including CD4-1, CD4-2, CD4-3, Tregs and a cluster highly expressing IL17A and markers of dysfunction, increased or remained unchanged over these same conditions. Integrins α4 and β7 were expressed in all T-cell clusters before vedolizumab. Most cells post-immune-suppression, particularly Tregs, lacked ITGB7.

Circulating cytotoxic T-cell dynamics

We found five CD4+ circulating clusters (figure 1F,G; online supplemental figure S3), including one expressing cytotoxicity markers (GZMB, PRF1). Six CD8+ clusters included naïve T-cells, MAITs, and three Teff clusters expressing cytotoxicity and early-to-late activation markers (‘Teff1-3’; figure 1G; online supplemental figures S4 and S5). Teff1/2 abundance remained stable. Teff3 dramatically expanded following pembrolizumab and contracted after immune-suppression. Like colonic Teff-GZMK cells, Teff3 highly expresses GZMK and CCL5, late activation markers (HLA-DRA, IFNG, CD38), and dysfunction markers (TIGIT, LAG3).

Clonal expansion

We mapped clonotypes onto CD3+ clusters. TCRαβ pairs were assigned for 4022/4872 circulating and 4740/5224 colon-derived T-cells. We defined ‘expanded’ clonotypes as representing >0.1% and with >2 observations in any sample. Consistent with ICI-induced inflammation, clonal expansion of circulating cytotoxic CD8+ and CD4+ cells was observed at Tox-1 with reduced percentages of unique clonotypes (figure 2A, online supplemental figure S6A). Increased clonotype diversity was observed in the post-vedolizumab colon, concurrent with clinical/pathologic improvement (online supplemental figure S6B).

Figure 2

(A) Abundance of clonotypes by expansion status in blood. (B) The most prevalent clonotypes in the indicated clusters were visualized across colon samples via alluvial plot (compareClonotypes function in scRepertoire). (C) Select clonotypes shared between blood and colon at Tox-1 are represented for each of the indicated samples by alluvial plot. No Teff-GZMK or cycling cells were observed post-immune suppression (Cl. #, Clonotype number). (D) Shared clonotypes that were expanded in blood were visualized by alluvial plot across blood samples. (E) Clonotypes of interest were highlighted as above on the tSNE plot representing all four blood samples. (F) Differentially expressed genes for circulating-only clonotypes versus shared clonotypes. (G) Expression of genes involved in targetable T-cell trafficking and homing mechanisms in the indicated circulating clonotypes. MAIT, mucosal-associated invariant T cell; Teff, T effector cell; Teff-GZMK, GZMK-positive Teff; Trm, resident memory T cell; tSNE, t-distributed stochastic neighbor embedding.

Expanded colonic clonotypes localized primarily to Trm-1, Cycling, and Teff-GZMK subsets (figure 2B). Clonotype-1 was most abundant, comprizing 18% and 55% of Tox-1 Cycling and Trm-1, respectively. At Tox-2, Clonotype-1 represented 3% of Cycling cells and 15% of Trm-1. Markers for activation, cytotoxicity, and memory were enriched in Clonotype-1 compared with other Trm-1 cells and all other colonic T cells (online supplemental table S2). Meanwhile, other Trm clonotypes (7, 30, 40) expanded from Tox-1 to Tox-2. Clonotype-40, which was rarer at earlier stages, persisted after vedolizumab.

Supplemental material

TCR sharing between blood and colon

One hundred and twelve circulating clonotypes were observed in the colon, excluding those with multiple alpha-chains or beta-chains. Nine were expanded in blood and colon (figure 2C,D, online supplemental table S2). The proportion of shared clonotype cells decreased with clinical improvement at Tox-2 and post-immune-suppression (online supplemental figure S6C). Clonotype-1 was never expanded in the blood. TCR sharing was enriched in Teff2–3 in blood and Teff-GZMK in colon compared with other clusters (online supplemental figure S6D,E).

Abundance and transcriptional shifts

Colonic clones responded differentially to infliximab. Clonotypes 5, 11, and 40, for example, persisted from Tox-1 to Tox-2, while Clonotype-1 contracted. Genes enriched in Clonotype-1 compared with other Trm-1 cells and to all other colonic T cells included the proinflammatory chemokine CCL5 along with markers of activation, cytotoxicity, and memory (IFNG, GZMB, KLRD1, CD160, ZNF683, ITGAE; online supplemental table S3).

Supplemental material

Clonotype-3, observed in Teff2/3, was the most prevalent circulating clonotype at all time points (figure 2C,D). At Tox-1, Clonotype-3 upregulated cytotoxicity markers and expanded dramatically, comprizing 32% of circulating CD8+ cells (online supplemental table S2; online supplemental figure S8). Colonic CD8+ clusters were more diverse, with Clonotype-3 comprizing only 1.8% at Tox-1 (online supplemental table S2). Compared with circulating Clonotype-3 cells, colonic Clonotype 3 cells exhibit higher overall levels of the proinflammatory chemokine CCL4, cytotoxicity marker granzyme B, and tissue retention marker CD69 (online supplemental figure S7A).

Clonotype-2 MAITs were observed in both colon and blood throughout. Both populations highly expressed S1PR5, associated with trafficking to the gut. As in Clonotype-3, Colon-infiltrating Clonotype-2 cells expressed greater activation marker levels (CCL4, GZMB; online supplemental figure S7B) than circulating cells.

We defined differentially-expressed circulating T-cell markers between shared expanded clonotypes and those unique to blood (figure 2E, online supplemental table S4). Shared clonotypes were enriched for CD8, CCL5, and activation markers. To evaluate known T-cell trafficking and homing mechanisms, we evaluated expression of vedolizumab target genes ITGA4/ITGB7 and Sphingosine 1-phosphate receptor (S1PR) modulator targets, S1PR1 and S1PR5, in shared clonotypes (figure 2F and online supplemental figures S9 and S10). Gut-trafficking genes were variably expressed among circulating clones, with highest expression in Teffs and CD4 cells. S1PR expression in the colon was observed primarily in MAITs (online supplemental figure S9).

Supplemental material


We present the first analysis of clonal dynamics in blood and colon in ICI-colitis progression and therapy. Observations in inflamed colon tissue echo Luoma et al, who identified a ‘Cycling’ cluster, MAITs, Tregs, and CD8+T cells with memory and cytotoxicity markers.2 We extended these findings by longitudinal tracking and assessing if colon-infiltrating T-cell subsets are identifiable in peripheral blood and could present an avenue for non-invasive prognostication and monitoring. Thomas et al reported a modest increase in circulating GZMK-expressing CD8+ T cells in ICI-colitis, which shared TCRs and transcriptional features with a subset of colonic T-cells.6 We expand this observation, describing a population (Teff-GZMK) sharing clonotypes and transcriptional features (CD8A, GZMK, EOMES) with hyperexpanded circulating Teffs (Teff1-3), preferentially depleted in the colon after vedolizumab. As the disproportionate depletion of Teff-GZMK and other CD8+ subsets (MAIT, Trm) coincided with clinical improvement, our data support further investigation of this therapeutic strategy.

Treg trafficking inhibition is a potential mechanism for vedolizumab resistance in inflammatory bowel disease (IBD).7 Interestingly, post-vedolizumab persistent Tregs lacked integrin-α4β7 expression, suggesting a vedolizumab-independent homing mechanism.

We identified MAIT cells in both blood and colon with high gut-homing gene expression. MAIT blockade can reduce inflammation in experimental models,8 while IBD studies showed increased MAIT activation relative to control patients.2 Interestingly, transcriptional differences were noted in tissue-infiltrating ICI-colitis MAITs compared with IBD, including GZMB upregulation.9 Similarly, we show substantial GZMB and CCL4 upregulation in colonic versus circulating clonotype-2 MAITs, with modestly enhanced EOMES, CCL5, GNLY, and CD38, suggesting an activated state. Thus, activation and persistence of MAITs and their antigens may contribute to ICI-colitis and therapeutic response.

Mirroring Thomas’ findings in ICI-colitis and healthy controls, ~2% of clonotypes were shared between blood and colon. Some colonic T-cells shared features with circulating cells, including Clonotype-3, which expanded dramatically in the blood and remained abundant, but this disproportionate representation is not reflected in Teffs, raising the possibility of a tumor-specific or bystander clone.

We observed colon-exclusive Trm populations that may contribute to colitis. Clonotype-1, shared between Cycling and Trm-1, expresses cytotoxicity markers, suggesting activation and proliferation of a colon-resident population. Clonotype-1 contracted after corticosteroids and infliximab and was not observed after vedolizumab, suggesting a locally active colonic clone. Meanwhile, clonotype 40 was the only Trm clonotype persisting after vedolizumab despite representing only a small proportion of Trms at Tox-1 and Tox-2. The persistence of this clone is interesting given that the ICI-colitis recurred. These observations highlight the complex contributions of diverse T-cell populations to ICI-colitis. Further they raise the question of whether targeting T-cell trafficking is sufficient to manage ICI-colitis driven partially by resident populations.

Importantly, our study does not compare T-cell changes to those in ICI-treated patients without colitis, and larger studies will be needed to interrogate whether the features we report are ICI-colitis specific. Additionally, interpretation of post-immune suppression colonic T-cell subsets is limited by low cell count (45 CD3+ transcriptomes). Interestingly, Teffs were entirely diminished while Tregs and CD4 cells were disproportionately maintained. Trm and MAIT cells were also observed after immune-suppression. These persisting populations may be key to understanding recurring colitis.

Our data nominate multiple T-cell subsets as potential drivers of ICI-colitis,. Due to shared clonotypes and transcriptional programs, we raise the possibility of using circulating GZMK-expressing Teffs (Teff1-2) for non-invasive monitoring of ICI-colitis. Populations identified responded differentially to more systemic (anti-TNF-α) versus gut-targeted (anti-α4β7 integrin) therapies, indicating multiple disease mechanisms that may complicate ICI-colitis management. Furthermore, our data suggest that trafficking of CD8+ T cells to the colon plays a major role in ICI-colitis, and drugs inhibiting homing mechanisms should be investigated for prophylaxis in patients re-challenged with ICIs. Future studies comparing pretreatment and on-treatment samples from ICI-treated patients with and without colitis will be crucial to identifying colitis-associated biomarkers in circulating and colonic cells.

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and was approved by Yale Human Investigation Committee #0608001773. Participants gave informed consent to participate in the study before taking part.


The authors thank the staff of the Yale Center for Genome Analysis for technical assistance with single-cell genomics and the staff of the Yale Flow Cytometry Facility for technical assistance with flow cytometry.


Supplementary materials


  • Contributors JEM: Data Curation, Formal Analysis, Visualization, Funding Acquisition, Writing—Original Draft Preparation, Writing—Review and Editing. LL: Data Curation, Formal Analysis, Investigation, Methodology, Writing—Review and Editing. MRA: Resources, Writing—Review and Editing. RDM: Investigation, Writing—Original Draft Preparation, Writing—Review and Editing. MER: Visualization, Writing—Review and Editing. BAB: Resources. KR: Investigation. LA: Writing—Review and Editing. NJ: Writing—Review and Editing. DH: Project Administration, Funding Acquisition. CA: Writing—Review and Editing. KCH: Writing—Review and Editing, Funding Acquisition, Supervision. HMK: Conceptualization, Funding Acquisition, Project Administration, Resources, Supervision, Writing—Review and Editing.

  • Funding JEM was supported by the Yale Human and Translational Immunology Training program (1T32AI155387091A1). This work was also supported in part by the National Institutes of Health grant R01 CA227472 to HMK and KCH and the Yale SPORE in Skin Cancer (P50 CA121974 to HMK, MB, SK and YK).

  • Competing interests MER reports consultant fees from Takeda, Bristol-Myers Squibb, and ImmunogenX. DH received research funding from Bristol-Myers Squibb, Sanofi, and Genentech. He has been a consultant for Bristol-Myers Squibb, Compass Therapeutics, EMD Serono, Genentech, Juno Therapeutics, Novartis Pharmaceuticals, Proclara Biosciences, Sage Therapeutics, and Sanofi Genzyme. HMK reports institutional research grants from Merck, Bristol-Myers Squibb and Apexigen and financial support from Iovance, Celldex, Merck, Elevate Bio, Instil Bio, Bristol-Myers Squibb, Clinigen, Shionogi, Chemocentryx, Calithera, Signatero, Gigagen, GI reviewers, Seranova, Pliant Therapeutics. No other disclosures were reported.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.