Single-cell analysis reveals clonally expanded tumor-associated CD57+ CD8 T cells are enriched in the periphery of patients with metastatic urothelial cancer responding to PD-L1 blockade

Abstract

Background: A growing body of evidence suggests that T-cell responses against neoantigens are critical regulators of response to immune checkpoint blockade. We previously showed that circulating neoantigen-specific CD8 T cells in patients with lung cancer responding to anti-Programmed death-ligand 1 (PD-L1) (atezolizumab) exhibit a unique phenotype with high expression of CD57, CD244, and KLRG1. Here, we extended our analysis on neoantigen-specific CD8 T cells to patients with metastatic urothelial cancer (mUC) and further profiled total CD8 T cells to identify blood-based predictive biomarkers of response to atezolizumab.

Methods: We identified tumor neoantigens from 20 patients with mUC and profiled their peripheral CD8 T cells using highly multiplexed combinatorial tetramer staining. Another set of patients with mUC treated with atezolizumab (n=30) or chemotherapy (n=40) were selected to profile peripheral CD8 T cells by mass cytometry. Using single-cell transcriptional analysis (single-cell RNA sequencing (scRNA-seq)), together with CITE-seq (cellular indexing of transcriptomes and epitopes by sequencing) and paired T-cell receptor (TCR) sequencing, we further characterized peripheral CD8 T cells in a subset of patients (n=16).

Results: High frequency of CD57 was observed in neoantigen-specific CD8 T cells in patients with mUC responding to atezolizumab. Extending these findings to bulk CD8 T cells, we found higher frequency of CD57 expressing CD8 T cells before treatment in patients responding to atezolizumab (n=20, p<0.01) but not to chemotherapy. These findings were corroborated in a validation cohort (n=30, p<0.01) and notably were independent of known biomarkers of response. scRNA-seq analysis identified a clonally expanded cluster enriched within CD57⁺ CD8 T cells in responding patients characterized by higher expression of genes associated with activation, cytotoxicity, and tissue-resident memory markers. Furthermore, compared with CD57^- CD8 T cells, TCRs of CD57⁺ CD8 T cells showed increased overlap with the TCR repertoire of tumor-infiltrating T cells.

Conclusions: Collectively, we show high frequencies of CD57 among neoantigen-specific and bulk CD8 T cells in patients responding to atezolizumab. The TCR repertoire overlap between peripheral CD57⁺ CD8 T cells and tumor-infiltrating lymphocytes suggest that accumulation of peripheral CD57⁺ CD8 T cells is reflective of an ongoing antitumor T-cell response. Our findings provide evidence and rationale for using circulating CD8 T cells expressing CD57 as a readily accessible blood-based biomarker for selecting patients with mUC for atezolizumab therapy.

Keywords: CD8-positive T-lymphocytes; immunotherapy; receptors, antigen; urinary bladder neoplasms.

Figure 1

Patients responding to atezolizumab treatment exhibit several unique neoantigen specificities with a differentiated effector T-cell phenotype. (A) Number of unique neoantigen specificities detected from a total of 656 neoantigen candidates tested across atezolizumab responder and non-responder patients. (B) Frequencies of neoantigen-specific CD8 T cells detected within all patients. Matching colored dots and corresponding numbers indicate the same neoantigen specificities detected at baseline and on-treatment time points. Connecting lines visualize the detection of the same antigen specificities at both time points. (C) Expression of phenotypical markers by neoantigen-specific CD8 T cells from responder and non-responder patients at baseline and on atezolizumab treatment. Data shown are median values. *P<0.05, **P<0.01. Wilcoxon rank-sum test. P values were adjusted for multiple testing using the Benjamini-Hochberg method to control the false discovery rate.

Figure 2

Late-differentiated CD57⁺ CD8 T cells are enriched in atezolizumab responder patients at baseline. (A, B) Frequencies of CD8 T cells positive for different phenotypical markers (A) or CD57 (B) in responders (n=11) and non-responders (n=9) assessed in baseline blood. Data shown are median values. *P<0.05, **P<0.01. Wilcoxon rank-sum test. P values were adjusted for multiple testing using the Benjamini-Hochberg method to control the false discovery rate. (C) Representative staining examples for CD57 expression on CD8 T cells. (D) Frequency of CD57⁺ CD8 T cells at baseline (pre) and on-treatment (post). Wilcoxon matched-pairs signed rank test. (E) Average expression of CD57 on CD8 T cells from responders and non-responders at baseline. **P<0.01. Wilcoxon rank-sum test. (F) Representative histogram examples for CD57 expression on CD8 T cells. (G) Correlogram showing correlation between frequency of CD57⁺ CD8 T cells and frequency of other phenotypical markers analyzed on bulk CD8 T cells. Pearson’s correlation coefficients were indicated by a heat scale whereby red color shows positive linear correlation, and blue color shows negative linear correlation. *P<0.05, **P<0.01. ns, non-significant.

Figure 3

CD57⁺ CD8 T cells are enriched in patients responding to atezolizumab treatment but not chemotherapy. (A) CD57⁺ CD8 T-cell frequency validation cohort of atezolizumab-treated patients in baseline peripheral blood (n=15 in each group). (B) CD57 expression on CD8 T cells in baseline peripheral blood from responder and non-responder patients treated with chemotherapy (n=20 each group). Wilcoxon rank-sum test. (C) Heatmap showing the frequencies of all phenotypical markers assessed on CD57⁺ CD8 T cells and CD57^ー CD8 T cells from atezolizumab responders and non-responders at baseline. (D) Expression of CD57 on major CD8 T-cell subsets (naïve): CD45RA⁺, CCR7⁺; T_CM: CD45RA^ー, CCR7⁺; T_EM: CD45RA^ー, CCR7^ー; and T_EMRA: CD45RA⁺, CCR7^ー in atezolizumab-treated responders and non-responders in peripheral blood. *P<0.05, **P<0.01. Wilcoxon rank-sum test. ns, not significant; T_CM, T central memory; T_EM, T effector memory; T_EMRA, T effector memory cells expressing CD45RA.

Figure 5

Single-cell analysis of CD57⁺ CD8 T cells reveals a unique cluster enriched in responding patients. (A) CD57⁺ frequencies among CD8 T cells at baseline calculated using CITE-seq analysis, p=0.037; two-tailed unpaired Student’s t-test. (B) Volcano plot showing DEGs in CD57⁺ CD8 T cells at baseline. DEGs are nominated by requiring at least 1.25 times fold change and an adjusted p value of <0.05 with gene expression detected in at least 10% of cells in either one of the two comparison groups. Selected DEGs are labeled in the volcano plot, and red dots represent DEGs. (C) Heatmap displaying the scaled expression of upregulated genes ranked by fold change between CD57⁺ CD8 T cells from responders and non-responders at baseline. (D) UMAP (uniform manifold approximation and projection) of CD57⁺ CD8 T cells profiled by scRNA-seq (n=8406). Cluster 1 is highlighted by a dotted line. (E) Proportions of individual clusters (as defined in D) out of all CD57⁺ CD8 T cells were compared in responders versus non-responders at baseline. Two-tailed unpaired Student’s t-test. (F) UMAP plot showing the expression of top genes identified to be enriched in cluster C1. Cluster 1 is highlighted by a dotted line. DEG, differentially expressed gene; ns, not significant; scRNA-seq, single-cell RNA sequencing.

Figure 4

Comparison of tumor PD-L1 score and TMB with CD57⁺ CD8 T cells in peripheral blood from responders and non-responders. (A) Heatmap showing frequency of CD57⁺ peripheral blood CD8 T cells, tumor PD-L1 immunohistochemistry score on IC and TC, as well as TMB score in all patients. (B) Bar chart representing the proportion of responders to non-responders by PD-L1 IC score status. P value is calculated by Χ² test. (C) Bar chart representing the median TMB score in each response group. (D) Frequency of CD57⁺ peripheral blood CD8 T cells in patients separated based on PD-L1 IC score status. Wilcoxon rank-sum test. (E) Correlation plot showing the association between TMB score and frequency of CD57⁺ peripheral blood CD8 T cells at baseline. P value (two-tailed t-test) is shown. (F) Association of CD57⁺ CD8 T-cell frequency and overall survival in atezolizumab-treated patients (discovery and validation cohorts are combined) in the IMvigor210 cohort. CD57 cut-offs are defined based on the frequency of CD57⁺ CD8 T cells in all patients and using median cut-off (25.9%) to divide them into CD57 high or CD57 low groups. HR and CI are calculated using Cox proportional hazard regression models, and p value is calculated using log-rank test. IC, immune cell; ns, not significant; TC, tumor cell; TMB, tumor mutation burden.

Figure 6

Increased clonal expansion in CD57⁺ CD8 T cells in responders compared with non-responders. (A) Box plots showing T-cell receptor evenness (evaluated by the Gini index) across CD57⁺ and CD57⁻ CD8 T cells in responders and non-responders at baseline. (B) Box plots showing the proportion of clonal cells in CD57⁺ and CD57⁻ CD8 T cells in responders and non-responders at baseline. Percentage of cells defined based on clone size (n≥2, left; n≥10, right) among CD57⁺ and CD57⁻ CD8 T cells are shown. (C) Box plots showing Gini index at baseline in three clusters C0 (left), C1 (middle) and C2 (right) identified among CD57⁺ CD8 T cells (referring to figure 5D, E). P values shown are calculated based on two-tailed unpaired Student’s t-test. Gini index measured in (A) and (C) is limited to samples with at least 10 cells. (D) Frequency distribution of CD57⁺ (left) and CD57^ー (right) CD8 T cells based on individual TCRB clonotype sequences. Each slice size represents the percentage of cells with individual TCR clonotypes. TCR sequences that overlap between blood and tumor are highlighted in color.