Immune-awakening revealed by peripheral T cell dynamics after one cycle of immunotherapy

Abstract

Our understanding of how checkpoint inhibitors (CPI) affect T cell evolution is incomplete, limiting our ability to achieve full clinical benefit from these drugs. Here we analyzed peripheral T cell populations after one cycle of CPI and identified a dynamic awakening of the immune system revealed by T cell evolution in response to treatment. We sequenced T cell receptors (TCR) in plasma cell-free DNA (cfDNA) and peripheral blood mononuclear cells (PBMC) and performed phenotypic analysis of peripheral T cell subsets from metastatic melanoma patients treated with CPI. We found that early peripheral T cell turnover and TCR repertoire dynamics identified which patients would respond to treatment. Additionally, the expansion of a subset of immune-effector peripheral T cells we call TIE cells correlated with response. These events are prognostic and occur within 3 weeks of starting immunotherapy, raising the potential for monitoring patients responses using minimally invasive liquid biopsies."

Conflict of interest statement

Conflict of Interest: RM is a consultant for Pfizer and has a drug discovery programme with Basilea Pharmaceutica. PL serves as paid advisor/speaker for Bristol-Myers Squibb, Merck Sharp and Dohme, Roche, Novartis, Amgen, Pierre Fabre, Nektar, Melagenix. PL reports travel support from Bristol-Myers Squibb and Merck Sharp and Dohme, and receives research support from Bristol-Myers Squibb. AG received honoraria and consultancy fees from BMS and Novartis.

Figures

Extended Data Fig. 1. Schematic summarizing T cell maturation and life-cycle.

a Pro-T cells undergo sequential somatic recombination of their T cell Receptor β (TCR) loci in attempts to generate functional TCR with unique CDR3 antigen binding regions. Cells that fail to generate a functional TCRβ at the first attempt can recombine their second TCR allele, but cells which fail to produce a functional TCR at the end of the process (crossed red box) are eliminated (β-selection) and their DNA, which encodes the CDR3unique regions, enters the blood as circulating cell-free DNA (cfDNA). Surviving cells retain the T cell receptor excision circle (TREC) generated during TCR locus rearrangement as an episome in the nucleus. The TREC does not replicate so is diluted during subsequent cell divisions. b T cells with a functional TCR undergo positive and negative selection (+/- selection) for HLA and self-antigen recognition. TheCDR3 DNA from T cells eliminated during this step is released into the blood. c Naive T cells enter the circulation as early thymic emigrants (ETE). d T cells primed by antigen presenting cells (APC) in the lymphatic system undergo clonal expansion, which dilutes the TREC amongst the daughter cells. e T cell homeostasis is maintained by subsequent contraction (turnover cycles), releasing further CDR3 DNA into the blood.

Extended Data Fig. 2. Gating strategy for the identification of T cell subsets in peripheral blood of melanoma patients.

Multiparametric fluorescence activated cell sorting analysis using the indicated gates. a Lymphocyte gate on side scatter/forward scatter; b single cell gate to exclude doublets; clive gating to exclude dead cells from subsequent gates; dCD3+ gate for T cells; e,f CD4+ and CD8+ gates for “helper” and “killer” T cell subsets, CD8 was detected with a PE-Cy7 labelled antibody for the Treg panel (e) and with a FITC labelled antibody for the T maturation panel (f); gCD4+/CD25+/CD127-/low regulatory T cells (Treg); h naive (top left) and memory (bottom right) gates total T cells; i ETE (top) and CD31-naive (bottom) gates for naive T cells; j naive (top left) and memory (bottom right) gates for CD8+ T cells; k CD8+ memory T cell subsets, the left bottom subset (CCR7-/CD27-) represents the TIE cells.

Extended Data Fig. 3. Clonal relatedness in tumor infiltrating T cells and PBMC.

a Clonal relatedness changes in PBMC-private and TIL-private TCR pools; comparison of week 3 (W3)CDR3 clonal relatedness in patients with progressive disease (PD, n=11 patients) and disease control at week 12 (DC, n=7 patients) in the PBMC-private (P=0.724, median=0.6x10-6 and 0.6x10-6, respectively; two-sided Mann-Whitney U test) and TIL-private pools (P=0.246, median= 0.5x10-4 and 0.8x10-5, respectively; two-sided Mann-Whitney U test). Dot represents one patient; green indicates DC; orange indicates PD; error bar is standard deviation.

Extended Data Fig. 4. Identification of TIE in CPI-treated patient PBMC.

a Comparison of differential abundance of TIE in CD8+ memory T cells in the PBMC of The Christie NHS Foundation Trust patients with best response progressive disease (PD, orange, n=14) and disease control (DC, green, n=16) at T0 (n=30, light shade) and week 9 (W9; n=10, dark shade; PD, n=4, DC, n=6). Differences over time were not significant for PD (median=15.2 and 35.5; P=0.375; two-sided Wilcoxon test) or DC (median=7.9 and 24; P=0.219; two-sided Wilcoxon test); PD vs DC patient values did not differ at T0 (P=0.275; two-sided Mann-Whitney U test) or W9 (P=0.762; two-sided Mann-Whitney U test). b Distributions of marker intensities of the T cell surface markers in the 20 cell populations (clusters) for PBMC from a published cohort3 (n=20 patients). Cluster 5 was identified as the TIE subset. Blue densities are calculated over all the cells and serve as a reference and red densities represent marker expression for cells in a given cluster. Arrows highlight the TIE subset. c T-stochastic neighbor embedding of single cell profiles (dots) performed in an external cohort3 using the T cell surface markers CD3, CD4, CD8, CD45RA, CD45RO, CCR7 and CD27; different colors are attributed by clustering. Arrow highlights the TIE subset. d Comparison of the differential abundance of the TIE cluster in the PBMC from a published cohort3 of patients with PD (orange, n=9) or DC (green, n=11) at pre-treatment (light shade, n=20; PD, n=9; DC, n=11) and at week 12 (W12, dark shade, n=20) on treatment with pembrolizumab or nivolumab in the external cohort. Horizontal bars indicate the differences over time for the PD (median at T0=5.9 and W12=9.1; P=0.164; two-sided Wilcoxon test) or DC patients (median at T0=3.8 and W12=3.3; P=0.831; two-sided Wilcoxon test), and difference in the two response groups at T0 or W12 (P=0.37 and P=0.201, respectively; two-sided Mann-Whitney U test). Light and dark orange indicate PD for T0 and W9-W12, respectively, light and dark green indicate DC for T0 and W9-W12, respectively; n represents patients; ns means not significant P values; error bars are standard deviation.

Extended Data Fig. 5. Characterization of TIE in PBMC.

Analysis of published cohort of PBMC single cell data from reference #27. a Violin plots of the expression level of selected phenotypic and transcriptomic features of the clusters identifying peripheral T cell subsets (n=7488 single cells), the cluster with TIE phenotype is indicated in red; the plots represent the density probability, the area shapes reflect the data distribution; horizontal lines represent the minima and maxima values; central dots represent the medians. Overall minima, mean and maxima values: surface CD3=0, 0.3785, 4.1396; surface CD8a=0, 0.96327, 6.21476; surface CD45RA=0, 0.8161, 4.8508; surface CD45RO=0, 0.6628, 4.6468; surface CD197/CCR7=0, 0.8961, 5.7975; surface CD69=0, 0.5219, 4.2200; surface CD279=0, 0.09787, 3.84886; surface CD25=0, 0.08653, 4.00428; surface TIGIT=0, 0.4663, 4.2381; surface CD155=0, 0.4850, 4.6679; surface CD40=0, 0.6003, 5.5083; surface CD154=0, 0.4062, 3.8159; surface CD357=0, 0.1193, 4.0316; LGALS2=0, 0.561, 6.089; TYROBP=0, 1.337, 6.662; FCN1=0, 1.290, 6.789; CST3=0, 1.404, 6.504; LST1=0, 1.042, 6.097; LYZ=0, 1.775, 6.859. b T-SNE plot showing the clusters identified by means of the antibody derived tags (ADT) targeted to surface markers (n=7488 single cells); the black arrow indicates the cluster with TIE phenotype. c Plot showing the proportion of cells with the TIE phenotype from the same published cohort after standard in vitro culture (CTRL, n=3 sorted healthy donor peripheral blood CD8+ naïve T cell samples in standard culture) or following stimulation with anti-CD3/anti-CD27 Dynabeads (STIM, n=3 sorted healthy donor peripheral blood CD8+ naïve T cell samples after stimulation) (P=0.0267, two-sided paired t test, two degrees of freedom) andd Volcano plot representing the transcriptomic differential expression of the cells with the TIE phenotype in PBMC presented in a (n=7488 single cells) or expanded from naive CD8+ T cells from the experiment presented inc(n=12217 single cells; two-sided Wilcoxon test with Bonferroni correction for multiple comparisons).

Extended Data Fig. 6. Expression of Ki-67 and PD-1 in peripheral TIE cells before and after 1 cycle of CPI.

a Expression of Ki67 and PD1 in the TIE subset as measured by FACS in n=5 frozen samples of PBMC from The Christie NHS Foundation Trust metastatic melanoma patients treated with CPI, at pre-treatment (T0) and after 1 cycle of CPI (W3); horizontal line indicates median; error bar indicates standard deviation. The small sample size did not allow statistical comparison of the outcome groups.

Figure 1. CPI induced peripheral TCR repertoire divergence.

a Graph showing early thymic emigrants in pre-treated patients’ blood (% ETET0 relative to total naive T cells; determined by FACS) relative to age (P=0.002, linear regression R2=-0.17; n=50). b Levels of ETE in pre-treatment (T0) and week 3 (W3) of CPI in paired patient samples (P=0.274, two-sided Wilcoxon test, n=50). c TREC (T cell receptor excision circle) concentration relative to genomic DNA was measured by droplet digital PCR in sorted CD3+ peripheral T cells at T0 (median 0.5x10-3) and W3 (median 0.1x10-2)(P=0.129, two-sided Wilcoxon test, n=17).d Tumor infiltrating T lymphocyte (TIL) CDR3sequences also present in peripheral PBMC and cfDNA for one patient at T0 and W3. See Table 1 for specific DNA sequence; tot = total. e Venn diagram showing unique predicted productive CDR3 sequences in PBMC and TIL for patient #01 at T0 (Supplementary Table 1). Numbers show unique nucleotide sequence counts for PBMC-private (pink), TIL-private (brown) and tePBMC (tumor emigrant PBMC; intersection, orange) pools. fClonal relatedness (the proportion of amino acids sequences that are related by maximum edit distance=3) for CDR3 in the PBMC-private pool, tePBMC and TIL-private pools at T0. Horizontal lines: comparison of clonal relatedness between PBMC-private and TIL-private TCR sequences at T0; ***: P=0.003; n=18, two-sided Wilcoxon test; median=0.4x10-6 and 0.4x10-3, respectively; ****: P<0.0001; n=18, two-sided Wilcoxon test; median=0.4x10-6 and 0.2x10-2, respectively.g Clonal relatedness (maximum edit distance=3 amino acids) for CDR3 sequence in PBMC TCR pools at T0 and W3. Comparison between the clonal relatedness of PBMC-private TCR of patients with progressive disease (PD, orange, n=11, median=4.3x10-5 and 5.6x10-5, respectively) or disease control (DC, green, n=7, median=4.0x10-5 and 8.0x10-5, respectively) after 12 weeks of treatment; ns: not significant (P=0.413 and P=0.999, two-sided Wilcoxon test) and between the clonal relatedness of tePBMC TCR of patients with PD (n=11) or DC (n=7); ns: not significant (P=0.638; two-sided Wilcoxon test; median=0.002 and 0.0008, respectively); *: P=0.031; n=7, two-sided Wilcoxon test; median=0.0017 and 0.0007, respectively). Dot is one patient; line is median; error bar is standard deviation; connecting line is paired samples; ns indicates not significant P values, n represents patients.

Figure 2. CPI induced peripheral T cell turnover.

a Venn diagram showing unique predicted productiveCDR3 sequences in PBMC (left, pink), PBMC/cfDNA-shared pool (intersection, purple) and cfDNA (right, blue) for patient #27 at T0 (Supplementary Table 2).b Total number of CDR3 clones at T0 (pink) and W3 (purple) in the PBMC/cfDNA-shared pool (P=0.010, two-sided Wilcoxon test).c Clonal relatedness (maximum edit distance=3 amino acids) for CDR3 in the PBMC-private pool, PBMC/cfDNA-shared pool and cfDNA-private pools at T0. Horizontal lines: comparison of clonal relatedness between PBMC-private and cfDNA-private TCR sequences T0 (P<0.0001; two-sided Wilcoxon test; median was 0.3x10-3 and 0.01, respectively); comparison of clonal relatedness between PBMC-private and PBMC/cfDNA shared TCR sequences T0 (P<0.0001, two-sided Wilcoxon test; median was 0.3x10-3 and 0.06). d Clonal relatedness of CDR3 sequence in PBMC/cfDNA-shared pool at T0 and W3 for patients with progressive disease (the number of patients is 12, PD, orange) or disease control (the number of patients was 16, DC, green) at week 12. Comparison (horizontal lines) of clonal relatedness between: T0 PBMC/cfDNA-shared pool TCR sequences for patients with PD or DC (P=0.623, two-sided Mann-Whitney U test; median is 0.04 and 0.08); W3 PBMC/cfDNA-shared pool TCR sequences for patients with PD or DC (P=0.026; two-sided Mann-Whitney U test; median=0.06 and 0.03); T0 and W3 for the PBMC/cfDNA-shared pool TCR sequences for patients with PD (P=0.733; the number of patients was 12, two-sided Wilcoxon test; median=0.04 and 0.06) or DC (P=0.039; n=16, Wilcoxon test; median=0.08 and 0.03).e Pre-treatment TCR rearrangement efficiency score (REST0) of rearranged CDR3 in healthy donors (HD) and patients on CPI in PBMC (P=0.445; median 0.83 and 0.81; the number of HD was 77 batches and the number of patients was 29; two-sided Mann-Whitney U test) and cfDNA (P=0.09, median 0.44 and 0.62; n=3 and 28; two-sided Mann-Whitney U test). Comparisons (horizontal lines) between: HD PBMC and cfDNA REST0 (P<0.0001, two-sided Mann-Whitney U test); matched samples of patients’ PBMC and cfDNA RES T0 (P<0.0001, Wilcoxon test). f ΔW3RES (change in RES from T0 to W3) according to response group at week 12; **: P=0.008, two-sided Wilcoxon test, median was 0.001 and 0.08; *: P=0.037, two-sided Mann-Whitney U test, median=0.02 and 0.08. Total number of melanoma patients equal to 28; dot is one patient; error bar is standard deviation; connecting line is paired samples; horizontal line is median; T0 indicates pre-treatment; W3 indicates week 3; ns is not significant; * indicates P=0.05-0.01; **** indicates P<0.0001.

Figure 3. Identification of TIE cells.

a Correlation between TIE cell abundance (ΔW3TIE) and changes in cfDNA RES (ΔW3RES: RESW3-REST0,) at W3 relative to T0 (P=0.001; linear regression R2=0.34, n=28 patients); dotted line is the linear regression line. b Similarity matrix ofTCR sequences in cfDNA and peripheral CD8+ T cell subsets. TCM, TIE Tnaive and ETE similarity with cfDNA in 6 patients at T0 (median=0.026, 0.045, 0.004 and 0.003, respectively; P=0.0013; Friedman analysis of variance, Friedman statistics=15.64; patient #16 naive subset not assessed) and W3 (median=0.043, 0.136, 0 and 0.003, respectively; P<0.0001; Friedman analysis of variance, Friedman statistics=23.05; patient #16 naive subset not assessed).c Clonality (Gini coefficient) in peripheral CD8+ T cell subsets. TIE subset clonality relative to other subsets at baseline (T0, TIE median clonality=0.46) and after the first cycle of CPI (W3, TIE median clonality=0.61) in 6 matched patient samples (P=0.0006 and 0.0002, Friedman analysis of variance Friedman analysis of variance, Friedman statistics=12.6 and 13.08; patient #16 naive subset not assessed); error bar is standard deviation. The small sample size did not allow the comparison between responders (#16-18) and patients who progressed (#12,19,29); horizontal line is median; error bar is standard deviation.d Graph showing the frequency of pre-treatment TILCDR3 sequences in patient #12 sorted peripheral CD8+ T cell subsets at T0 and W3. Dot in a and c is a single patient.

Figure 4. TIE cells infiltrate tumors that respond to immunotherapy.

T-SNE (t distributed stochastic neighbor embedding) plots of biopsy cell clusters according to T cell surface markers in (a) melanoma (n=18 patient samples), (b) renal cell carcinoma (RCC, n=3 patient samples), (c) glioblastoma (GMB, n=3 patient samples), and (d) non cancerous tonsils (n=4 patient samples). Samples for this analysis were from reference #24. Black arrows highlight the clusters with the TIE phenotype.

Figure 5. Peripheral TIE cell expansion in response to first-line CPI.

a ΔW3TIE in patients with best response PD (orange, n=14, median=-0.58%) or best response DC (green, n=16, median=10.04%; ***: P=0.0007, two-sided Mann-Whitney U test) in the training set. Arrow: patient #20. b ΔW3TIE in patients receiving anti-PD1 monotherapy (αPD1, n=18, median=1.35%) or combination of ipilimumab plus nivolumab (I+N, n=12, median=10.84%; P=0.2, two-sided Mann-Whitney U test). c Variations over time (days from treatment initiation) of variant allele frequency (VAF) of mutantNRASQ61R measured in circulating tumor DNA by droplet digital PCR over the course of CPI treatment for patient #20 (arrow in a). Blood collection ceased after day 90 due to patient complications leading to patient death. d Receiver operating curve showing the sensitivity and false positive rate (1-specificity) of ΔW3TIE values in identifying the patients that will achieve disease control; *: maximum accuracy (cut-off=+0.8%).e Kaplan-Meier survival curves for patients with TIE expansion ≥0.8% (pink, n=12; median survival not reached) at W3 compared to patients without TIE expansion <0.8% (blue, n=18; median survival=9.6 months; P=0.013; log rank test), in the training set.f ΔW3TIE in patients with best response PD (orange, n=3, median=-1.3%) or best response DC (green, n=17, median=3.3%; *: P=0.019, two-sided Mann-Whitney U test) in the validation set.g Kaplan-Meier survival curves for patients with TIE expansion≥0.8% (pink, n=15; median survival not reached) at W3 compared to patients without TIE expansion≥0.8% (blue, n=5; median survival=4.2 months; two-sided log rank test, P=0.003), in the validation set. OS is overall survival; dot represents one patient; n represents patients; horizontal line is median; error bar is standard deviation; dotted vertical line is landmark at week 3.

Figure 6. Expansion of a peripheral regulatory T cell subset associated with toxicity.

a Graph showing expansion of TIE at W3 (ΔW3TIE) in patients with no ≥grade 3 toxicity (median=1.6; n=33) or ≥grade 3 toxicity (median=3.72; P=0.347; n=17; two-sided Mann-Whitney U test). Line, median; error bar, standard deviation. Orange=PD, green=DC, dot=single patient, triangles=single agent anti-PD1, squares=combination ipilimumab+nivolumab. b Expansion of CD3+/CD4+/CD8-/CD25+/CD127-/low Treg cells (Extended Data Figure 2g) at W3 (ΔW3Treg) according to toxicity at any time between 2 weeks and 6 months (P<0.0001; n=50, two-sided linear regression analysis, R2=0.29). Horizontal line is the median; error bar is standard deviation; orange represents PD; green represents DC; triangle represents one patient treated with single agent anti-PD1 drug; square represents one patient treated with combination ipilimumab+nivolumab; n represents patients; dotted line is linear regression line; ns is not significant; T0 is pre-treatment; W3 is week 3.

Figure 7. TCR repertoire evolution after immune-stimulation.

a Changes in CDR3 clonality (Δclonality, Gini coefficient) and diversity (Δdiversity, Renyi index, alpha=1) in peripheral T cells from T0 to W1-2 in healthy donors who received anti-viral vaccination (n=25 healthy donor samples). b Changes inCDR3 clonality (Δclonality, Gini coefficient) and diversity (Δdiversity, Renyi index, alpha=1) in peripheral T cells from T0 to W3 in the training cohort (The Christie NHS Foundation Trust) of advanced melanoma patients receiving first line anti-PD1 based immunotherapy (n=17 patients) who progressed (n=9 patients) or responded (n=8 patients) at week 12.c Δclonality (Gini coefficient) and Δdiversity (Renyi index, alpha=1) in peripheral T cells from T0 to W3 in the validation cohort (The Christie NHS Foundation Trust, Huang et al. and Amaria et al.cohort n=12 patients, n=4 patients and n=11 patients, respectively) of advanced melanoma patients who progressed (n=11 patients) or responded (n=16 patients) at week 12 of anti-PD1 based treatment. Dot is one healthy donor or patient; maroon represents healthy donors in the vaccination cohort; orange represents patients who progressed after 12 weeks of immunotherapy; green represents patients who achieved disease control after 12 weeks of immunotherapy; dotted line is the linear regression line.