Self-Renewing CD8⁺ T-Cell Abundance in Blood Associates with Response to Immunotherapy

Abstract

Treatment with immune checkpoint blockade (ICB) often fails to elicit durable antitumor immunity. Recent studies suggest that ICB does not restore potency to terminally dysfunctional T cells, but instead drives proliferation and differentiation of self-renewing progenitor T cells into fresh, effector-like T cells. Antitumor immunity catalyzed by ICB is characterized by mobilization of antitumor T cells in systemic circulation and tumor. To address whether abundance of self-renewing T cells in blood is associated with immunotherapy response, we used flow cytometry of peripheral blood from a cohort of patients with metastatic non-small cell lung cancer (NSCLC) treated with ICB. At baseline, expression of T-cell factor 1 (TCF1), a marker of self-renewing T cells, was detected at higher frequency in effector-memory (CCR7⁻) CD8⁺ T cells from patients who experienced durable clinical benefit compared to those with primary resistance to ICB. On-treatment blood samples from patients benefiting from ICB also exhibited a greater frequency of TCF1⁺CCR7⁻CD8⁺ T cells and higher proportions of TCF1 expression in treatment-expanded PD-1⁺CCR7⁻CD8⁺ T cells. The observed correlation of TCF1 frequency in CCR7⁻CD8⁺ T cells and response to ICB suggests that broader examination of self-renewing T-cell abundance in blood will determine its potential as a non-invasive, predictive biomarker of response and resistance to immunotherapy.

Introduction

Antibodies targeting immune checkpoint proteins are now essential components of standard treatment in non-small cell lung cancer (NSCLC) and other tumor types. Nonetheless, only a small subset of patients experience an initial response to treatment with immune checkpoint blockade (ICB), with even fewer patients achieving durable responses (1). Biomarkers that predict initial and durable responses for patients treated with ICB remain elusive.

Recent advances in understanding the mechanism-of-action of ICB have revealed that PD-1 blockade drives proliferation and differentiation of self-renewing progenitor T cells into fresh, effector-like T cells (2-5). A proliferative burst of CD8⁺ T cells in peripheral blood upon initiation of ICB has been associated with initial responses to treatment of lung cancer and melanoma (6, 7). Moreover, T-cell clones in regressing tumor lesions are evident in systemic circulation (8-11). Some studies have suggested that sufficient numbers of TCF1⁺CD8⁺ T cells from melanoma lesions might predict response to ICB, but the utility of tissue-based assays remains uncertain (4, 5, 12). The potential for non-invasive, peripheral blood analyses focused on TCF1 expression in T cells to associate with response to ICB and to inform treatment decisions has not been adequately explored. Herein, we used multi-parametric flow cytometry to analyze the frequency of self-renewing TCF1⁺ cells in peripheral blood T-cell subsets of patients with locally advanced, unresectable, or metastatic NSCLC before and during treatment with PD-1 blockade.

Materials and Methods

Study design and participants

We conducted a retrospective study using banked biospecimen samples of peripheral blood mononuclear cells (PBMCs) from 23 patients and formalin-fixed paraffin-embedded (FFPE) pretreatment tumor tissue from 6 patients, as well as historical clinical data to identify characteristics of peripheral blood and tumor-associated T cells and their relationship with clinical outcomes (response and progression free survival). The research participants were all patients with locally advanced, unresectable, or metastatic NSCLC who received ICB with single-agent PD-1 blockade as part of standard of care or clinical trial participation (Table 1). All were treated at Columbia University Irving Medical Center. Randomization was not performed for this analysis. Patients were eligible for analysis if they had available banked pre- and on-treatment samples, serial radiographic assessment and met the criteria for durable clinical benefit (DCB; > 1 year of treatment or clinical response ≥ 1 year) or primary resistance (RES; < 6 months of treatment with evidence of clinical or radiographic progression). Pre-treatment blood samples were collected at the time of initiation of ICB. On-treatment blood samples were collected between 2 and 29 weeks from initiation of treatment. Tumor tissue samples were obtained at time of initial diagnosis. For analyses involving one on-treatment data point, either the only available sample or, from patients with multiple samples, the sample nearest the 3-to-6-week range, corresponding to the peak of peripheral CD8⁺ T-cell proliferation induced by PD-1 blockade (6, 7), was used. Blood from healthy donors was obtained from New York Blood Center.

Table 1.

Clinical characteristics of patients

Characteristics		All Patients (N = 23)	DCB (n = 11)	RES (n = 12)
Age, Median (Range)		65 (46–79)	68 (51–79)	65 (46–79)
Gender	Male	11 (48%)	5 (45%)	6 (50%)
	Female	12 (52%)	6 (55%)	6 (50%)
Histology	Adenocarcinoma	18 (78%)	9 (82%)	9 (75%)
	Squamous	3 (13%)	1 (9%)	2 (17%)
	Adenosquamous	1 (4.5%)	0 (0%)	1 (8%)
	Sarcomatoid	1 (4.5%)	1 (9%)	0 (0%)
PD-L1 Expression	≥ 50%	11 (48%)	5 (46%)	6 (50%)
	1–49%	4 (17%)	1 (9%)	3 (25%)
	< 1%	5 (22%)	2 (18%)	3 (25%)
	Not reported	3 (13%)	3 (27%)	0 (0%)
Prior Platinum Tx		14 (61%)	5 (45%)	9 (75%)
Prior Radiation	Non-CNS	9 (39%)	3 (27%)	6 (50%)
	CNS	2 (9%)	0 (0%)	2 (17%)
Concurrent Radiation	Non-CNS	4 (17%)	4 (36%)	0 (0%)
	CNS	3 (35%)	2 (18%)	1 (8%)
Therapy	anti–PD-1	22 (96%)	10 (91%)	12 (100%)
	anti–PD-L1	1 (4%)	1 (9%)	0 (0%)

Abbreviations: DCB, durable clinical benefit; RES, primary resistance; Tx, treatment; CNS, central nervous system.

Patient specimen collection

All participants provided informed written consent for blood and tissue collection under an approved protocol in accordance with the Institutional Review Board of Columbia University Irving Medical Center and in accordance with recognized ethical guidelines. Patients included in the study participated in one of the following biospecimen protocols: IRB-AAAO5706 and IRB-AAAL5871.

Flow cytometry

Cryopreserved PBMC samples from pre-treatment and on-treatment samples were thawed and stained according to standard flow cytometry protocol using a master mix of antibodies for surface markers including CD3 (BD, UCHT1; BUV395), CD4 (BD, SK3; BUV737), CD8 (Biolegend, RPA-T8; BV605), CD27 (Biolegend, O323; BV421), CD57 (Biolegend, HCD57; Pacific Blue), CD45RA (Biolegend, HI100; BV510), CD39 (Biolegend, A1; BV711), CX3CR1 (Biolegend, 2A9-1; BV650), PD-1 (Biolegend, EH12.2H7; AF488/PE), and CD127 (Biolegend, A019D5; PE-Cy5), and intracellular markers including Eomes (eBioscience, Dan11mag; PE or WD1928; FITC), T-bet (Biolegend, 4B10; PE-Cy7), CTLA-4 (Biolegend, L3D10; PE-Dazzle594), FOXP3 (eBioscience, PCH101; PE-Cy5.5), TCF1 (Cell Signaling, C63D9; AF647), Ki67 (Biolegend, Ki-67; AF700). Live/dead cell discrimination was performed using fixable Zombie NIR viability kit (Biolegend; APC-Cy7). Permeabilization prior to staining for intracellular markers was performed using the FOXP3 Fixation/Permeabilization Concentrate and Diluent kit (TONBO bioscience). To detect PD-1 molecules potentially coated with therapeutic anti–PD-1, anti-human IgG4 PE capable of binding pembrolizumab and nivolumab (Southern Biotec, HP6025; PE) was added to the staining panel of on-treatment blood samples. Cells were collected by Bio-Rad ZE5 and analyzed using FlowJo V10.

The primary criterion for determining the threshold of TCF1 expression was the intrinsic bimodal distribution of TCF1 staining in PD1-vs-TCF1 plots of batched samples (Supplementary Fig. S1). If bimodal TCF1 distribution was not clear from a given PD1-vs-TCF1 plot, then the threshold of TCF1 expression was secondarily resolved using the extremes of CD27^lo and CD27^hi cells, which are generally TCF1^– and TCF1⁺, respectively (Supplementary Fig. S1).

Tissue immunofluorescence assay

Quantitative multiplex immunofluorescence (qmIF) was performed on formalin-fixed paraffin-embedded (FFPE) tissue sections of pre-treatment tumor samples to measure biomolecular targets in defined tissue compartments. Validated and standardized protocols for qmIF staining and signal measurement were carried out as previously described (13-15). Primary unconjugated antibodies against CD8 (BioLegend, C8/144b Mouse IgG1) and TCF1 (Cell Signaling, C63D9 Rabbit IgG) were coincubated with tissue and then sequentially detected with horseradish peroxidase (HRP)-conjugated species-specific secondary antibodies (Goat Anti-Mouse IgG1 HRP, Abcam, ab97240; and EnVision+ Single Reagent HRP Rabbit, Agilent/Dako, K400311-2; respectively). Tyramide Signal Amplification was performed with HRP activators Cy3 Tyramide (Perkin Elmer, SAT704A001KT) for CD8, and Cy5 tyramide (Perkin Elmer, SAT705A001KT) for TCF1. Tumor epithelium was identified with a cocktail of monoclonal antibodies to pan-cytokeratins (AE1/AE3, ThermoFisher; 53-9003-82) directly conjugated to AlexaFluor 488. DAPI was used to identify cell nuclei. Coverslips were mounted with Prolong Gold (Molecular Probes), and whole slide imaging was performed on a Vectra Polaris (Akoya) using the 20X objective.

Statistical analyses

One-way Analysis of Variance (ANOVA), unpaired t test, Wilcoxon rank sum test, and correlative analyses (Spearman) were used as appropriate, as indicated in Figure legends. For longitudinal data sets with non-uniform sampling intervals from patients with single or multiple on-treatment samples, a linear mixed-effects model that adjusts for sampling time was used to determine longitudinal TCF1 frequency of DCB and RES groups. The Kaplan–Meier method and log-rank test were used in survival analysis. A significance level of 0.05 was used in all statistical tests.

Data availability

All data were generated by the authors and are available in the main text or the supplementary materials or on request from the corresponding author. Information regarding patient age, sex, histology, PD-L1 score, immunotherapy, other treatments received, response, progression-free survival are not publicly available due to patient privacy requirements but are available from the corresponding author upon reasonable request.

Results and Discussion

Patient characteristics

We analyzed PBMC samples collected from a cohort of 23 patients with NSCLC treated with ICB (Table 1), 11 of the patients had durable clinical benefit (DCB; > 1 year of treatment or clinical response ≥ 1 year) and 12 patients had primary resistance (RES; < 6 months of treatment with evidence of clinical or radiographic progression). All patients' disease was stage IV except one patient in the DCB group who was stage IIIB. Thus, all treatment was in the locally advanced, unresectable, or metastatic setting. The major histology for both groups was adenocarcinoma (78%, n = 9 DCB; n = 9 RES), followed by squamous cell carcinoma (13%, n = 1 DCB; n = 2 RES), adenosquamous (4.5%, n = 1 RES) and sarcomatoid (4.5%, n = 1 DCB). Baseline tumor PD-L1 expression of ≥ 50% was detected in 48% of patients (n = 5 DCB; n = 6 RES); 1–49% in 17% of patients (n = 1 DCB; n = 3 RES) and < 1% in 22% of patients (n = 2 DCB; n = 3 RES). No PD-L1 expression data were available for 13% of patients (n = 3 DCB). One patient received anti–PD-L1 therapy as ICB while the remainder of the patients received anti–PD-1 therapy.

Baseline TCF1 frequency in blood CCR7⁻CD8⁺ T cells associates with response to anti-PD-1

Of 22 available pre-treatment blood samples (11 DCB; 11 RES), there was no significant difference in the proportion of CD4⁺ and CD8⁺ T cells between patients with DCB those with RES (Fig. 1A; Supplementary Fig. S1A). To capture the spectrum of peripheral T-cell differentiation and renewal of blood CD8⁺ T cells, the markers CCR7 and CD45RA were used to delineate naïve (CCR7⁺CD45RA⁺), central memory (TCM; CCR7⁺CD45RA⁻), effector memory (TEM; CCR7⁻CD45RA⁻), and effector memory re-expressing CD45RA (TEMRA; CCR7⁻CD45RA⁺) subsets (Fig. 1B; Supplementary Fig. S1A). The frequencies of naive, TCM, TEM, TEMRA, and CCR7⁻ (TEM + TEMRA) CD8⁺ T-cell subsets were similar between DCB and RES patients (Fig. 1C).

Figure 1. Pre-treatment frequency of self-renewing, blood CD8⁺ T cells in patients with DCB and RES.

A, Frequency of CD3⁺ T cells that are CD8⁺ or CD4⁺ in peripheral blood of DCB and RES patients with NSCLC treated with ICB. B, Representative flow cytometry of CD45RA vs. CCR7 identifying the CCR7⁻ (TEM + TEMRA) subset of CD3⁺CD4⁻CD8⁺ T cells from a DCB and RES patient, and healthy donor. In all flow cytometry plots, numbers next to polygons or inside quadrants signify % cells in that gate. In all Figures and Supplementary Figs., CD8⁺ refers to gating of CD3⁺CD4⁻CD8⁺ T cells (Supplementary Fig 1A). C, Frequency of Naïve (CCR7⁺CD45RA⁺); TCM (CCR7⁺CD45RA⁻); TEM (CCR7⁻CD45RA⁻); TEMRA (CCR7⁻CD45RA⁺); CCR7⁻ (TEM + TEMRA) subsets within blood CD8⁺ T cells of DCB and RES patients. D, Representative flow cytometry of PD-1 vs. TCF1 in a DCB and RES patient, and healthy donor. De-identified patient label in blue or red lettering. E, Frequency distribution of TCF1⁺ cells among blood CCR7⁻CD8⁺ T cells from DCB (n = 11) and RES (n = 11) patients. Graphed values of TCF1⁺ correspond to the sum of the two values above the horizontal line of flow cytometry plots (both left- and right-upper quadrants of panel D and Supplementary Fig. S1B-S1D). F, Kaplan–Meier graph of time from ICB start vs. % of patients with PFS. Patients in upper half (blue, n = 11) vs. lower half (red, n = 11) of baseline TCF1 values (regardless of DCB/RES designation) are compared. In all figures, graphs’ mean ±SEM error bars are shown. Flow cytometry data in all panels are representative of 1–4 technical replicates of the stains indicated. Statistical significance was determined by unpaired two tailed t tests (A), Analysis of Variance (ANOVA) (C), Wilcoxon rank sum test (E), Kaplan–Meier method and log-rank test (F); *P < 0.05.

Naïve and TCM CD8⁺ T cells (both CCR7⁺) are capable of self-renewal and enriched for higher frequency and intensity of expression of TCF1, a marker of self-renewing human CD8⁺ T cells (16). The more differentiated (effector-specified), CCR7⁻ (TEM + TEMRA) subset exhibits lower frequency of TCF1 expression and more intermediate levels of TCF1 within TCF1⁺ cells (16). Because proportions of CCR7⁻CD8⁺ T cells can expand in frequency during active immune responses (16, 17), we examined TCF1 expression within the CCR7⁻CD8⁺ T-cell population (Figs. 1D, 1E; Supplementary Fig. S1A-S1D). TCF1⁺ frequency in CCR7⁻CD8⁺ T cells was significantly higher in DCB than in RES patients (Fig. 1E; 41% vs. 22%, *P = 0.013).

Lack of uniformity in the non-ICB treatments that patients received (Table 1) prompted us to stratify the analysis for prior radiation and chemotherapy, which could be confounding variables that impact blood lymphocyte subsets (18). We found that prior radiation or chemotherapy may increase baseline TCF1 frequency in RES patients, but that same trend was not evident in DCB patients (Supplementary Figs. S2A, S2B). Future studies with larger sample sizes will be needed to further address the impact of these treatment variables in a statistically rigorous manner.

Like the significant difference in TCF1 frequency in CCR7⁻CD8⁺ T cells (TEM + TEMRA) from DCB and RES (Fig. 1E; 41% vs. 22 *P = 0.013), CCR7⁺CD8⁺ T cells (Naive + TCM) exhibited a similar trend of higher TCF1 frequency in DCB, but statistical significance was not achieved in the CCR7⁺ subset (Supplementary Fig. S2C; 62% vs. 38%, P = 0.051, ns). Unlike the difference in TCF1 frequency between DCB and RES CCR7⁻CD8⁺ T cells (Fig. 1E), frequency of the T-cell memory-associated markers CD27 and CD127 (19) were not different in CCR7⁻CD8⁺ T cells of DCB and RES patients (Supplementary Fig. S2D).

Despite the clinical extremes represented by the DCB and RES phenotypes, the spectrum of baseline TCF1 frequency in CCR7⁻CD8⁺ T cells values falls along a continuum (Fig. 1E). To test whether TCF1 frequency in CCR7⁻CD8⁺ T cells correlates with progression-free survival (PFS) independently of DCB/RES status, we stratified patient samples into upper vs. lower half of all baseline values of TCF1 frequency. Patients in the upper half of TCF1 frequency in CCR7⁻CD8⁺ T cells had significantly greater PFS than those in the lower half (Fig. 1F; 17 months vs. 3 months, respectively; HR 0.39, *P = 0.019), suggesting the association of TCF1 frequency in CCR7⁻CD8⁺ T cells and clinical response was significant, as was evident in, but not dependent on, comparison of DCB vs. RES groups (Fig. 1E; *P = 0.013).

For six patients with pre-treatment tumor tissue and evaluable responses to anti–PD-1, we quantified the intensity of TCF1 expression within CD8 events. Analysis of lesional response to ICB vs. intensity of TCF1 within CD8 events revealed a non-significant trend of higher TCF1 expression with greater lesional response (Supplementary Fig. S3A-S3C). Analysis of the systemic clinical response (PFS) in relation to tissue TCF1 intensity, however, did not demonstrate a correlation (Supplementary Fig. S3D). Analysis of tissue PD-L1 scores derived from pathology records (Table 1) also showed no significant difference between DCB and RES patients (Supplementary Fig. S3E).

The foregoing findings support a model wherein greater abundance of TCF1⁺CCR7⁻CD8⁺ T cells in baseline blood associates with systemic clinical response to ICB. A limitation of the present study population is the absence of samples from a continuous clinical spectrum that would include disease progression between 6 and 12 months, reflecting the gap between the RES and DCB groups. Additionally, for the many possible covariates, including non-ICB treatment history (18), total systemic tumor mass (6), tumor mutational burden (20), and other patient characteristics, proper multivariate statistical analysis was not possible due to the small cohort size. It is, thus, anticipated that the size and spectrum of samples in future studies will need to be expanded in order to determine whether baseline (or on-treatment) TCF1 frequency in CCR7⁻CD8⁺ T cells is truly a predictive biomarker of ICB response across the spectrum of patient outcome, and to clarify the potentially confounding effects of other disease-specific covariates such that any predictive power of blood TCF1 frequency might be further refined.

On-treatment TCF1 frequency in blood CCR7⁻CD8⁺ T cells associates with response to ICB

For the 20 patients (10 DCB; 10 RES) with one or more on-treatment blood samples (Supplementary Fig. S1B-S1D), we analyzed the sample nearest the 3-to-6-week range from start of ICB, because it corresponds to the reported peak of peripheral CD8⁺ T-cell proliferation induced by PD-1 blockade (6, 7). The frequency of CCR7⁻ (TEM + TEMRA) T cells within the CD8⁺ T-cell pool was not significantly different between DCB and RES groups (Fig. 2A).

Figure 2. On-treatment frequency of self-renewing, blood CD8⁺ T cells in patients with DCB and RES.

A, Frequency of CCR7⁻ (TEM + TEMRA) T cells among blood CD8⁺ T cells from single-value, on-treatment samples of DCB (n = 10) and RES (n = 10) groups. For patients with multiple on-treatment samples (n = 8), the sample nearest 3–6 wk (range 3–11 wk) on-treatment was selected as being closest to peak ICB-induced blood CD8⁺ T-cell proliferation (Supplementary Fig. S1B-S1D). Patients with single on-treatment samples (n = 12) were also included, expanding range to 2–24 wk. Of 20 samples used, the majority were ≤ 8 wk (n = 15), with the rest ≥ 10 wk (n = 5). B, Linear mixed-effect-model for longitudinal TCF1 frequency within blood CCR7⁻CD8⁺ T cells of DCB (blue; n = 10) and RES (red n = 9) patients. For each subject, samples span pre-treatment baseline (week 0) through paired single or sequential on-treatment samples timepoints connected by a solid line. The sole closed symbol on each solid line designates which sample was used for single-value, on-treatment analyses (panels C, D). TCF1 frequency for DCB and RES groups over the sampling range were interpolated into a best-fit model of the longitudinal kinetics of TCF1 (dashed blue and red lines) and was found to be significantly different (*P = 0.019). C, TCF1 frequency distribution within blood CCR7⁻CD8⁺ T cells from single-value, on-treatment samples of DCB (n = 10) and RES (n = 10) groups. Graphed values of TCF1⁺ in panels B, C correspond to the sum of values above the horizontal line of flow cytometry plots (both upper quadrants of panel E and Supplementary Fig. S1B-S1D). D, Kaplan–Meier graph of time from ICB start vs. % of patients with PFS. Patients in upper half (blue, n = 10) vs. lower half (red, n = 10) of single on-treatment TCF1 values are compared. E, Representative flow cytometry of PD-1 vs. TCF1 from gated CCR7⁻CD8⁺ T cells of on-treatment samples in 2 DCB, and 2 RES patients. Note the right-sided expansion of PD-1⁺ cells in these on-treatment samples compared to pre-treatment pair (Fig. 1D), and across entire cohort (Supplementary Fig. S1B-S1D). F, Average frequency of PD-1⁺ cells within CCR7⁻CD8⁺ T cell subset in pre-treatment and single on-treatment samples from DCB and RES patients. G, Distribution of TCF1⁺/TCF1⁻ ratio among single on-treatment samples of PD-1⁺CCR7⁻CD8⁺ T cells from DCB (n = 10) and RES (n = 10) groups. Graphed values of TCF1⁺/TCF1⁻ ratio correspond to right-upper quadrant value divided by right-lower quadrant value of panel E and Supplementary Fig. S1B-S1D. In all graphs, mean ±SEM error bars are shown. Flow cytometry data in all panels are representative of 1–4 technical replicates of the stains indicated. Statistical significance was determined by Wilcoxon rank sum test (A, B, G), linear mixed-effects model (C), Kaplan–Meier method and log-rank test (D) ANOVA, (F); *P < 0.05, ***P < 0.001, P ≥ 0.05 ns (not significant).

There were 19 patients (10 DCB; 9 RES) with paired sets of a baseline plus one or more on-treatment samples (Supplementary Fig. S1B-S1D). To visualize the dynamics of TCF1 frequency from the onset of ICB treatment, the longitudinal data sets were analyzed with a linear mixed-effects model that can quantify differences between groups despite non-uniform sampling intervals among the subjects. (Fig. 2B). TCF1 frequency in CCR7⁻CD8⁺ T cells in the DCB and RES groups over the sampling range (up to 29 weeks) were interpolated into a best-fit model of the longitudinal kinetics of TCF1 (dashed blue and red lines). Statistical analysis indicated significantly higher TCF1 frequency in DCB than in RES groups over the timespan of on-treatment samples (*P = 0.019). The linear mixed-effects model analysis also evaluated and found no interaction of sample timing and treatment response.

We also performed on-treatment single-value analyses of the TCF1 frequency of CCR7⁻CD8⁺ T cells at the timepoints closest to the ICB-associated proliferative burst (6, 7). Akin to the differences from baseline samples, the on-treatment TCF1 frequency values maintained the trend of higher values in DCB vs. RES patients, but statistical significance was not achieved (Fig. 2C; 47% vs. 26%, P = 0.052, ns). By contrast, stratification of single on-treatment samples into upper vs. lower half of TCF1 frequency in CCR7⁻CD8⁺ T cells, regardless of DCB/RES status, revealed that patients with higher single-value on-treatment TCF1 frequency had significantly greater median PFS than those with lower TCF1 frequencies (Fig. 2D; 18.0 months vs. 3.5 months, respectively; HR 0.38, *P = 0.017). The foregoing findings suggest that greater TCF1 frequency in on-treatment blood CCR7⁻CD8⁺ T cells, like baseline blood, associates with response to ICB. Stratification of concurrent radiation therapy as a potentially confounding variable of the on-treatment TCF1 values, revealed a non-significant trend of increased TCF1 frequency in CCR7⁻CD8⁺ T cells in DCB patients receiving radiation (Supplementary Fig. S2A). The effect of concurrent radiation was unevaluable in RES patients for lack of samples.

A notable feature comparing baseline to on-treatment samples is the ICB-associated increase in frequency of PD-1⁺ cells within the CCR7⁻CD8⁺ T-cell subset (Fig. 2E, 2F; Supplementary Fig. S1B-S1D), which is consistent with the reported ICB-induced proliferation of blood CD8⁺ T cells in melanoma and NSCLC patients (6, 7). There was a trend of more PD-1⁺ T cell expansion in RES than DCB patients (Fig. 2F), but the difference was not statistically significant (Supplementary Fig. S4A). In ICB-treated patients, PD1⁺ T cells are more proliferative than PD1⁻ T cells (6, 7), but we found that DCB and RES patients had similar increases in proliferative index (Ki67 frequency) of PD1⁺CD8⁺ T cells comparing paired baseline to on-treatment samples (Supplementary Fig. S4B).

Within the PD-1⁺CCR7⁻CD8⁺ T-cell subset of on-treatment blood, the ratio of TCF1⁺ to TCF1⁻ cells was significantly greater in DCB than RES groups (Fig. 2G, ratios of 1.03 vs. 0.37, respectively, *P = 0.012). Stratification of those values into upper vs. lower half of TCF1 ratio within the PD-1⁺CCR7⁻CD8⁺ T-cell subset, regardless of DCB/RES status, revealed that patients with higher on-treatment TCF1 ratios had greater median PFS than those with lower TCF1 ratios (15 vs 4 months), but that difference was not statistically significant (Supplementary Fig. S4C; HR 0.46, P = 0.061, ns). Tumor-specific T-cell clones have been detected in circulation (9-11) and can undergo ICB-induced peripheral expansion (6, 7). We, therefore, hypothesize that some (but not all) of the CCR7⁻CD8⁺ T cells and many (but not all) PD-1⁺CCR7⁻CD8⁺ T cells being examined in blood are tumor-responsive, but this will require specificity analyses of the T cells using antigen-specific tetramer staining or single-cell T-cell receptor sequencing for confirmation. We cannot exclude the possibility that ICB-responsive patients have an intrinsic difference in the nature of their immunity, which is reflected in higher TCF1 frequency in CCR7⁻CD8⁺ T cells, but further investigation will be required to interrogate that possibility. A systemic CD4⁺ T-cell response in antitumor immunity has been proposed (8, 20), but our analysis of the self-renewal parameters of the blood CD4⁺ T-cell compartment did not reveal a significant association of TCF1 with ICB response in baseline or on-treatment samples (Supplementary Fig. S5).

The present results reveal significant associations between the response of NSCLC patients to ICB and heightened frequency of TCF1⁺ cells in baseline and on-treatment blood CCR7⁻CD8⁺ T cells, as well as for the fraction of TCF1⁺ cells in the ICB-expanded subset of PD-1⁺CCR7⁻CD8⁺ T cells in on-treatment blood. Currently approved biomarkers for ICB such as PD-L1 and tumor mutational burden require tissue, which can be influenced by a number of factors including biopsy site, handling prior to and during fixation, antibody selection, inter/intra-tumor heterogeneity, and variability of interobserver interpretation (21, 22). Our current findings and their extensions, thus, have potential to influence clinical practice: by clarifying whether insufficiency of T-cell self-renewal represents a novel class of ICB resistance; by leading to a non-invasive, predictive biomarker of ICB response and resistance; and by offering a biomarker-driven strategy of resistance intervention for patients with lower TCF1 frequency in CCR7⁻CD8⁺ T cells through brief preconditioning before and between cycles of ICB using an oral anti-anabolic medication that can expand the size of the TCF1⁺ pool and improve ICB efficacy, as was recently employed in preclinical models (23).

Synopsis.

The authors find a correlation between abundance of self-renewing, CD8+ T cells in blood and response of cancer patients to PD-1 blockade, suggesting assessment of T-cell regenerative status might serve as a non-invasive, predictive biomarker of immunotherapy response.