Circulating T-cell Immunosenescence in Patients with Advanced Non–small Cell Lung Cancer Treated with Single-agent PD-1/PD-L1 Inhibitors or Platinum-based Chemotherapy

This article is featured in Highlights of This Issue, p. 369

Immunosenescence is a global remodeling of immune functions which involves both adaptive and innate immunity and is related to the chronic antigenic stimulation occurring throughout life (1, 2). T-cell immunosenescence focuses on the phenotypic characteristics of lymphocytes and refers mainly to a low proliferative activity, whereas the functions of senescent T lymphocytes and other immune cells do not necessarily decrease (3). Different markers have been associated with low replicative potential and senescence in T lymphocytes. The loss of CD28 (4) and the gain of CD57 (5, 6) and killer-cell lectin-like receptor (KLRG1; ref. 7) in CD8⁺ T lymphocytes correlated with lower proliferation and shorter telomeres, with oligoclonal T-cell receptor (TCR) repertoire and reduced capacity to recognize antigenic diversity (8). T-cell immunosenescence has been described as a multistep process where, under persistent antigenic exposure, T cells acquired a terminal differentiation status (reexpressing CD45RA; ref. 5) and in some preclinical models also resistance to apoptosis (9). Although both senescent and exhausted T cells are characterized by low replicative potential and could share some common phenotypic features (expression of PD-1 and/or loss of CD28), they may engage different pathways to induce cell-cycle arrest (10). Furthermore, severely exhausted T cells are mainly dysfunctional (11) while senescent T lymphocytes retain their cytotoxic potential and ability to secrete high levels of cytokines and soluble factors, mainly TNFα and IFNγ (12, 13).

The expansion of low replicative, proinflammatory, and oligoclonal senescent T cells occurring upon persistent antigenic stimulation (i.e., due to aging, tumor, chemotherapy, chronic inflammation, or infections) may negatively affect treatment outcomes with ICI in patients with advanced cancer. To explore whether T-cell immunosenescence is associated with age, chemotherapy exposure or patients' clinical characteristics and to define its impact on efficacy from single-agent PD-1/PD-L1 inhibitors (ICI) in patients with advanced non–small cell lung cancer (aNSCLC), we assessed a senescent immune phenotype (SIP; % CD28⁻CD57⁺KLRG1⁺ among circulating CD8⁺ T cells) at baseline in a discovery cohort of ICI-treated patients. The SIP cut-off generated in the discovery cohort was subsequently tested in a validation cohort at baseline and during treatment with ICI. To investigate whether T-cell immunosenescence may also affect chemotherapy outcomes, we assessed SIP in a control cohort of treatment-naïve aNSCLC receiving platinum-based chemotherapy (PCT).

Data were collected from consecutive patients with aNSCLC enrolled in PREMIS (NCT03984318; ICI discovery cohort), CEC-CTC (NCT02666612; ICI validation cohort), and MSN (NCT02105168; PCT cohort) prospective studies. These studies allowed the collection of aNSCLC patients' clinical information and fresh whole blood samples for research purposes (maximum 10 samples in 3 years for each patient), after signature of informed consent. All these studies were approved by the institutional review board and the ethical committee of Gustave Roussy (Villejuif, France) and were conducted in accordance with ethical principles for medical research involving human subjects reported in the Declaration of Helsinki.

In the ICI cohorts, patients with aNSCLC were treated with single-agent ICI from July 2019 to March 2020 (discovery cohort) or from March 2017 to June 2018 (validation cohort). In the control cohort, patients with treatment-naïve aNSCLC received PCT from September 2017 to July 2018. To be eligible, patients had to be 18 years or older, with histologically or cytologically confirmed stage III or IV NSCLC and available fresh blood samples right before PCT and/or single-agent ICI.

All radiological evaluations were centrally reviewed by a senior radiologist. Tumor response was assessed by RECIST v.1.1 (14). Objective response rate (ORR) was defined as the sum of complete response (CR) and partial (PR) response, disease clinical benefit (DCB) as CR/PR and stable disease (SD) lasting at least 6 months. Atypical patterns of progressive disease (PD), such as pseudoprogression, defined as initial PD, followed by CR/PR or SD lasting more than 6 months (15) and hyperprogressive disease (HPD) defined as RECIST v1.1 PD at first CT scan during treatment and delta-tumor growth rate (TGR; ref. 16; computed as previously reported) ≥ 50% (17), were also assessed. HPD was assessed only for patients with measurable disease on two CT scans before and one during treatment.

The procedures to perform blood immune phenotyping on fresh whole blood samples, functional experiments in human peripheral blood mononuclear cells (PBMC) and flow cytometry antibodies and fluorochromes used are described in the Supplementary Methods and Supplementary Table S1, respectively. SIP was measured as percentage of CD28⁻ CD57⁺ KLRG1⁺ among CD8⁺ circulating lymphocytes. Flow cytometry gating strategy for T-cell immunosenescence is shown in Supplementary Fig. S1. Unsupervised analysis of flow cytometry data was performed using t-distributed stochastic neighbor embedding (t-SNE) algorithm with the online R software (version 3.5.0, cytofkit package). After setting the compensation matrix, CD3⁺ cells events were extracted and logicle transformation was applied. t-SNE analysis was achieved on 7940 CD3⁺ cells for each sample, using “T-cell senescence” panel markers. Supervised analysis of flow cytometry data was performed using Kaluza Flow Cytometry Software (Beckman Coulter) and was done by a single operator, blinded to the clinical patients' information. Percentages of each CD4⁺ or CD8⁺ subsets was calculated in total CD8⁺ or CD4⁺ T cells. Hierarchical clustering of flow cytometry data was performed by Morpheus software. Unsupervised cell populations analysis was performed by hierarchical clustering using euclidean distance on percentages of immune populations after data adjustment by z-score (Morpheus software, https://software.broadinstitute.org/morpheus). PBMCs cultures were established from patients with aNSCLC at baseline to perform functional experiments.

Graphs for flow cytometry data were performed by GraphPad Prism 7.0 or 8.0 (GraphPad Software).

To better stratify progression-free survival (PFS) risk an optimal cut-off for SIP level was chosen on the base of maximization of log-likelihood ratio method as proposed by Hothorn and colleagues (18) and patients were dichotomized accordingly. Associations between SIP and categorical or continuous variables were performed by logistic regression/Fisher exact test or Mann–Whitney test, respectively. For longitudinal analyses (before and after ICI), Wilcoxon matched-pairs signed-rank test was used. Overall survival (OS) and PFS curves were estimated with the Kaplan–Meier method and compared by the log-rank test. Median follow-up was estimated by inverse Kaplan–Meier method. Cox proportional hazards regression model was used to estimate HR and to perform multivariable analysis adjusting for potential cofounding effect of variables associated with outcomes from single-agent ICI. All P values were two-sided, and values less than 0.05 were considered statistically significant. Statistical analyses were performed using R (free software environment for statistical computing and graphics). Analyses were reported according to Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK) guidelines for prognostic studies (19).

Overall, SIP analysis was performed at treatment baseline in 144 consecutive patients with aNSCLC. Number of patients evaluable for responses and flow cytometry analyses in each cohort are reported in Supplementary Fig. S2.

In the single-agent ICI discovery cohort (n = 37), the median follow-up was 6.9 (95% CI, 5.4–9.2) months, ORR was 22% (8/37) and among 33 patients having a follow-up longer than 6 months DCB was 18% (6/33). Median PFS and OS were 1.7 (95% CI, 1.3–3) months and not reached (NR) (95% CI, 3.2-NR), respectively.

In the single-agent ICI validation cohort (n = 46), median follow-up was 26.6 (95% CI, 24.8–29.9) months. Among 43 patients evaluable for response, the ORR was 21% (9/43), DCB was 37% (16/43). One (2%) of 43 patients had pseudoprogression, 27 (63%) of 43 patients had measurable disease on at least 3 CT scans (2 before and one during ICI) and were eligible for TGR analysis, among them 4 (14.8%) were classified as HPD. Median PFS and OS were 5.2 (95% CI, 1.9–8) months, 12.7 (95% CI, 5–24.2) months, respectively.

The main patients' characteristics of ICI discovery and validation cohort are listed in Supplementary Table S2. Clinical characteristic did not significantly differ between cohorts with the exception of PD-L1 inhibitors administration which was less frequent in the validation compared with the discovery cohort. Median OS and PFS also did not significantly differ between discovery and validation cohorts (Supplementary Fig. S3).

Overall, in the pooled ICI cohort (n = 83), ORR and DCB occurred in 21% (17/80) and 29% (22/76) of evaluable patients, respectively. At a median follow-up of 22.1 (95% CI, 11–26.6) months, median PFS and median OS were 2.01 (95% CI, 1.7–6.2) and 12.3 (95% CI, 5.5–22.8).

In the control cohort, SIP analysis was performed in 61 consecutive patients with treatment-naïve aNSCLC eligible for PCT. The main patients' characteristics are reported in Supplementary Table S3. Median follow-up was 26.2 (95% CI, 23.9–29.1) months, ORR was 39% (24/61), and DCB was 36% (22/61). Median PFS and OS were 5.1 (95% CI, 4.2–7.8) months and 15.1 (95% CI, 7.9-NR) months, respectively.

t-SNE algorithm, performed on the first consecutive 4 patients with PD (non-DCB) to ICI and 4 patients with CR/PR or SD lasting at least 6 months (DCB) upon ICI, showed that patients who progressed had lower density of CD4⁺ and CD8⁺ T cells expressing CD28 and higher density of CD8⁺ and CD4⁺ T cells expressing CD57 and KLRG1 at baseline, compared with patients with long-lasting PR/CR/SD (Fig. 1). Therefore, the assessment of CD28, CD57, and KLRG1 expression on circulating T cells by supervised analysis of flow cytometry data (Supplementary Fig. S1) was subsequently extended to a larger set of patients to validate these preliminary findings.

In the ICI discovery cohort (n = 37), SIP median value was 23.8% (min 1.4%, max 66.5%; Supplementary Fig. S4). A total of 39.5% was the cut-off computed by log-rank maximization method (Supplementary Fig. S5A). At this value, the HR for PFS overcame the threshold of one and continued to increase with a proportional higher risk for patients to experience progression or death (Supplementary Fig. S5B). The specificity and sensitivity of this cut-off were 100% and 35%, respectively. 27% (10/37) patients had >39.5% CD28⁻CD57⁺KLRG1⁺ among CD8⁺ lymphocytes, being classified as SIP⁺. SIP⁺ patients had significantly worse median PFS [1.3 (95% CI, 0.9-NR) vs. 1.8 (95% CI, 1.4-NR) months, P = 0.03] compared with SIP⁻ patients (Supplementary Fig. S6, left). Median OS did not significantly differ between SIP⁺ and SIP⁻ patients [3.2 (95% CI, 2.2-NR) vs. NR (95% CI, 5.5-NR), P = 0.2; Supplementary Fig. S6, right], probably due to the short median follow-up [6.9 (95% CI, 5.4–9.2) months].

In the validation cohort (n = 46), SIP median value was 21.8% (min 2.3%, max 63.7%; Supplementary Fig. S4). A total of 28% (13/46) of patients were SIP⁺ according to the 39.5% cut-off previously identified in the discovery cohort. SIP⁺ patients had significantly lower ORR (0% vs. 30%, P = 0.04) and DCB (8% vs. 50%, P = 0.01) compared with SIP⁻ patients (Supplementary Fig. S7). Furthermore, 3 of 4 patients with HPD were SIP⁺, having from 47.7% to 63.7% of senescent circulating CD8⁺ T lymphocytes (Supplementary Fig. S8) and HPD was more frequent in SIP⁺ compared with SIP⁻ patients (50% vs. 4.5%, P = 0.02). SIP⁺ patients had significantly worse median PFS [1.8 (95% CI, 1.3-NR) vs. 6.4 (95% CI, 2–19) months, P = 0.009; Supplementary Fig. S6, left] and median OS [2.8 (95% CI, 2.0-NR) vs. 20.8 (95% CI, 6.0-NR) months, P = 0.02] compared with SIP⁻ patients (Supplementary Fig. S6, right).

In the ICI-pooled population (n = 83), SIP median value was 22.9 % (min 1.4%, max 66.5%; Supplementary Fig. S4). 28% (23/83) had >39.5% CD28⁻ CD57⁺ KLRG1⁺ among CD8⁺ circulating lymphocytes, being classified as SIP⁺. SIP was not significantly associated with age, PD-L1 expression on tumor cells, immune-related adverse events (irAE) rate, previous chemotherapy exposure, cytomegalovirus (CMV) antibody positivity or any other clinical characteristics (Table 1). SIP⁺ patients had significantly lower ORR (0% vs. 30%, P = 0.002) and DCB (4% vs. 40%, P = 0.002; Fig. 2A) compared with SIP⁻ patients. SIP⁺ patients had significantly worse PFS [1.7 (95% CI, 1.3–2.8) months vs. 3.8 (95% CI, 1.8–10.3), HR 2.4 (95% CI, 1.4–4.2), P < 0.0001] (Fig. 3A, left) and OS [3.1 (95% CI, 2.4–13.3) vs. 20.8 (95% CI, 8.0-NR), HR 2.3 (95% CI, 1.25–4.2), P = 0.007); Fig. 3A, right] compared with SIP⁻ patients. In a multivariate Cox regression model, SIP⁺ status remained associated with worse OS [HR 2.4 (95% CI, 1.2–5.0), P = 0.02; Supplementary Table S4A] and worse PFS [HR 2.2 (95% CI, 1.2–4.1), P = 0.01; Supplementary Table S4B] after adjustment for several factors (i.e., performance status, number of metastatic sites, lung immune prognostic index, type of ICI, line of ICI and previous chemotherapy exposure).

In the PCT cohort, SIP median value was 22.5% (min 0.8%, max 76.5%; Supplementary Fig. S4), 11% (7/61) of patients had >39.5% CD28⁻CD57⁺KLRG1⁺ among CD8⁺ lymphocytes being classified as SIP⁺. As in the ICI-pooled population, SIP did not significantly correlate with age or other clinical characteristics (Supplementary Table S3). SIP⁺ patients had similar ORR (29% vs. 41%, P = 0.69), DCB (71% vs. 31%, P = 0.09; Fig. 2B), PFS [11.7 months (95% CI, 4.8-NR) vs. 5 months (95% CI, 4.1–6.9), HR 0.39 (95% CI, 0.14–1.1), P = 0.07; Fig. 3B, left] and OS [NR (95% CI, 11.5-NR) vs. 13.9 months (95% CI, 7.1–23.7), HR 0.35 (95% CI, 0.1–1.4) P = 0.15; Fig. 3B, right] compared with SIP⁻ patients.

At the time of data analysis, polarization panel, activation/T-regulatory (Treg) panel and dynamic evolution of SIP after ICI start were performed in 40 (87%), 22 (48%), and 21 (46%) of 46 ICI-treated patients, respectively (Supplementary Fig. S2). Polarization panel allowed a phenotypic characterization of CD4⁺ T-helper (T_H) and CD8⁺ T-cytotoxic (T_C) cell subsets (T_H/C1, T_H/C2, T_H/C9, T_H/C17, T_H/C17.1, T_H/C17 double negative, T_H/C17 double positive and T_H/C22; Supplementary Table S5). Hierarchical clustering showed that high proportion of terminally differentiated (TEMRA) CD4⁺ and CD8⁺ T cells, lower naïve CD8⁺ T cells and higher T_H1 and T_C1 cells clustered with SIP⁺ (Fig. 4A). These populations were significantly associated with SIP⁺ status (Fig. 4B). Activation/Treg panel allowed a phenotypic characterization of activation markers on conventional and Treg cells. SIP⁺ patients had significantly increased activated OX40⁺ Treg cells compared with SIP⁻ (Fig. 4C) [mean 6.16%; SD (4.5) vs. 2.15%; (1.6), P = 0.01]. Variable PD-1 expression was observed among SIP⁺ CD8 T cells [mean: 39.3%; SD (24.0); Supplementary Fig. S9A] and did not significantly differ according to DCB upon ICI (Supplementary Fig. S9B).

Precomparison and postcomparison of SIP⁺ population did not show any significant variation during ICI treatment in 21 patients with available SIP analysis performed between 28 and 65 days (D) after ICI start (Supplementary Fig. S10A). Among these 21 patients, 9 did not progress while 12 had progression as best response to ICI. No statistically significant variation of SIP⁺ status, assessed as % SIP⁺ population at D_28/65 minus % SIP⁺ population at D₀, was found in nonprogressing compared with progressing patients. No statistically significant variations of immunosenescence markers, (CD28 absence and/or CD57 expression and/or KLRG1 expression) on CD4 and CD8⁺ T cells and no difference in the variation of circulating TEMRA, naïve or T_C1 CD8⁺ T cells were observed neither in the overall population with available pre and postimmunophenotype analyses nor according to DCB upon ICI (Supplementary Fig. S10B–S10D).

We next investigated the proliferation capacity and the functional properties of SIP⁺ CD8⁺ T cells, assessing the expression of Ki67 and of IL2, TNFα, IFNγ, respectively, after PBMCs in vitro activation. PBMCs in vitro cultures were established from 22 patients with treatment-naïve aNSCLC. Cytokine production was measured in 22 patients, while only 13 patients could be assessed for proliferation. Ki67 expression was significantly lower in CD28⁻ and in CD28⁻CD57⁺KLRG1⁺ CD8⁺ T cells compared with CD28⁺ CD8⁺ T cells under all the stimulation conditions (Fig. 5A). CD28⁻ CD57⁺ CD8⁺ T cells produced significantly higher levels of IFNγ and lower level of IL2 compared with CD28⁺ CD8⁺ T cells, TNFα expression was also higher in CD28⁻ CD57⁺ compared with CD28⁺ CD8⁺ T cells, although not statistically significant (Fig. 5B).

T-cell immunosenescence, defined by the loss of CD28 and expression of CD57 and KLRG1 on peripheral CD8⁺ T cells, was observed at baseline in 28% and 11% of patients with aNSCLC treated with ICI or PCT, respectively. The 39.5% cut-off was initially generated in a discovery cohort of patients with aNSCLC and subsequently tested in a larger validation cohort with longer follow-up. The relatively high threshold and specificity suggest that the 39.5% cut-off is able to detect primary immunotherapy resistance and patients with aNSCLC with high probability to progress to single-agent ICI.

T-cell immunosenescence did not correlate with age, previous chemotherapy exposure or irAEs, while it was associated with worse outcomes only in the ICI cohort. The efficacy and safety of single-agent ICI among elderly patients with aNSCLC is a debated topic. Although the age cut-offs used in clinical trials were different, one metanalysis (20) and some studies including patients older than 75 years (21–23) suggested an absence of a significant survival benefit of ICI in this subgroup. Furthermore, age ≥65 years was associated with HPD in patients with cancer treated with ICI (24). On the other hand, a pooled analysis of three trials comparing pembrolizumab with chemotherapy in patients with NSCLC (25), a phase II study of nivolumab including 34% of patients older than 70 years (26) and real-world data (27–30) showed a survival improvement with single-agent ICI in the elderly population. The reasons for this discrepancy are unknown; however, they might potentially be related to the heterogeneity of the studies, the different age cut-offs used and to the fact that elderly patients are largely unrepresented in clinical trials and results come mainly from subgroup analyses. Alpert and colleagues (31) recently demonstrated that immune and chronologic ages do not overlap and that an IMM-AGE score, based on the variation of multiple genes over an older adult lifetime, exhaustively describes a person's immune status and efficiently predicts all causes related mortality. As for IMM-AGE score, immunosenescence does not simply reflect patients' age and may be a more reliable biomarker compared with chronologic age to assess efficacy from ICI.

Increased senescent CD28⁻CD57⁺ (32) or highly differentiated CD8⁺CD28⁻ T cells (33) were reported, respectively, in patients with lung cancer compared with healthy volunteers and in patients with NSCLC receiving PCT compared with treatment-naïve subjects, suggesting that T-cell immunosenescence may be associated with cancer diagnosis or chemotherapy exposure. Contrary to patients in PCT cohort who were chemotherapy naïve, in the ICI-pooled population only 14% had not been exposed to chemotherapy. Furthermore, although the median percentage of CD28⁻CD57⁺KLRG1⁺ CD8⁺ T cells did not differ between ICI-pooled population and PCT cohort (≃23%), SIP⁺ patients were fewer in PCT cohort compared with ICI-pooled population (11% vs. 28%), suggesting a possible role of previous chemotherapy treatment in increasing circulating T-cell senescence.

Similarly, lack of CD28 and expression of CD57 correlated with chronic viral infections (i.e., CMV) in humans (34). In this study, no significant association between T-cell immunosenescence and positive CMV status was found; however, only 37 (45%) of 83 ICI-treated patients had anti-CMV antibodies assessment and definitive conclusions cannot be drawn. In addition, we did not include a control cohort of healthy volunteers to assess a potential role of cancer itself in inducing T-cell immunosenescence.

SIP⁺ status is significantly associated with higher TEMRA CD8⁺ and CD4⁺ cells, T_H1, T_C1 lymphocytes, OX40⁺ Treg cells and with reduced naïve and effector memory CD8⁺ lymphocytes. Although we included CD45RA in our T-cell immunosenescence flow cytometry panel, we did not directly assess the coexpression of such terminal differentiation marker on senescent T lymphocytes. Interestingly, an overlap between TEMRA and senescent phenotypes (35) and low CD28 expression among circulating TEMRA CD8⁺ T cells in patients with NSCLC before ICI start (36) have been reported. The association between T-cell immunosenescence and terminal differentiation along with the reduction of circulating naïve CD8⁺ lymphocytes in senescent patients suggest that the ability of the immune system to react against tumor antigenic diversity and emerging neoantigens might be impaired in T-cell immunosenescence.

Furthermore, the increase in circulating T_H1, T_C1 lymphocytes and OX40⁺ Treg may reflect a chronic systemic proinflammatory state. In fact, both CD28⁻CD57⁺ T cells (12) and TEMRA CD8⁺ cells (13) have been described as functionally activated and able to produce high levels of proinflammatory cytokines (i.e., TNFα and IFNγ), and extracellular matrix remodeling proteases (35). In line with these studies, we found that senescent T cells produced more TNFα and IFNγ compared with the nonsenescent counterparts. Sustained increased levels TNFα and IFNγ have been associated with smoldering inflammation (37), tumor development, and acquisition of cancer stemness and aggressive features (38). Altogether these factors may promote resistance to anticancer treatments. Of note, blood inflammatory parameters such as derived neutrophils/lymphocytes ratio and high lactate dehydrogenase have recently been associated with lack of benefit to ICI (39–41), probably due to the suppression of effective T-cell antitumor responses (42). In this study, no significant correlation between T-cell immunosenescence and these blood parameters was found, therefore how the chronic proinflammatory status associated with T-cell senescence affects ICI activity remains unclear.

We demonstrated that circulating senescent T cells produced less IL2 compared with nonsenescent CD8⁺ lymphocytes. Low level of IL2 preferentially promotes Treg homeostasis due to the high constitutive IL2R-α expression by these cells (43). In addition, lowering IL2 levels impairs CD8⁺ T-effectors development (44), tempering anticancer T-cell response upon ICI. A crosstalk between T-cell immunosenescence and Treg lymphocytes has been recently described. Human Treg cells were shown to suppress effector T cells and initiate DNA damage response triggered by glucose competition, resulting in senescence and functional changes in T cells (45). In this regard, we could have expected a lower incidence of irAEs in SIP⁺ patients due to the higher proportion of OX40⁺ Treg cells. However, the rate of any grade irAEs did not differ according to SIP status. These data would need further exploration due to the small sample size of patients experiencing irAEs in this study (N = 15, 18%) or having an assessment of circulating Treg cells before irAEs occurrence (N = 4, 9%).

To our knowledge, this study is the first to demonstrate that T-cell immunosenescence is significantly associated with lack of response, low DCB, HPD, and poor survival upon ICI in patients with aNSCLC.

Two recent manuscripts have reported an association between circulating highly differentiated CD28^low/negative CD27^negative CD4⁺ T cells (46) or CD62^low CD4⁺ T cells (47) and clinical responses in aNSCLC treated with ICI. In particular, in Zuazo and colleagues highly differentiated CD28^low/negative CD27^negative CD4⁺ T cells were mainly central memory (CD45RA⁻ CD62L⁺) or effector memory (CD45RA⁻ CD62L⁻) T cells, not anergic, not exhausted and not senescent which expressed high level of Ki67 and expanded upon PD-1 blockade (46). Similarly, in Kagamu and colleagues CD62L^low CD4⁺ T cells belonged to a primed Th₁ subpopulation and expressed aurora kinase A, a gene involved in G₂–M mitotic phase (47). Of note, highly differentiated CD4⁺ T cells were able to recover CD8⁺ systemic immunity with expansion of CD28⁺ CD8⁺ subsets upon PD-1 blockade (46). Although we did not assess these circulating CD4⁺ populations in relation to T-cell senescence, it is likely that they were lower in SIP⁺ patients. However, the real implication of these cells in the response to ICI was not definitively proven due to the absence of a control cohort of patients not treated with ICI.

In this study, T-cell senescence did not significantly correlate with response or survival to PCT. Although results in the PCT cohort may be influenced by the low number of SIP⁺ patients, the numerically longer PFS and the higher disease control observed in SIP⁺ patients further suggest a truly differential effect of T-cell senescence according to treatment type. Therefore, immune aging could specifically affect functions of T lymphocytes which are necessary for an effective anticancer immune response upon ICI. It is likely that the low proliferative potential of senescent T cells, their reduced capacity of recognizing antigenic diversity and their ability to promote immune tolerance or protumoral chronic systemic inflammation may all play a role in explaining the lack of efficacy from ICI in SIP⁺ patients.

Interestingly, Hui and colleagues (48) reported that CD28 is the primary target for PD-1–mediated inhibition and is strongly preferred over TCR for dephosphorylation by PD-1–recruited phosphatases. This finding suggests that abundance of CD28⁻ T cells may correlate with absence of efficacy from anti-PD-1/PD-L1 therapy. In vitro experiments showed that blocking PD-1 had no effect on proliferation or functionality of circulating human TEMRA (CD45RA⁺ CD27⁻) T cells, while both were increased by blocking senescence (13), supporting the hypothesis that senescence and exhaustion may involve distinct pathways and that reversing senescence together with PD-1 blockade may be a potential therapeutic strategy. In this study, the mean PD-1 expression on SIP⁺ CD8 T cells was approximately 40%, suggesting a relatively limited overlap between exhaustion and senescence.

Despite its prospective design and the use of a control cohort of chemotherapy-treated patients, this study has some limitations such as the small number of patients tested for the dynamic evolution of SIP or for the association between T-cell senescence and specific circulating immune populations, the absence of available tissue samples to assess senescent markers on tumor-infiltrating lymphocytes and the lack of mechanistic insights about how T-cell immunosenescence may negatively impact ICI outcome. Furthermore, SIP cut-off generation method suffers of low sensitivity, therefore a subgroup of SIP⁻ patients may still experience primary progression to ICI, as observed in 1 of 4 patients with HPD who was classified as SIP⁻, having 28.03% of circulating CD28⁻ CD57⁺ KLRG1⁺ CD8⁺ T cells. Although SIP status did not significantly change according to PD-L1 expression, it was not possible to evaluate the correlation between T-cell senescence and other known biomarkers of response (i.e., tumor mutational burden; ref. 49) or progression (i.e., LKB1 and KEAP mutations; ref. 50) to ICI. Finally, this study included mainly pretreated patients with NSCLC who received single agent ICI in second or further lines. Considering that chemoimmunotherapy and double immune checkpoint blockade have emerged as effective first-line treatment options in patients with NSCLC (51), the role of T-cell senescence in the context of immunotherapy combinations need to be determined.

A better characterization of the proliferation, the secretory phenotype (inflamm-aging; ref. 52), and the activated intracellular signaling pathways (signal-aging; ref. 53) of both circulating and tumor-infiltrating senescent T cells is currently ongoing and could further validate immunosenescence as a negative predictive biomarker for ICI, providing also novel immunological mechanisms associated with immunotherapy resistance.

In conclusion, circulating T-cell immunosenescence is observed in 28% and 11% of aNSCLC treated with ICI or PCT, respectively, is not associated with clinical characteristics and correlates with poor outcome upon ICI but not PCT, being a novel negative predictive biomarker in pretreated patients with aNSCLC. T-cell senescence was associated with specific circulating immune phenotypes and it was characterized by low CD8⁺ T cells proliferation and increased functional activation. Additional studies are needed to characterize the biological mechanisms of resistance to ICI involving T-cell immunosenescence.

R. Ferrara reports personal fees from MSD (advisory board) outside the submitted work. E. Auclin reports other from Mundipharma (travel expenses) and personal fees from Sanofi Genzymes (lecture) outside the submitted work. B. Duchemann reports personal fees and nonfinancial support from Roche (expert testimony, travel accommodation) and nonfinancial support from Pfizer (travel accommodation) and AstraZeneca (travel accommodation) outside the submitted work. L. Mezquita reports grants, personal fees, and other from Boehringer (research funding), Amgen (research funding), BMS (research funding), Roche Diagnostics (consultant and advisory role), Bristol-Myers Squibb (lectures and educational activities), Tecnofarma (lectures and educational activities), Roche (lectures and educational activities), AstraZeneca (lectures and educational activities), and Chugai (travel, accommodations, expenses), Roche (travel, accommodations, expenses); in addition, L. Mezquita reports mentorship program with key opinion leaders (funded by AstraZeneca) outside the submitted work. C. Caramella reports personal fees from Astra Zeneca, BMS, MSD, and Pfizer outside the submitted work. L.E.L. Hendriks reports grants and other from Roche-Genentech [grants for IIS (institution), fees for adboards (institution), fees for interview sessions (institution), travel reimbursement (self), local PI Roche study], AstraZeneca [grants for IIS (institution), mentorship program with key opinion leaders funded by AstraZeneca, local PI AZ study], and Boehringer Ingelheim [grants for IIS (institution), fees for adboards (institution)]; other from MSD [fees for adboards (institution), speaker (institution)], BMS [fees for adboards (institution), travel reimbursement (self), local PI BMS study]; Eli Lilly [fees for adboards (institution)], and Pfizer [fees for adboards (institution)]; personal fees from Quadia (fees for webinars); other from Novartis (local PI Novartis study), GSK (local PI GSK study), Merck (local PI Merck study), Takeda [fees for adboards (institution), local PI Takeda study], and Blueprint Medicines (local PI Blueprint medicines study) outside the submitted work. D. Planchard reports personal fees from AstraZeneca (consulting, advisory role or lectures), Bristol-Myers Squibb (consulting, advisory role or lectures), Boehringer Ingelheim (consulting, advisory role or lectures), Celgene (consulting, advisory role or lectures), Daiichi Sankyo (consulting, advisory role or lectures), Merck (consulting, advisory role or lectures), Novartis (consulting, advisory role or lectures), Pfizer (consulting, advisory role or lectures), Roche (consulting, advisory role or lectures), Samsung (consulting, advisory role or lectures), prIME Oncology (consulting, advisory role or lectures), and Peer CME (consulting, advisory role or lectures) outside the submitted work; and reports clinical trials research as principal or coinvestigator (institutional financial interests) from AstraZeneca, Bristol-Myers Squibb, Abbvie, Boehringer Ingelheim, Eli Lilly, Merck, Novartis, Pfizer, Roche, Medimmun, Sanofi-Aventis, Taiho Pharma, Novocure, and Daiichi Sankyo. J. Remon reports nonfinancial support from OSE Immunotherapeutics (travel); personal fees from Boeringher ingelheim (advisory), MSD (advisory), and Pfizer (speaker); personal fees and other from Astrazeneca (advisory), BMS (advisory/travel), and Roche (advisory/travel) outside the submitted work. C. Proto reports personal fees from BMS and Roche (advisory board) and other from BMS (travel accommodation), MSD (travel accommodation), and Roche (travel accommodation) outside the submitted work. M.C. Garassino reports grants and personal fees from Eli Lilly (PI, MISP in Thimic malignancies; speaker, advisory board), Otsuka Pharma (local PI, enrollment in clinical trials in NSCLC; speaker; advisory board), AstraZeneca (PI, enrollment and steering committee in clinical trials in NSCLC; consulting, advisory boards, lectures; steering committee), Novartis (PI, enrollment in clinical trials in NSCLC; advisory board), BMS (PI, enrollment in clinical trials in NSCLC; speaker, advisory board), Roche (PI, enrollment in clinical trials in NSCLC; speaker, advisory board), Pfizer (PI, MISP in Thimic malignancies; advisory board), Celgene (PI, enrollment in clinical trials in NSCLC; speaker, advisory board), Incyte (institutional grants; advisory board; speaker), Inivata (advisory board), Bayer (PI, enrollment in clinical trials in mesothelioma; advisory board), MSD (PI, enrollment in clinical trials in NSCLC; consulting, advisory boards, lectures; steering committee), GlaxoSmithKline S.p.A. (local PI, enrollment and steering committee in clinical trials in NSCLC; advisory board), Spectrum Pharmaceuticals (PI, enrollment in clinical trials; advisory board; steering committee), and Blueprint Medicine (PI, enrollment in clinical trials; advisory board; steering committee); personal fees from Takeda (speaker, advisory board; lectures), Boehringer Ingelheim (advisory board), Sanofi-Aventis (advisory board), Daiichi Sankyo (advisory board), and Janssen (advisory board); nonfinancial support from MSD (principal investigator Keynote 189;MISP pembrolizumab in low expressors PD-L1(>50%)), Pfizer (MISP sunitinib in thymic malignancies), and Eli-Lilly (MISP ramucirumab plus carbo-taxol in thymic malignancies); grants from Tiziana Sciences (PI, enrollment in clinical trials Thimic malignancies), Clovis (PI, enrollment in clinical trials in NSCLC), Merck Serono (PI, enrollment in clinical trials in NSCLC), UNITED THERAPEUTICS CORPORATION (institutional grant), Merck KGaA (institutional grant), TURNING POINT (P.I. TRIDENT-1 enrollment in clinical trials), IPSEN (P.I. MM-398-01-03-04 RESILIENT enrollment in clinical trials), MedImmune (PI, enrollment in clinical trials), and EXELISIS (PI, enrollment in clinical trials) outside the submitted work. J.-C. Soria reports personal fees from AstraZeneca, Abbvie, Bayer, Blend Therapeutics, Boehringer Ingelheim, Cytomix, Daiichi Sankyo, Eli Lilly, Genmab, Guardant Health, Inivata, Merck, Netcancer, Pharmamar, Roche, Servier, and Tarveda outside the submitted work; and was a full-time employee at AstraZeneca from Sep 2017 to Dec 2019. A. Marabelle reports grants from Fondation Malakoff Médéric and grants and personal fees from Sanofi during the conduct of the study; grants, personal fees, and nonfinancial support from BMS; personal fees and nonfinancial support from MSD; grants from Fondation MSD Avenir; personal fees from Roche/Genentech, Astra Zeneca, Servier, Merck Serono/Pfizer, and GSK outside the submitted work. B. Besse reports grants from Abbvie, Amgen, AstraZeneca, BeiGene, Blueprint Medicines, BMS, Boehringer Ingelheim, Celgene, Cristal Therapeutics, Daiichi-Sankyo, Eli Lilly, GSK, Ignyta, IPSEN, Inivata, Janssen, Merck KGaA, MSD, Nektar, Onxeo, OSE immunotherapeutics, Pfizer, Pharma Mar, Roche-Genentech, Sanofi, Servier, Spectrum Pharmaceuticals, Takeda, Tiziana Pharma, and Tolero Pharmaceuticals during the conduct of the study. N. Chaput reports grants from BMS Fondation during the conduct of the study and grants from SANOFI, GSK, ROCHE, and BMS Fondation, grants and personal fees from Astrazeneca, and grants and other from Cytune Pharma (BSA) outside the submitted work. No disclosures were reported by the other authors.

R. Ferrara: Conceptualization, data curation, formal analysis, funding acquisition, validation, investigation, visualization, writing-original draft, writing-review and editing. M. Naigeon: Conceptualization, data curation, formal analysis, supervision, validation, investigation, visualization, writing-original draft, writing-review and editing. E. Auclin: Software, formal analysis, methodology. B. Duchemann: Investigation. L. Cassard: Conceptualization, data curation, formal analysis, supervision, methodology. J.M. Jouniaux: Investigation. L. Boselli: Investigation. J. Grivel: Investigation. A. Desnoyer: Investigation. L. Mezquita: Data curation. M. Texier: Software, formal analysis, methodology. C. Caramella: Investigation. L. Hendriks: Data curation. D. Planchard: Visualization. J. Remon: Investigation. S. Sangaletti: Writing-review and editing. C. Proto: Writing-review and editing. M. Garassino: Writing-review and editing. J.-C. Soria: Visualization. A. Marabelle: Visualization. A.-L. Voisin: Data curation. S. Farhane: Data curation. B. Besse: Conceptualization, supervision, funding acquisition, visualization, writing-review and editing. N. Chaput: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, writing-original draft, writing-review and editing.

This work was supported by a grant from “Fondation Bristol-Myers Squibb pour la recherche en Immunoncologie” and by “SIRIC SOCRATE 2.0 INCa-DGOS-Inserm_12551.”

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.