Circulating CD28⁻KLRG1⁺CD8⁺ T cells involve in systemic and local immunity that predicts chemoimmunotherapy outcomes in advanced NSCLC

Abstract

Background

Chemoimmunotherapy has become the standard first-line treatment for advanced non-small cell lung cancer (NSCLC). Deciphering the T-cell subset responsible for chemoimmunotherapy and easily tested conveniently is critical in predicting the treatment outcomes.

Methods

Based on peripheral blood collected from patients enrolled from a phase 2 clinical study (ClinicalTrials.gov NCT04836728), we performed multi-color flow cytometry and unsupervised analysis to explore correlations with therapeutic outcomes. We integrated single-cell RNA and T-cell receptor (TCR) sequencing in 36 samples, including peripheral blood, tumors and non-tumor tissues, from 8 NSCLC patients to interpret the correlation, which was further verified using blood samples, orthotopic and subcutaneous lung cancer mouse model.

Results

The baseline CD28⁻KLRG1⁺CD57⁺ and on-treatment CD28⁻KLRG1⁺ CD8⁺ T cells in peripheral blood were independent factors which indicated improved treatment outcomes in advanced NSCLC patients receiving first-line chemoimmunotherapy. While being in a late-differentiated T-cell status, these cells were clonally expanded and reinvigorated during chemoimmunotherapy, serving as a peripheral T-cell pool for supplying potential tumor-reactive T cells in tumors, and reversely differentiating into less-differentiated subsets. The zinc-metallothionein pathway regulated the CD28⁻KLRG1⁺ CD8⁺ T-cell subset. Zinc supplementation combined with chemoimmunotherapy improved both local and systemic antitumor immune responses in mouse model.

Conclusions

Circulating CD28⁻KLRG1⁺ CD8⁺ T cells are valuable and convenient biomarkers for first-line chemoimmunotherapy in advanced NSCLC and provide insight into how late-differentiated or senescent T cells engage in the antitumor immunity when immunotherapy is added to conventional therapies.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12967-026-08484-5.

Introduction

Immune checkpoint inhibitors, such as PD-1/PD-L1 inhibitors, have greatly changed the treatment strategies for non-small cell lung cancer (NSCLC). Chemoimmunotherapy (CIT) has become the first-line standard of care for advanced NSCLC [1, 2]. However, only 49.2%–57.9% of the advanced NSCLC patients benefit from the combination treatment [1, 2]. Therefore, predictive biomarkers are urgently required. Cytotoxic CD8⁺ T cells played a crucial role in antitumor immune response. Tumor-infiltrating CD8⁺ T cells (TILs) always exhibited an exhausted status [3]. Reinvigoration of exhausted TILs by PD-1/PD-L1 inhibitors was associated with better survival in metastatic melanoma, non-small cell lung cancer and renal cell cancer [4, 5]. However, accurate assessment of TILs before treatment or at advanced stages is limited by biopsy tissues. Furthermore, T-cell clonal replacement during treatment implied that TILs detected at a fixed time point could not represent the dynamic alteration [6, 7]. Consequently, convenient sample collection is also important for exploration and application of biomarkers.

Several studies have demonstrated the association of T-cell biomarkers in peripheral blood with treatment outcomes of immunotherapy [8]. Peripheral blood, as a pool of systemic immune systems, harmonizes responsiveness to multiple immune inducers, which might comprehensively reflect the long-term clinical efficacy [9]. The previous study showed that high circulating CD28⁻KLRG1⁺CD57⁺ among CD8⁺ T cells correlated with lack of benefit from single-agent PD-1/PD-L1 inhibitors in patients with advanced NSCLC after multiple-line treatments [10]. The CD28⁻KLRG1⁺CD57⁺ phenotype is identified as late-differentiated T cells [11], and has been studied in age-related senescence [12]. In contrast, in patients with advanced NSCLC treated with second-line nivolumab (PD-1 antibody), circulating terminally differentiated CD8⁺ T cells were associated with an improved response to PD-1/PD-L1 blockade [13]. NSCLC patients responding to atezolizumab (PD-L1 antibody) had tumor neoantigen-specific T cells enriched in the circulating late-differentiated CD8⁺ T cells [14]. Although the KLRG1⁺ CD8⁺ T cells exhibited limited capacity to differentiate into other circulating memory population [15], they protected nonlymphoid tissues from a second pathogen attack by supplying effector T (Teff) cells and tissue-resident memory T cells (Trm) followed [16]. However, the associations of circulating differentiated CD8⁺ T cells with clinical efficacy and TILs has not been reported for first-line CIT in NSCLC.

Zinc (Zn) is a vital trace metal necessary for all forms of life. Both deficiency and hyperzincemia have been linked to various diseases, including infection, cancer and neurodegeneration [17]. In lung cancer patients, serum Zn levels are significantly lower than that in controls [18, 19]. Deficiency in Zn may contribute to the pathogenesis and progression of lung cancer [20], but the molecular mechanism is unclear. Zn induced immunogenic cell death [21, 22], activated dendritic cells (DCs), degraded various collagen components of tumors by upregulating the activity of matrix metalloproteinase [23], and alleviated the immunosuppression of macrophages [24], thus enhancing the tumor infiltration of CD8⁺ T cells. Combination of Zn biomaterials with anti-PD-1 antibody synergistically increased the anti-tumor effect in colorectal cancer therapy [25]. However, the direct regulation of CD8⁺ T cells by Zn remains poorly understood.

Metallothioneins (MTs) are small, cysteine-rich proteins that participate in metal ions homeostasis, detoxification and scavenging of free radicals [26]. MTs are involved in overall Zn signaling. When cells experience stress and oxidants, MTs release Zn and influence various cellular structures, including chromatin, by regulating transcription or acting as an antioxidant [27]. The role of MT expression in lung cancer prognosis has been debated [28, 29]. MT expression is downregulated in lung cancer compared to peri-tumor tissues [28, 30], which is related to methylation [29]. High MT expression has been associated with T-cell dysfunction in TILs [31] as well as in adipose-resident T cells [32]. Peripheral blood lacks the intense metabolic and inflammatory conditions found in the tumor microenvironment (TME). MTs might have a distinct impact on T cells in the peripheral blood.

In the present study, we demonstrated the predictive potential of circulating CD28⁻KLRG1⁺ CD8⁺ T-cell subset in the advanced NSCLC patients receiving first-line CIT, based on a multicenter, randomized phase 2 clinical study (ClinicalTrials.gov NCT04836728, Registration date: April 6, 2021) of the CIT control group. T-cell receptor (TCR) tracking revealed this subset proliferated responsively and provided a circulating T-cell pool under CIT. Meanwhile, we proved that the Zn-MT signaling pathway was involved in regulation of this subset, as revealed by single-cell transcriptomic analysis. Zn supplementation in combination with CIT improved the antitumor effect. Our results revealed that the CD28⁻KLRG1⁺ CD8⁺ T-cell subset was a predictive candidate for first-line CIT in advanced NSCLC, providing new insights into the role of late-differentiated T cells in antitumor immunity.

Methods

Study design

This study was designed to investigate the role of circulating CD28⁻KLRG1⁺ CD8⁺ T cells in predicting first-line CIT outcomes in advanced NSCLC and the associated mechanisms. A total of 98 individuals were included in the study, including 72 patients with NSCLC, 5 patients with other tumors, and 21 healthy donors (Figure S1A). Of the 72 patients with NSCLC, a cohort with 30 patients were from in a multicenter, randomized, open-label phase 2 study evaluating the effects of a PD-1 inhibitor (Sintilimab developed in China) and chemotherapy with or without autologous cytokine-induced killer cell immunotherapy in the first-line treatment of advanced NSCLC (ClinicalTrials.gov NCT04836728, Registration date: April 6, 2021). In our study, only patients enrolled in the CIT group were included. We used flow cytometry analyses to determine the association of circulating CD28⁻KLRG1⁺ CD8⁺ T cells with the treatment outcomes, and characterize the differentiated statuses, phenotypes, functionality, and senescence markers. We combined scRNA-seq with scTCR-seq analyses to illustrate the TCR repertoire, dynamic alterations and compartmental association of circulating CD28⁻KLRG1⁺ CD8⁺ T cells. As single-cell/TCR sequencing is technically challenging in advanced NSCLC because tumor tissue is rarely available from stage IV patients, making neoadjuvant samples the most feasible source for mechanistic exploration. We performed scRNA-seq analysis and in vitro experiments to investigate Zn-MT pathway activated to regulate CD28⁻KLRG1⁺ CD8⁺ T cells. Samples collected from the patients or health donors were summarized in Figure S1A. Figure S1B shows the correspondence of CD8⁺ T-cell populations across flow cytometry, scRNA-seq, and TCR analysis. We used ex vivo model and orthotopic and subcutaneous lung cancer transplantation mouse models to evaluate the effect of Zn supplementation on CIT treatment. The flow cytometry experiment and data analyses were performed in a blinded manner. The design of the scRNA-seq analyses, TCR analyses and mouse treatment were not blinded. Tumor-burdened mice were randomly assigned to the control and treatment groups. All the available data were included in the analyses. One patient included in the phase 2 study refused treatment after one cycle of CIT treatment was treated as the censoring in the survival analysis. This patient was not available in response evaluation. One patient suffered from a sudden death after finishing three cycles of CIT treatment. These two patients were excluded from treatment response analysis. The n values represent the number of biologically independent subjects/samples (patients or healthy donors and mice), with each replicate originating from distinct individuals, unless otherwise indicated. The n values and statistical tests for the analyses are included in the legends of each figure.

Patient selection and peripheral blood collection for biomarker assessment

Peripheral blood samples for biomarker exploration in this study were collected from 30 patients enrolled in a multicenter, randomized, open-label phase 2 study (ClinicalTrials.gov NCT04836728, Registration date: April 6, 2021) (Figure S1C). All the 30 patients were diagnosed with stage IV NSCLC, were treatment-naïve, lacked EGFR/ALK/ROS1 gene mutations, and had an Eastern Cooperative Oncology Group (ECOG) performance status score of 0–1. The inclusion and exclusion criteria were shown in Supplementary Methods. And these patients were all from the PD-1 inhibitor combined with chemotherapy group (control group). CIT was concurrently performed in the first 4–6 cycles, and then PD-1 inhibitor alone for squamous NSCLC or CIT maintenance for non-squamous NSCLC (Supplementary Methods). The baseline (before the first treatment) and on-treatment (during 3–6 cycles) peripheral blood samples were prospectively collected in EDTA anticoagulant tubes (Table S1). Peripheral blood mononuclear cells (PBMCs) were isolated from whole blood within 4 h of collection by density gradient centrifugation using Ficoll-Paque PLUS (DongFang HuaHui, China, Cat# 25610). PBMCs were stored in cryopreservation medium in liquid nitrogen for further exploration. This study was approved by the Institutional Review Board and Ethics Committee of Tianjin Medical University Cancer Institute & Hospital (E20210014) and conformed to the Declaration of Helsinki and the Good Clinical Practice guidelines. All the patients provided written informed consent for participation.

Flow cytometry detection and data analysis

The detailed procedures to assess the immunophenotype and function of PBMCs using flow cytometry method are described in the Supplementary Methods. FlowJo software (version 10.8.1 package) was used for data analysis. The flow cytometric gating strategies for immunophenotyping were shown in Figure S2. Unsupervised analysis was conducted using the t-SNE algorithm in the FlowJo software (Supplementary Methods). The main subsets we focused on were the CD28⁺CD8⁺ T cells, CD28⁻CD8⁺ T cells, and the CD28⁻CD8⁺ subsets with positive or negative expression of KLRG1 and CD57. To identify characteristics of these subsets, the expression levels of T-cell differentiated markers (naïve T cell: CD45RA⁺CCR7⁺, central memory T [Tcm] cell: CD45RA⁻CCR7⁺, effect memory T [Tem] cell: CD45RA⁻CCR7⁻, and effector memory-expressing CD45RA [TEMRA] cell: CD45RA⁺CCR7⁻), checkpoint molecules (PD-1, TIGIT, LAG3, and Tim-3), T-cell activation and proliferation markers (CD69, CD134, Ki67), T-cell function markers (IL-2, IFN-γ, ΤΝF-α), stem-like and exhaustion markers (TCF1 and TOX) were assessed. In addition, we evaluated senescence of the subsets using the combined senescence markers, including β-galactosidase activity, which was evaluated by SA-β-Gal staining; cell cycle arrested proteins, p53 and p21; DNA damage repair, p-ATM); and activated MAPK signaling pathways, p-ERK, p-p38 and p-JNK. Antibody information is presented in Table S2.

Clinical sample preparation for single-cell RNA and TCR sequencing

The inclusion of patients with stage IA–IIIB NSCLC in the scRNA- and scTCR-seq analyses was reported in our previous study [33]. To minimize the impact of gene mutations on the immune response in circulation as well as within the TME, we limited this study to 8 patients diagnosed with squamous NSCLC that lacked EGFR and ALK gene mutations (Table S3). The neoadjuvant chemoimmunotherapy (NCIT) regimen consisted of pembrolizumab combined with paclitaxel for two cycles both before and after surgical resection (Supplementary Methods). Thirty-six samples were finally included in our analyses, including 11 peripheral PBMCs (with four paired pre- and post- neoadjuvant treatment samples), eight tumor tissues, eight adjacent non-cancerous lung tissues, five distance lung tissues, and four regional draining lymph nodes (Table S4). Tissue samples were firstly digested into single-cell suspensions, followed by magnetic-activated cell sorting (MACS) to isolate of CD45⁺ immune cells (Supplementary Methods). This study was approved by the Institutional Review Board and Ethics Committee of Tianjin Medical University Cancer Institute & Hospital (BC2020060) and conformed to the Declaration of Helsinki and the Good Clinical Practice guidelines. Informed consent was obtained from all patients.

Library construction and data analysis for scRNA sequencing

After removing dead cells using the Dead Cell Removal Kit (Miltenyi Biotec, USA, Cat# 130-090-101), 7000–12,000 viable cells, ensuring ≥ 80% cell viability, were used to construct the scRNA-seq library on the 10x Genomic platform. The scRNA-seq libraries were prepared using Chromium Single Cell 5′ Library & Gel Bead Kit v1.0 (10x Genomics, USA, Cat# 1000006) and i7 Multiplex Kit (10x Genomics, USA, Cat# 120262) (details in Supplementary Methods).

TCR repertoire and shared clonotype analyses

The scTCR-seq data were processed to identify the productive TCR sequence (Supplementary Methods). Each unique productive TCR sequence was assigned a clone identifier (clonotype), enabling the quantification of individual TCR clones within the samples. To track the dynamic and spatial alterations of T-cell clones, we analyzed TCR repertoires in the peripheral blood. Key metrics such as diversity, unique TCR frequency, and relative TCR abundance were used to characterize the TCR repertoire. TCR diversity was assessed using the Chao1 index, Shannon-Wiener (Shannon) index, Inverse Simpson (Inv. Simpson) index, and clonality index, each reflecting distinct aspects of the TCR repertoire diversity (Supplementary Methods).

Shared TCR clonotypes between tissues were defined as TCRs with the same complementarity-determining region 3 (CDR3) nucleotide sequences. To visualize shared clonotypes across different CD8⁺ T-cell subsets and tissue types, we used the R package VennDiagram version 1.7.3 to generate Venn diagrams.

Potentially tumor-reactive (pTRT) T cells were defined, according to the previous study with slight modification [34], as tumor-infiltrated CD8⁺ T cells with clonal size ≥ 2. In addition, CD8⁺ T-cell clones detected in nontumor tissues that shared clonotypes with tumor-infiltrated expanded clones were also classified ad pTRTs.

Zn supplementation in vitro

1 ⊆ 10⁶/mL PBMCs or purified 5 ⊆ 10⁵/mL CD8⁺ T cells cultured in lymphocyte culture medium (LONZA X-VIVO 15 Cell Medium supplemented with IL-2 at the final concentration of 100 IU/mL) were stimulated by the TCR signaling stimuli, soluble anti-CD3 antibody OKT3 (0.01 µg/mL) and anti-CD28 antibody (0.1 µg/mL) for 24 h. Then, ZnSO4 solution (25 µM) was added in the culture medium. After culturing for an additional 72 h (short-term activation), or 17 days (one cycle of activation, expansion and contraction of T cells), cells were collected for the estimation of gene transcription (Table S5), protein expression, subcellular location and chromatin accessibility. Peripheral blood was collected from 8 cancer patients (Table S6).

Lung cancer model evaluates effect of Zn supplementation combined with chemoimmunotherapy

LLC-OVA cells (RRID: CVCL_4358), (2 ⊆ 10⁶ or 4 ⊆ 10⁶ cells each) were orthotopically or subcutaneously incubated into 4–5 weeks old male C57BL/6 mice. Tumor-burdened mice were randomly grouped into four groups: (1) control group treated with saline solution, (2) Zn group treated with ZnSO4 solution at a dose of 12.38 mg/kg (equivalent to Zn 5 mg/kg), (3) CIT group treated with cisplatin (1 mg/kg) + anti-PD-1 antibody (2 mg/kg, Biocell, USA, Cat# BE0273), and (4) Zn plus CIT group. Tumor volume was calculated as the formula: 1/2 ⊆ ab2, where a is the longest diameter of the tumor and b is the perpendicular diameter to length. In vivo imaging was performed on the orthotopic tumors using the IVIS Spectrum (PerkinEler, USA), where the luciferase substrate luciferin, at a dose of 150 µg/g, was intraperitoneally injected into anesthetized mice, and bioluminescence was captured within 10–15 min post injection to quantify tumor growth. At the end of the treatment, immune cell profiles in the TME, peripheral blood, and spleen were evaluated using flow cytometry. The cytokine production and secretion were quantified using flow cytometry and an ELISA assay, respectively (details in Supplementary Methods). This study was approved by the Laboratory Animal Ethics Committee of Tianjin Medical University Cancer Institutes & Hospital (AE-2022040), and conformed to the international guidelines for the humane treatment of animals. Animal experiments were conducted following the Basel Declaration and institutional guidelines, with adherence to the International Council for Laboratory Animal Science (ICLAS) principles to ensure ethical standards. The information of antibody and subpopulation phenotype for flow cytometry assessment are shown in Table S7 and S8, respectively.

Other methods, including splenocyte functional assessment, anti-tumor CD8⁺ T-cell cytotoxicity assessment, real-time PCR, western blot, immunofluorescence staining, intracellular Zn level measurement, and ATAC-seq to assess chromatin accessibility, are described in detail in the Supplementary Methods.

Statistical analyses

IBM SPSS statistics (version 29.0.1.0) and R (version 4.3.2) were used for statistical analyses. Significance was determined using unpaired, Mann-Whitney U test, Independent-samples t-test, or paired, Wilcoxon signed-rank test, paired t-test for comparison of dependent groups; using Kruskal-Wallis test for nonparametric comparison and one-way or one-way repeated ANOVA test for parametric comparison. Post-hoc pairwise comparisons were adjusted using the Benjamini-Hochberg method to control the false discovery rate. Two-way ANOVA test followed by Mann-Whitney U test was used in treatment response analysis; the two factors being analyzed were group (response vs. non-response) and time (baseline vs. treatment). Two-way repeated measures ANOVA was used in comparisons of tumor size in mouse model, with the experiment factors, treatment groups as the between-subjects factors and time points of tumor measurement as the within-subject factors, adjusted by the least significance difference (LSD) method for multiple measurements over time between groups. Durable clinical benefit (DCB) was defined as complete response/partial response or disease stable > 6 months. Non-DCB was defined as disease stable ≤ 6 months/progressive disease. Major pathological response (MPR) was defined as residual viable tumor cells ≤ 10% of total tumor tissue after treatment. OS was defined as the time from inclusion to death from any cause or the last date of follow-up date. PFS was defined as the time from inclusion until the date of objective disease progression or death from any cause in the absence of progression. The Kaplan-Meier method was used to estimate OS and PFS. Differences in survival rates were compared using the log-rank tests. The best cutoff for the Kaplan–Meier survival analysis was determined by iteratively testing all possible thresholds of the variables using the R package survminer. Specifically, surv_cutpoint function was applied, which implements maximally selected rank statistics to identify the threshold that yield the most significant separation of survival curves. Each potential cutoff was evaluated by log-rank tests, and the values producing the maximal between-group difference was selected under strict significance criteria (p < 0.05). No alternative cutoff methods were applied. All variables were processed using the same standardized procedure to ensure methodological and to minimize the risk of overfitting. P-values were determined using a two-sided α = 0.05. The data cutoff date for all analyses was August 30, 2024. The software and algorithms used are listed in Table S9.

Role of funding source

The funding sources had no role in study design, data collection, data analyses, interpretation, or writing of this manuscript. All authors have full access to the data in the study and accept responsibility to submit for publication.

Results

Circulating CD28⁻KLRG1⁺ CD8⁺ T cells are associated with survival in patients with advanced NSCLC receiving first-line CIT

A total of 43 peripheral blood samples (13 pairs prior to and during CIT) of 30 advanced NSCLC patients (Table S1) enrolled in a phase 2 clinical study were available to explore the correlation between circulating CD8⁺ T-cell subsets and treatment outcomes (Fig. 1A; Figure S1). All patients were treatment-naïve, diagnosed with stage IV advanced NSCLC, and received first-line CIT treatment (Table S10). The flow cytometric gating strategies were shown in Fig. 1B and Figure S2. Unsupervised analysis of the flow cytometry data was firstly performed using t-distributed stochastic neighbor embedding (t-SNE) algorithm (Fig. 1C) to screen CD8⁺ T-cell subsets associated with the treatment response. The t-SNE algorithm, performed on four DCB and four non-DCB patients, showed that DCB patients had a higher density of CD28⁻KLRG1⁺ CD8⁺ T-cell subset, especially the CD28⁻KLRG1⁺CD57⁺ CD8⁺ T-cell subset at baseline, compared with non-DCB patients (Fig. 1D). Therefore, we assessed the KLRG1, CD57, and CD28 expression on circulating CD8⁺ T cells in all 30 patients.

Fig. 1.

Circulating CD28⁻KLRG1⁺ CD8⁺ T cells predict treatment outcomes in patients with advanced NSCLC. (A) Experimental strategy for quantification, isolation, and characterization of PBMCs for flow cytometry detection. (B) Flow cytometric gating strategy for CD28⁻CD8⁺ T cells. (C) t-SNE plots overlaid with differentiated status of 40,000 CD8⁺ T cells from 8 patients (4 DCB vs. 4 non-DCB) at baseline. (D) t-SNE plots of CD8⁺ T cells from each response group (left) or overlaid with the expression of selected markers (right) (red: higher density, blue: lower density). (E) Frequencies of CD28⁻CD8⁺ T-cell subsets with positive or negative KLRG1 and CD57 expression in peripheral blood at baseline and during chemoimmunotherapy (Baseline n = 21; Treatment n = 22). Data are presented as median with interquartile. Wilcoxon signed-rank test. (F) Frequencies of CD28⁻KLRG1⁺CD57⁺ CD8⁺ T cells in each response group (Baseline n = 19; Treatment n = 22). Data are presented as median with interquartile. Two-way ANOVA test. P < 0.05 only for main effect of group (DCB vs. non-DCB), F = 9.342, P = 0.004. Main effect of time, F = 0.184, P = 0.670. Interaction effect between group and time, F = 0.046, P = 0.832. Mann-Whitney U test followed for DCB vs. non-DCB in each time point. (G–H) Baseline (G) and on-treatment (H) CD57⁺ and CD57⁻ subsets within CD28⁻KLRG1⁺ CD8⁺ T-cell population associated with overall survival. Cut-off: CD57⁺ subset, 14.0% (baseline) and 16.2% (on-treatment); CD57⁻ subset, 28.6% (baseline) and 10.4% (on-treatment). (I) On-treatment CD28⁻KLRG1⁺ CD8⁺ T-cell population positively associated with overall survival. Cut-off, 25.7%. (J–K) On-treatment CD57⁺ and CD57⁻ subsets within CD28⁻KLRG1⁺ CD8⁺ T-cell population (J) and the entire CD28⁻KLRG1⁻ population (K) negatively correlated with overall survival. Cut-off, 9.8%, 8.6%, and 20.4%, respectively. Log-rank test (G-J). (Baseline n = 21; Treatment n = 22). DCB, Durable clinical benefit

The median proportion of CD28⁻/CD8⁺ T cells in the peripheral blood was 48.3% (95% confidence interval [CI]: 41.3–57.3%) and 48.9% (95% CI: 43.2–54.7%) at baseline and during CIT, respectively (Figure S3A). The median proportion of KLRG1⁺CD57⁺, KLRG1⁺CD57⁻, KLRG1⁻CD57⁺ and KLRG1⁻CD57⁻ subset in the CD28⁻CD8⁺ T cells was 20.6% (95% CI: 18.3–34.8%), 37.6% (95% CI: 20.8–58.1%), 5.3% (95% CI:1.0–10.1%) and 16.5% (95% CI: 6.5–20.2%) at baseline, and 34.0% (95% CI: 28.9–41.2%), 30.9% (95% CI: 25.2–44.2%), 5.2% (95% CI: 2.6–14.7%) and 16.0% (95% CI: 8.8–22.9%) during CIT (Fig. 1E). Comparing the paired blood samples, we found that the proportion of CD28⁻/CD8⁺ cells was not altered, but the proportions of KLRG1⁺CD57⁺/CD28⁻CD8⁺ cells were increased under CIT (Fig. 1E; Figure S3). Total 28 patients were available for DCB evaluation (Table S1). Consistent with the preliminary finding (Fig. 1D), the proportions of CD28⁻KLRG1⁺CD57⁺/CD8⁺ T cells were higher in DCB patients than those in non-DCB patients both at baseline and during CIT (Fig. 1F). In survival analysis, we did not find any association between the patient survival and CD8⁺ T cells, CD4⁺ T cells and CD28⁻CD8⁺ T cells. At baseline, the CD57⁺ and CD57⁻ subsets within baseline CD28⁻KLRG1⁺ CD8⁺ T-cell population showed opposite association with OS; however, both subsets predicted better OS during CIT (Fig. 1G and H). Consistently, a higher number of on-treatment CD28⁻KLRG1⁺ CD8⁺ T cells correlated with improved OS (Fig. 1I). In contract, both the CD28⁻KLRG1⁻ CD8⁺ T-cell population and its CD57⁺ and CD57⁻ subsets were related to worse OS during CIT (Fig. 1J and K; Table S11). Similar associations were observed in the PFS analysis (Table S12). The univariate and subsequent multivariate analyses indicated that the circulating CD28⁻KLRG1⁺CD57⁺ CD8⁺ T-cell subset at baseline, and CD28⁻KLRG1⁺ CD8⁺ T-cell subset during treatment, were independent factors predicting better OS in NSCLC patients receiving first-line CIT (Table S13 and S14). Additionally, we found that the on-treatment KLRG1⁻CD57⁻ CD8⁺ naïve subset had a positive association with the treatment response (DCB vs. non-DCB patients) and OS, respectively (Figure S4). This naïve subset deserves further study in future. Collectively, our findings indicate that the circulating CD28⁻KLRG1⁺CD57⁺ CD8⁺ T-cell subset at baseline and CD28⁻KLRG1⁺ CD8⁺ T-cell subset during treatment have predictive potential for patients with advanced NSCLC who received first-line CIT.

Late-differentiated CD28⁻KLRG1⁺ CD8⁺ T cells are active and functional

Next, we investigated the phenotype and functional characteristics of the five CD8⁺ T-cell subsets in peripheral blood, including the four CD28⁻CD8⁺ T-cell subsets with positive or negative KLGR1 and CD57 expression, and the CD28⁺CD8⁺ T-cell subset, using flow cytometry. The four CD28⁻CD8⁺ T-cell subsets, except for CD28⁻KLGR1⁻CD57⁻ subset, had the highest percentages of TEMRA cells, followed by Tem cells, naïve cells, and Tcm cells. The proportions of TEMRA cells were decreased in the CD28⁻KLGR1⁻CD57⁻ CD8⁺ T-cell subset, but still higher than Tem cells. The CD28⁺CD8⁺ T cells displayed a different differentiated composition compared with the CD28⁻CD8⁺ T-cell subsets, showing an increase in less-differentiated naïve, Tcm and Tem cells, with Tem cells being the largest component (Fig. 2A; Figure S5). The CD28⁻KLRG1⁺ CD8⁺ T-cell subsets exhibited moderate expression of PD-1 and CD69, which was lower than that in the CD28⁺CD8⁺ subset but higher than that in the CD28⁻KLRG1⁻ CD8⁺ T subsets (Fig. 2B). Upon TCR signaling activated by anti-CD3 and anti-CD28 antibodies, the CD28⁻KLRG1⁺ CD8⁺ T-cell subsets presented elevated proliferation (Ki67) and production of the antitumor cytotoxic cytokines (IFN-γ, IL-2 and TNF-α), compared with the CD28⁻KLRG1⁻ CD8⁺ T-cell subsets (Fig. 2C, D). Dynamically analyzed, most of the phenotype and functional markers remained relatively stable levels during CIT compared with those at baseline, in all five CD8⁺ T-cell subsets (Figure S6). The proportions of Tem cells were enlarged in the CD28⁻KLRG1⁺CD57⁺ subset (Fig. 2A). CD134 and CD69 expression were decreased in the CD28⁻KLRG1⁺ subsets under CIT (Figure S6A), but they were still higher than those in the CD28⁻KLRG1⁻ subsets (Fig. 2B). As CD69 is an important marker for Trm cells in addition to early activated T cells [35], these activated T cells might migrate into tissues when both the systemic and local antitumor immune responses are induced by CIT. Collectively, these results indicate that although being the late-differentiated status, the CD28⁻KLRG1⁺ CD8⁺T-cell subsets remain active and functional.

Fig. 2.

Circulating CD28⁻KLRG1⁺ CD8⁺ T cells remain active and functional. (A) Composition of differentiated stages. (B) PD-1 expression of and activation (CD69 and CD134). (C–D) Proliferation (Ki67) and cytokine production (D) of subsets in response to TCR stimulation. Overnight-rested T cells were stimulated by OKT3 and anti-CD28 for 3 h (cytokine detection) and 4 days (Ki67 detection). (A–D), NSCLC patients from the phase 2 clinical trial (NCT04836728). Baseline n = 21, Treatment n = 22 for all markers except for Ki67 (Baseline n = 17; Treatment n = 14) because of limited peripheral PBMCs available in several patients. (E) TCF1 and TOX expression. (F) Levels of senescence-related markers. (E–F), 15 treatment-naïve NSCLC patients. Percentages on Y-axis represent the proportion of labeling in each CD8⁺ T-cell subset. Data are presented as median with interquartile. One-way repeated ANOVA followed by pairwise comparisons using estimated marginal means (emmeans), with p-values adjusted using Benjamini-Hochberg method. MFI, median fluorescence intensity

We also evaluated the expression of TCF1, TOX, and other checkpoint molecules (TIGIT, LAG3 and Tim-3) in the CD8⁺ T-cell subsets in 15 treatment-naïve advanced NSCLCL patients (Table S15). The CD28⁻KLRG1⁺ CD8⁺ subsets had higher levels of TCF1 and TOX, as well as the checkpoint molecules, among the CD28⁻CD8⁺ subsets (Fig. 2E; Figure S7A). The concurrent increase in TCF1 and TOX indicated that the CD28⁻KLRG1⁺ CD8⁺ subsets maintained a progenitor exhausted T-cell (Tpex) state. Since the loss of CD28 expression and acquisition of KLRG1 and CD57 expression are associated with T-cell senescence [36], we identified senescence features in these subsets using combined senescence markers in the 15 NSCLC patients. While the senescence markers were expressed higher in the CD28⁻KLRG1⁺CD57⁺ subset than the other CD28⁻CD8⁺ subsets, they were comparable to or significantly lower than those expressed in the CD28⁺CD8⁺ subset, except for β-galactosidase activity (Fig. 2F). Lastly, to investigate whether the distinct expression of above molecules among CD8⁺ T-cell subsets was associated with the tumor-burden condition, we evaluated several markers that were closely associated with T-cell status and functionality, including PD-1, TIGIT, TCF-1, TOX and β-galactosidase activity, in the peripheral blood of 11 non-elderly healthy donors (Table S16), and found the similar expression patterns among the five CD8⁺ T-cell subsets as the findings in the NSCLC patients, except decreased PD-1 expression on the CD28⁺CD8⁺ subset (Figure S7B; Table S17). We did not test LAG3 and Tim-3 in the healthy donors, because the above results from NSCLC patients and our previous findings from health donors both indicated minimum expression of LAG3 and Tim-3 in circulating CD8⁺ T cells. In summary, our results demonstrate the heterogenous features of the circulating CD28⁻CD8⁺ T-cell subsets; the CD28⁻KLRG1⁺ CD8⁺ T cells remain the capacity for activation and functionality, as well as a less exhausted status, implying that these cells might be responsible for PD-1/PD-L1 inhibitors.

scRNA-seq analysis profile transcription of circulating CD28⁻KLRG1⁺ CD8⁺ T cells

To investigate the transcriptional regulation of the circulating CD28⁻KLRG1⁺ CD8⁺ T cells and their relationship with TILs while minimizing the confounding effects of gene mutations, we selectively extracted and integrated the scRNA-seq and scTCR-seq data from eight patients with squamous NSCLC from the dataset that we previously published [33] (Table S3). Thirty-six samples, including peripheral blood and various tissue samples, were finally analyzed in the present study (Fig. 3A; Table S4).

Fig. 3.

Single-cell transcriptome profiles circulating CD28⁻KLRG1⁺ CD8⁺ T cells. (A) Experimental diagram. NSCLC patients were either treated with neoadjuvant chemoimmunotherapy or chemotherapy, or were treatment-naïve, and all patients subsequently underwent surgery. P0, before neoadjuvant treatment. P1, surgery. (B) tSNE visualization of five major immune cell types in PBMCs. (C) tSNE visualization of circulating CD8⁺ T cells. (D) Cluster identification of circulating CD8⁺ T cells. (E) Expression of canonical markers across clusters. (F–G) Expression of CD28, KLRG1, and B3GAT1 (CD57) in circulating CD8⁺ T cells. (H) Volcano plot showing differentially expressed genes (DEGs) between Teff cluster and the other clusters in circulating CD8⁺ T cells. DEGs were nominated by requiring at least 2 times fold change and a Bonferroni-adjusted p-value less than 0.05. (I) GO and KEGG enrichment analyses of DEGs. Q value, Benjamini-Hochberg for multiple comparisons to control the false discovery rate

A total of 56,185 cells of 11 peripheral blood mononuclear cell (PBMC) samples, with 4 paired before and after NCIT, passed the quality control and were clustered into 5 major immune cell types, along with T cells, natural killer (NK) cells, myeloid cells, mast cells and B cells (Fig. 3B). 14,069 circulating CD8⁺ T cells were selected based on the clustering analysis of total cell population, re-clustered into 14 clusters (Fig. 3C; Figure S8A–C), and annotated as six subpopulations, including the naïve, Tcm, Tem, Teff, Trm, and mucosal-associated invariant T (MAIT) cells, according to their specific expression of classic markers (Fig. 3D, E). The expression of CD28, KLRG1, and B3GAT1 (CD57 gene) for each CD8⁺ T cells was shown in Fig. 3F. The average KLRG1expression was high in the clusters 3, 4, 8, 11 and 12, and middle expression in clusters 0, 2, and 7. High CD28 expression was observed in the clusters 1, 3, 9 and 12 (Fig. 3F, G). B3GAT1 was slightly expressed in all CD8⁺ T clusters (Fig. 3G), probably owing to the short lifetime of B3GAT1 mRNA in T cells as reported previously [37]. Therefore, CD28⁻KLRG1⁺ CD8⁺ T cells were distributed in the Teff clusters, especially concentrated in the clusters 4, 8 and 11. As our above results revealed that the survival benefit and potential antitumor efficacy of CD28⁻CD8⁺ T cells more depended on the positive expression of KLRG1, but not CD57, we therefore addressed the Teff cells in the following analysis. Compared with the other CD8⁺ T-cell clusters, the Teff clusters highly expressed MT family members (MT1E, MT2A), FGFBP2, ITGB1, GZMH, PRF1, KLRG1 and KLF6 genes (Fig. 3H; Figure S8D). Subsequent Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses based on the differentially expressed genes (DEGs) revealed that the upregulated signaling pathways were enriched in processes related to cellular structure, immune function, and cytoskeletal regulation (Fig. 3I) and downregulated signaling pathways were enriched in ribosome related process (Figure S8E). Similar results were found when clusters 4, 8 and 11 were compared with the other Teff clusters or all the other clusters (Figure S8F-H). These results from differentially expressed genes (DEGs) indicated that the CD28⁻KLRG1⁺ distributed Teff clusters experienced alterations under NCIT, presenting improved immune activation.

Next, we performed trajectory analysis to explore the potential transition from naïve to memory and Teff in the circulating CD8⁺ T cells, which displayed heterogeneous and orchestrated differentiation patterns (Fig. 4A). Specifically, we identified two bifurcation events emerging from a common progenitor state (Figure S9). Teff cells mainly distributed in state 1 and 4, which arose independently from branching points 1 and 2 (Fig. 4A). Compared with the pretreatment samples (P0), the proportion of state 1 cells increased after treatment (P1, at surgery), while the proportions of state 5, 3 and 4 decreased (Fig. 4B). To further characterize these states within Teff, Tcm, and Tem populations and their association with treatment outcomes, we found that a higher proportion of state 1 cells in the Teff population were observed in patients with improved survival. Similarly, higher proportions of state 1 and state 4 cells in Tcm population, were associated with better survival, whereas Tem cells from patient with worse outcomes were enriched for state 3 (Fig. 4C). We then analyzed the DEGs among the five states. MTs (MT1E, MT1X and MT2A) were upregulated at the late stage of state 1, and from the mid- to late stages of states 4 and 3. In state 4, their expression paralleled that of the cytotoxic genes GZMB and GZMH (Fig. 4D, E).

Fig. 4.

Trajectory analysis identifies key gene expressions during differentiation of circulating CD8⁺ T cells. (A-B) Monocle plots demonstrating the transition of circulating CD8⁺ T cells based on CD8⁺ T-cell types, sample origins, pseudotime (A), monocle clusters, and sampling time (B). (C) Cell compositions of the five states within Teff, Tem and Tcm populations based on patient survival time. Short, n = 2, PFS ≤ 10 months. Long, n = 5, PFS > 10 months. (D, E) Heatmap showing the divergent cell fates along with the pseudotime curve, focusing on branching point 1 (D) and 2 (E). (F) Trajectory distribution of CD8⁺ T-cell clones with different clonal size. (G) State composition across clones with different sizes. (H) Proportion of clonal size within each trajectory state. (I) Trajectory distribution of the two largest T-cell size (clone 1# and 2#)

Lastly, we examined the clonal architecture within the trajectory. The productive TCR clones were identified in 74.7% (20612/27585) T cells, including 67.4% (9481/14069) CD8⁺ and 82.4% (11131/13516) CD4⁺ T cells. Based on clonal size (number of identical clones), CD8⁺ TCR clones were classified into five groups: single clone, double clone, 2 < clone ≤ 5, 5 < clone ≤ 10, and > 10. These clonal categories were present across all trajectory states (Fig. 4F). Teff state 1 cells contributed the largest number of > 10-cell clones among all states (Fig. 4G). And when comparing within each state, Teff state 1 and state 4 cells showed the elevated proportion of large clones (> 10) (Fig. 4H). To further assess clonal dynamics along the trajectory, we traced two largest clones (clone 1# and 2#) detected at both P0 and P1 time points in patient 1# and 2#, respectively (Fig. 4I). At baseline, both the clones were preferentially enriched in state 4. After treatment, clone 1# showed a broader distribution across multiple states, and clone 2# remained mainly located in state 1 and 4 (Fig. 4I). Taken together, the CD28⁻KLRG1⁺ CD8⁺ Teff cells display an enrichment of large TCR clones alongside elevated expression of Zn-MT-associated genes, features that may be relevant to their involvement in the response to NCIT treatment.

Circulating CD28⁻KLRG1⁺ CD8⁺ T cells experience clonal expansion during CIT

Subsequently, we explored TCR clonal alternation of the circulating CD28⁻KLRG1⁺ CD8⁺ T cells. Three CD8⁺ T-cell subsets, CD28⁺CD8⁺ T cells (count 947), CD28⁻KLRG1⁻ CD8⁺ T cells (count 7010) and CD28⁻KLRG1⁺ CD8⁺ T cells (count 4999) were used for TCR repertoire analysis. CD28⁻KLRG1⁻ CD8⁺ T cells had the highest unique TCR clonotypes, as shown by the richness index (Fig. 5A), partially because of the relatively higher numbers of the subsets compared to the other subsets. To describe the TCR diversity more comprehensively, we applied several diversity indices that focused on different aspects of TCR diversity. The Chao 1 index estimates the total number of unique TCR clonotypes (richness), meanwhile considering the singleton and doubleton [38]. The Clonality index correlates closely with highly prevalent clonotypes [39]. The Inverse Simpson (Inv.Simpson) index emphasizes dominant clonotypes. The CD28⁺ and CD28⁻KLRG1⁺ subsets exhibited a substantial reduction in the Chao 1 index compared to the CD28⁻KLRG1⁻ subset (Fig. 5B), suggesting that the first two subsets had fewer single clones than the third subset. The CD28⁻KLRG1⁺ subset displayed the highest Clonality index and lowest Inv.Simpson index (Fig. 5C, D), implying significant clonal expansion in the subset. Accordingly, the frequency of unique TCR was decreased (Fig. 5E), and large clones (frequency: 0.01–0.1) were greatly increased in the CD28⁻KLRG1⁺ subset (Fig. 5F). These results demonstrated that the CD28⁻KLRG1⁺ CD8⁺ T-cell subset possessed a superior clonal amplification capacity compared to the other subsets.

Fig. 5.

TCR profiles characterize circulating CD8⁺ T-cell subsets. (A-D) Diversity indices of TCR repertoire (P0 = 4, P1 = 7). Kruskal-Wallis test followed by Wilcoxon signed-rank test. (E) Percentages of unique TCR clonotypes. Data are presented as median with interquartile. (F) Proportion of TCR clonotypes with different amplifications. (G) Venn diagram depicting the intersections and disjunctive unions of TCR clonotypes between P0 (baseline) and P1 (surgery). Shared clones were classified into expanded, contracted, and the same clones based on the clonal size alteration of each clonotype under neoadjuvant therapy. P1 vs. P0: >, expanded; <, contracted; =, same. (H) Composition of different clonal categories defined as (G) in each CD8⁺ subset. (I) Cluster distribution of specific clonal categories in tSNE visualization of total circulating CD8⁺ T cells (left) and relative abundance in each cluster (right). (J) Upset plot showing conversion pattern between three CD8⁺ T-cell subsets at P0 and P1. Top vertical bar represents the number of clonotypes at each unique intersection. In the medium matrix of dots, each row represents a subset and each column represents a unique intersection. Left plot shows the number of clonotypes in six cell subsets, respectively. Solid triangles, circles, and squares: one subset differentiated into either of the other two subsets or both. Representative image of patient 5

Next, we examined whether the expansion of CD28⁻KLRG1⁺ CD8⁺ T cells might be related to NCIT by dynamically tracking all TCR clones within CD8⁺ T-cell subsets across four paired PBMC samples (Fig. 5G, H). Comparing the clonal size of each specific TCR clonotype between baseline (P0) and surgery (P1), we categorized the P0 and P1 TCR clones into four types, respectively (Fig. 5G). P0 clones were classified as disappeared, expanded, contracted, and the same clones. The P1 clones were classified as emerged, expanded, contracted, and the same clones (Fig. 5G). The CD28⁻KLRG1⁺ CD8⁺ subset had the highest proportions of expanded clones compared to the other two CD8⁺ subsets in all patients. Meanwhile, in the CD28⁻KLRG1⁺ CD8⁺ subset, the expanded clones were the largest component among the four clonotype categories in 3/4 of the patients, both at P0 and P1 (Fig. 5H). However, both the CD28⁺CD8⁺ and CD28⁻KLRG1⁻ CD8⁺ subsets had higher proportions of the disappeared and newly emerged clones at P0 and P1, respectively, compared with the CD28⁻KLRG1⁺ CD8⁺ subset. The disappeared and newly emerged clones constituted the main components of the P0 and P1 CD28⁺CD8⁺ subsets, respectively (Fig. 5H). The contracted and the same clones presented low abundance in all the three subsets both at P0 and P1. Further mapping of these P0 and P1 clones into the circulating CD8⁺ T-cell clusters (Fig. 3B, C), we found that the P0 expanded clones were mostly distributed in the Teff clusters (cluster 0, 2, 4, 8 and 11). The P1 expanded clones were largely in the Teff clusters (cluster 0, 2, 4 and 8). The majority of P0 disappeared clones were in the naïve (cluster 6), Tcm (cluster 5) and Tem (cluster 1) clusters. And mass P1 new emerged clones were in naïve, Tcm, Tem and Teff (cluster 2) clusters (Fig. 5I; Figure S10A). These results clearly show that the three CD8⁺ T-cell subsets experience divergent clonal alterations under NCIT, and that the CD28⁻KLRG1⁺ subset is virtually revigorated by NCIT.

Finally, to investigate the pattern of conversion between the three CD8⁺ T-cell subsets across the two time points (P0 and P1), we analyzed the shared TCR clones among the six combinations formed by these subsets at the two time points (Fig. 5J). A specific TCR clone shared between ≥ 2 subsets indicated that this clone underwent interconversion between these subsets. We found that interconversion occurred between the subsets not only at the same time point (P0 or P1), but also at different time points (P0 and P1) (Fig. 5J; Figure S10B). Moreover, to substantiate the conversion direction of clones between the subsets, we identified the TCR clones that were shared between one subset at P0 and another or the other two subsets at P1. We found that each subset could serve as a derivation, capable of converting into one or more other subsets (Fig. 5J; Figure S10B). These findings suggested that the late-differentiated CD28⁻KLRG1⁺ and CD28⁻KLRG1⁻ subsets could reversely differentiated into the less-differentiated CD28⁺ subset, indicating the highly dynamics and plasticity of these subsets.

Circulating CD28⁻KLRG1⁺ CD8⁺ T cells sharing clones with TILs confer improved survival

Based on the newly emerged TCR clonotypes in circulating CD8⁺ T cells post-NCIT by matching P0 and P1 clonotypes detected in the circulating CD8⁺ T cells (Fig. 5G), we examined their shared TCR clones across multiple tissues, aiming to investigate the communication between circulating CD28⁻KLRG1⁺ CD8⁺ T cells and CD8⁺ TILs in TME. After clustering the total cells from 25 tissue samples, including tumor tissues, adjacent non-cancerous lung tissues, distance lung tissues, and draining lymph nodes (Figure S11A), we selected the CD8⁺ T-cell cluster and re-clustered the cells into 6 main T-cell subpopulations (Fig. 6A; Figure S11B). We evaluated the expression levels of CD28, KLRG1, and CD57 across multiple clusters of tissue-derived CD8⁺ T cells, and observed that the CD28⁻KLRG1⁺ CD8⁺ T-cell subset was distributed in cluster 1 and 6, which were classified as Teff cells (Figure S11C, D). These findings suggest that the CD28⁻KLRG1⁺ CD8⁺ T-cell subsets from both circulating blood and tissues shared a late T-cell differentiated status (Teff). Then, we isolated the newly emerged TCR clonotypes in circulating CD8⁺ T cells at P1 (Fig. 5G). Given that the same clonotypes may exhibit diverse phenotypes, as evidenced by the findings presented in Fig. 5J, we re-categorized the circulating CD28⁺, CD28⁻KLRG1⁺, and CD28⁻KLRG1⁻ CD8⁺ T cells into four categories according to their shared or unique TCR clonotypes (Fig. 6B; Figure S11E). Therefore, each subset had its unique clonotypes distinct from the other categories. We aligned these subsets with the peripheral CD8⁺ T-cell clusters (Fig. 3B and C). Increased Tcm and Tem cells were included within the CD28⁻KLRG1⁺ associated subset, alongside Teff cells (Figure S11F), due to this subset composed cells with either display a CD28⁻KLRG1⁺ phenotype or exhibit a CD28⁻KLRG1⁻ or CD28⁺ phenotype, sharing the same clonotypes as CD28⁻KLRG1⁺ subset (Fig. 6B; Figure S11E). We matched the unique clonotypes of the four circulating subsets to their corresponding clonotypes in patients’ tissues using paired blood-tissue samples. All circulating subsets had overlapping clones existed in various tissues (Fig. 6C-E). These shared clones were distributed across various tissue T-cell clusters, with a greater overlap of tissue clones in the circulating CD28⁻KLRG1⁺ associated and CD28⁻KLRG1⁻ subset (Figure S11G). The tissue T-cell clones shared with these two circulating CD8⁺ T-cell subsets had high proportions of Teff and Tex cells (Figure S11H). Tracing the specific tissue type that presented the overlapped TCR clones shown in Fig. 6C, we found that the TILs were shared with the circulating CD28⁻KLRG1⁺ associated category clones in all patients, displaying high clone counts (194 and 20) in patients 5 and 6, moderate (8) in patient 2, and the least (2) in patient 1 (Fig. 6D, E). Interestingly, patients 5, 6 and 2 had longer survival than patient 1, with PFS of 55, 45, 50 and 6 months, and OS of 55, 45, 50 and 9 months, respectively (Table S3). We also found that the tumor tissue clones shared with the circulating CD28⁻KLRG1⁻ unique clones were higher in the patients receiving NCIT (patient 5, 6 and 1) than in patient 2 treated with neoadjuvant chemotherapy (Fig. 6D, E), which might be associated with the different neoadjuvant treatment strategies applied. The CD28⁺ unique and CD28⁺ & CD28⁻KLRG1⁻ shared categories shared few clonotypes with the tissues (Fig. 6C). These findings indicate that the circulating CD28⁻KLRG1⁺ CD8⁺ T cells shared more clones with TILs, contributing to better treatment outcomes.

Fig. 6.

Circulating CD28⁻KLRG1⁺ CD8⁺ T cells share clones with TILs. (A) tSNE visualization of CD8⁺ T cell clusters in 25 tissue samples. (B) Three newly emerged circulating CD8⁺ T-cell subsets were categorized into four classes. Clonotypes either uniquely within the CD28⁻KLRG1⁺ CD8⁺ subset or shared with the other subsets were defined as CD28⁻KLRG1⁺ associated clones. Each class indicated the number of clonotypes. Newly emerged clones were only identified at the P1 time point but not at P0 in four paired peripheral samples. (C) Total number of clonotypes shared between the four circulating CD8⁺ T-cell classes and all tissue samples. (D and E) Number (D) and percentage (E) of shared clones between four circulating CD8⁺ T-cell classes and each tissue derivation. (F) tSNE visualization of tumor-infiltrating CD8⁺ T-cell (TIL) clusters (left). Available productive clones are highlighted in red (right). 8 NSCLC patients. (G) tSNE visualization of potentially tumor-reactive T (pTRT) cells in TILs (left). pTRT, ≥ 2 clonal size. pTRT cells shared clones with circulating CD28⁻KLRG1⁺ CD8⁺ T cells, defined as shared pTRT cells (right, peripheral blood samples were available from 7 NSCLC patients). (H) Bar, pTRT or shared pTRT cells in each TIL cluster. Line, shared pTRT cells in pTRT cells in each TIL cluster. (I) Bar, pTRT or shared pTRT cells in the total TILs of each patient. Line, shared pTRT cells in the pTRT cells of each patient. Peripheral blood sample was not available for Pt8. (J) Cluster composition of pTRT cells in each patient. MPR, major pathological response

Next, we specified the shared clones between circulating CD28⁻KLRG1⁺ CD8⁺ T cells and TILs among the pTRT cells. In all tumor tissues from eight patients, the productive clones were identified in 70.4% (3465/4922) of CD8⁺ T cells (Fig. 6F), and pTRT cells comprised 43.4% (2137/4922) of CD8⁺ T cells (Fig. 6G). We also examined T cells specific to non-tumor antigens, such as viral antigens, in the pTRT clones from the public TCR database (VDJdb). The results showed that 7.9% (39/493) pTRT clonotypes were identified recognizing non-tumor antigens, occupying 6.7% (143/2137) of pTRT cells. Since tumor antigen-specific TCR clones are shared with virus antigens in viral-induced as well as nonviral-induced tumors [40–42], all the pTRT clones were included in the following study. We matched the overlapping clonotypes pTRT between TILs and three circulating CD8⁺ T-cell subsets: CD28⁻KLRG1⁺, CD28⁻KLRG1⁻ and CD28⁺, available at surgery in 7 patients. The shared clones were distributed across various clusters of the TILs (Fig. 6G; Figure S11I). CD28⁻KLRG1⁻ subset shared more pTRT clones compared with the CD28⁻KLRG1⁺ and CD28⁺ subsets, partially due to relative higher number of CD28⁻KLRG1⁻ CD8⁺ T-cell subset (7010 count) compared with the other two subsets (4999 and 947 count) (Figure S11J). Notably, pTRT clones shared with each circulating CD8⁺ T-cell subsets exhibited distinct T-cell differentiation patterns. The highest proportions of memory cells (Tcm and Tem) and the lowest proportions of Teff cells were observed in pTRT clones shared with the CD28⁺ CD8⁺ T-cell subset. pTRT clones shared with the CD28⁻KLRG1⁻ CD8⁺ T-cell subset displayed moderate and balanced proportions of Tcm and Tem cells. Meanwhile, a shift from Tcm to Tem cells was observed in pTRT clones shared with the CD28⁻KLRG1⁺ CD8⁺ T-cell subset (Figure S11K). These results suggested that, compared with the first two groups of shared pTRT clones, those shared with the CD28⁻KLRG1⁺ CD8⁺ T-cell subset might undergo smoother differentiation within the TME, maintaining components balance memory functionality as well as effector activity. We then addressed the pTRT clones shared with the CD28⁻KLRG1⁺ CD8⁺ T-cell subset, which were subsequently defined as shared pTRT clones. High proportions of shared pTRT were observed in the Teff and Trm TILs, but low in the Tex TILs (Fig. 6H). The proportions of the shared pTRTs relative to total pTRTs were high in the Teff, Trm, and Tem TILs (Fig. 6H). Subsequently, we analyzed the shared pTRT clones from each patient. Patients with a long PFS had high proportions of the shared pTRT clones either in total TILs or in pTRT cells, though two of them experienced non-MPR disease (Fig. 6I). In addition, pTRT cells from patients with long-PFS were composed of a high proportion of Tem cells. However, the proportion of Tex cells was increased in the pTRT cells in patients who were treatment-naïve, receive neoadjuvant chemotherapy, or received NCIT but with short PFS (Fig. 6J). Collectively, these findings indicate that the circulating CD28⁻KLRG1⁺ CD8⁺ T cells supply the pTRT TILs, promoting the specific antitumor immune response in the TME.

Zn-MT pathway regulates circulating CD28⁻KLRG1⁺ CD8⁺ T cells

To confirm the genes that were differentially expressed in the CD8⁺ T-cell subsets revealed by scRNA-seq analysis, we sorted the enriched PBMCs from six healthy donors into six CD8⁺ T-cell subsets with different levels of CD28, KLRG1 and CD57 expression, and qualified transcriptional levels of several associated genes in these subsets. MT1E and MT1G were highly expressed in the CD28⁻KLRG1⁺CD57⁺ and CD28⁻KLRG1⁺CD57⁻ subsets, respectively, either compared to the CD28⁺ subset or among the four CD28⁻ subsets (Fig. 7A). MT1X, MT1F and MT2A expressed comparably among all CD8⁺ T-cell subsets (Fig. 7A). KLF6 was expressed at a relatively low level in the CD28⁻KLRG1⁻CD57⁻ subset (Fig. 7A). TCF1 did not show significantly different expression among the four CD28⁻CD8⁺ subsets, whereas TOX expression was enhanced in the CD28⁻KLGR1⁺ CD8⁺ subsets (Fig. 7B). The protein levels of GZMB, GZMK and perforin varied across the CD28⁻CD8⁺ T-cell subsets, with high production observed in the CD28⁻KLRG1⁺ CD8⁺ subsets (Fig. 7C).

Fig. 7.

Zn-MT pathway regulates phenotype and function of CD28⁻KLRG1⁺ CD8⁺ T cells. (A and B) mRNA levels of associated genes in sorted CD8⁺ T-cell subsets (n = 6, healthy donors). Data are presented as median with interquartile. (C) Levels of GZMB, GZMK, and perforin by flow cytometry (n = 6, treatment-naïve NSCLC patients). Data are presented as median with interquartile. (D) Intracellular Zn²⁺ levels in CD8⁺ subsets analyzed using the ZP1 probe and flow cytometry (n = 4, treatment-naïve NSCLC patients). Data are presented as the mean with individual values. (1–6) in (A, B and D) are shown in (A). Statistical analysis for (C and D) was performed using a one-way repeated ANOVA test, followed by pairwise comparisons using emmeans, with p-values adjusted using the Benjamini-Hochberg method. (E-G) Purified CD8⁺ T cells were stimulated with anti-CD3 and anti-CD28 antibodies for 72 h, with Zn supplementation added after the first 24 h of stimulation. (E) Transcriptional levels of MT isoforms and KLF6 gene (n = 7). (F) Protein levels of MT isoforms and KLF6 (n = 3). Numbers, relative expression normalized to β-actin. (G) Subcellular localization of MT (red) and KLF6 (green) (n = 2). DNA was stained with DAPI (scale bar = 5 μm). (H and I) T cells were treated as described above, and tested after 4 days of stimulation using flow cytometry (n = 4). Data are presented as mean ± standard error of the mean (SEM). Paired T Test. (H) Intracellular Zn²⁺ levels in CD8⁺ subsets. Solid-filled untreated T cells; thin line, T cells treated for 30 min with the membrane permeable Zn²⁺ chelator TPEN (50 µM) where EDTA was depleted. (I) Proportions of CD8⁺ T-cell subsets with different levels of CD28, CD57, and KLRG1 expression. (E-I), Data include samples from 5 NSCLC patients and 3 patients with other tumors. (J) Associations between plasma Zn²⁺ levels and CD28⁻KLRG1⁺ CD28⁺ T cells in peripheral blood. n = 16 NSCLC patients. Grouped as median Zn²⁺ level: 27.69 µM. Data are presented as median with interquartile. Independent-Sample T test. (K-L) Purified CD8⁺ T cells were treated as described above and assessed using ATAC-seq analysis (n = 3, healthy donors). (K) GO enrichment of genes associated with different peaks. (L) Normalized ATAC-seq signal profiles across 12 gene loci in human CD8⁺ T cells. Differential peaks between groups are highlighted in gray (P < 0.05). (M) Purified CD8⁺ T cells were treated as described above. Proliferation was measured on the indicated day using CCK-8 assay (n = 2). Two-way repeated ANOVA test was followed by pairwise comparisons using T test, with p-values adjusted using the Benjamini-Hochberg method

Subsequently, we measured the Zn levels in each CD8⁺ T-cell subset using flow cytometry. Consistent with the high levels of MTs in the CD28⁻KLRG1⁺CD57⁺ and CD28⁻KLRG1⁺CD57⁻ subsets (Fig. 7A), enhanced Zn levels were observed in both subsets (Fig. 7D). During the early stage of CD8⁺ T-cell activation (within 72 h), which was stimulated by TCR signaling (anti-CD3 and anti-CD28 antibodies), both MT isoform transcription and total MT proteins were upregulated by Zn supplementation (Fig. 7E, F). And increased MT proteins translocated from the cytoplasm to nucleus (Fig. 7G). Zn supplementation also elevated KLF6 protein level (Fig. 7F) and promoted KLF6 to accumulate in the nucleus (Fig. 7G). Meanwhile, Zn supplementation increased the proportion of KLRG1⁺ subset and decreased the proportion of CD57⁺ subset in CD8⁺ T cells, although the difference was not significant (Fig. 7H, I). We also examined the phenotypic changes in stimulated CD8⁺ T cells throughout one full cycle of activation and expansion, most ending on day 10–12 and immediately followed by a contraction phase, while supplementing with Zn. KLRG1 maintained slightly increased expression until day 7, after which it recovered. CD57 expression remained relatively low, and CD28 expression did not change significantly throughout the study period (Figure S12A). Based on peripheral blood samples collected from 16 NSCLC patients (Table S18), we measured a median Zn level of 27.69 µM (95% CI: 26.68–28.96 µM) in peripheral blood. The plasma Zn levels were not associated with age, histological types, tumor locations, and stages in these patients. However, patients with higher plasma Zn levels had higher percentages of CD28⁻KLRG1⁺ CD8⁺ T cells in peripheral blood, with a markedly evaluated CD57⁺ subset and moderate increase in the CD57⁻ subset (Fig. 7J; Figure S12B). While MT regulated copper (Cu) hemostasis besides Zn, a role of Cu-MT signaling in regulating CD28⁻KLRG1⁺ CD8 T cells was not observed in our study (Figure S12C-E). These in vitro and vivo findings suggest that activation of Zn-MT signaling pathway may regulate CD28⁻KLRG1⁺ CD8⁺ T-cell subsets.

As Zn-MT interaction played a crucial role in chromatin structure, influencing the transcription expression [43], we analyzed the chromatin accessibility of Zn supplemented CD8⁺ T cells using the ATAC-seq method (Figure S13A-C). Zn supplementation did not significantly change the genomic distribution of peaks, although the total peaks increased (Figure S13D, E). We therefore analyzed the genes associated with these differential peaks. The upregulated GO pathways were enriched in structure-related development, localization and the JNK cascade, which indicated T-cell migration and activation (Fig. 7K; Figure S13F). Downregulated GO pathways were enriched in the leukocyte-related activation (Fig. 7K; Figure S13F). Upregulated and downregulated KEGG pathways were both enriched in the autophagy pathway (Figure S13G). Autophagosome formation in CD8⁺ T cells was similar regardless of Zn supplementation (Figure S13H). These results demonstrated that the CD8⁺ T cells remained the capacity for flexible and harmonious adjustment under Zn supplementation. Meanwhile, the chromatin accessibility of the promoter (≤ 1 kilo bases) of MT1E, MT1G, and MT1P genes was increased (Fig. 7L).

To elucidate the results of the GO enrichment analysis (Fig. 7K), we evaluated the proliferation of CD8⁺ T cells supplemented with Zn in vitro. Zn supplementation facilitated CD8⁺ T-cell proliferation, especially within eight days of T-cell activation period (Fig. 7M). However, Zn supplementation did not significantly affect the growth of human and mouse lung cancer cell lines compared to its effects on CD8⁺ T cells (Figure S14A). Assessing the transcriptional levels of TCR signaling downstream genes (FOS, JUN, NFATC3 and NFKBIA), we found similar transcriptional levels of these genes among the four sorted circulating CD28⁻ CD8⁺ T-cell subsets (Figure S14B). Zn supplementation increased the FOS transcription in activated CD8⁺ T cells (Figure S14C), which was reduced by siRNA MTs (Figure S14D). These results demonstrate that the Zn-MT signaling pathway partially induces CD8⁺ T-cell amplification through upregulating transcription of FOS gene. Subsequently, we evaluated the responsiveness of 3-day Zn supplemented CD8⁺ T cells to the second stimulus of T-cell activation. Zn supplementation significantly enhanced CD8⁺ T cells to secret IFN-γ, TNF-α, IL-2 and GZMB upon a second stimulation, either with stimulus of OKT3 (anti-CD3 antibody) or phorbol 12-myristate 13-acetate (PMA) + ionomycin (Figure S15). The relative lower sensitivity of CD8⁺ T cells to OKT3 compared to PMA + ionomycin might due to attenuated sensitivity of CD8⁺ T cells under repeated TCR signaling stimulation within short-term. Collectively, these findings reveal that the Zn-MT pathway is involved in the regulation of the phenotype and function of CD28⁻KLRG1⁺ CD8⁺ T cells.

Zn supplementation plus chemoimmunotherapy improves local and systemic immune response in vivo

To assess whether Zn supplementation combined with CIT enhances treatment outcomes in vivo, we employed the orthotopic and subcutaneous allogeneic transplanted lung cancer mouse models. Lewis lung cancer (LLC) cells stably expressing ovalbumin (OVA) were implanted either orthotopically or subcutaneously into male C57BL/6 mouse. CIT was administrated twice daily for a total of three doses. Zn supplementation was given concurrently with CIT and continued until the end of the study (Fig. 8A; Figure S16A and S17A). Zn supplementation at a dose of 5 mg/kg demonstrated no obvious toxicity, as evidence by comparing plasma alanine aminotransferase levels between the Zn supplemented and control group (Figure S16B).

Fig. 8.

Zn supplementation combined with chemoimmunotherapy improves antitumor response in an orthotopic model. (A) Schematic representation of LLC-OVA cell incubation and triple treatment strategy. Cisplatin (CDDP) at 1 mg/kg, anti-PD-1 antibody at 2 mg/kg, Zn at 5 mg/kg were administered. (B) In vivo bioluminescence imaging using luciferase after 8 days of initial treatments. Control: n = 12; Zn: n = 9; CDDP + PD-1: n = 11; CDDP + PD-1 + Zn: n = 10. (C, D) Lymphocyte infiltration (C) and checkpoint molecule expression (D) in tumor microenvironment. n = 7; 5; 6; 5. (E-G) Levels of T cells (E), and CD28, KLRG1, and PD1 expression on CD8⁺ (F) and CD4⁺ (G) T cells in peripheral blood. n = 6; 5; 6; 5. (H) Cytokine production in circulating CD8⁺ T cells. n = 3; 5; 6; 3. (I) Cytokine production in KLRG1⁺ CD8⁺ T cells. n = 6; 3. Data are presented as median with interquartile. One way ANOVA test followed by emmeans, and adjusted using the Benjamini-Hochberg method. (J) Concentrations of IFN-γ, TNF-α, and IL-2 in non-co-cultured and co-cultured supernatants from splenocytes (2 × 10⁶) across the four treatment groups (n = 6). Data are presented as mean with range. Independent-samples T Test. (K) Schematic representation of tumor antigen-specific assessment in vitro. (L, M) Representative histograms (L) and summary data (M) of CD107a, IFN-γ, and TNF-α levels in CD44⁺CD8⁺ T cells (n = 5). Data are presented as mean with individuals. Independent-samples T Test. MFI, median fluorescence intensity

Given the short survival time of the orthotopic lung cancer model (approximately 20 days in the control group), we euthanatized the mice on the eighth day following the initiation of treatment to evaluate the TME and systemic immune response. At the same time, we confirmed antitumor effects of CIT and CIT plus Zn treatment using in vivo bioluminescence assay. Our findings showed that Zn supplementation alone did not impact tumor growth (Fig. 8B). However, both CIT treatment and CIT in combination with Zn demonstrated notable antitumor effects, with inhibitory rate of 88.1% and 97.3%, respectively (Fig. 8B).

Analysis of the TME in the orthotopic lung cancer revealed that myeloid-derived cells, B cells and nature killer (NK) cells were not significantly affected by the treatments (Figure S16C). However, T-cell levels increased in the CIT and CIT plus Zn groups (Fig. 8C). The proportions of CD4⁺ T-cell in total T cells slightly decreased in the CIT and Zn plus CIT groups, but did not reach statistical significance (Fig. 8C). The differentiation status of T cells remained unchanged under treatments (Figure S16D). Notably, both CIT and CIT plus Zn groups showed elevated KLRG1 expression on CD8⁺ T cells (Fig. 8D). The PD-1 expression on CD8⁺ T cells remained consistent across groups, whereas its expression on CD4⁺ T cells significantly decreased in CIT and CIT plus Zn groups (Fig. 8D). Neither treatment increased TIM3 expression on CD8⁺ as well as CD4⁺ T cells (Figure S16E).

Circulating CD8⁺ and CD4⁺T-cell levels remained stable under treatments (Fig. 8E). In the CIT and CIT plus Zn groups, CD28 expression increased on both CD8⁺ and CD4⁺ T cells. KLRG1 expression increased on CD8⁺ T cells in the CIT groups. Additionally, CD28⁺KLRG1⁺ subset expanded among CD8⁺ T cells, while CD28⁺KLRG1⁻ subset increased among CD4⁺ T cells in the CIT and CIT plus Zn groups (Fig. 8F and G). PD-1 expression decreased on both CD8⁺ and CD4⁺ T cells (Fig. 8F and G). Furthermore, circulating CD8⁺ T cells exhibited heightened IFN-γ production in response to TCR signaling stimulation in the CIT and CIT plus Zn groups (Fig. 8H).

Further evaluating cytokine production in circulating KLRG1⁺ CD8⁺ T cells, we focused on the CIT and CIT plus Zn groups, due to limited numbers of the subset detected in the control and Zn groups. The CIT plus Zn group’s KLRG1⁺ CD8⁺ T cells produced significantly higher levels of IL-2 and TNF-α than those of the CIT group (Fig. 8I). Moreover, KLRG1⁺ subset showed greater cytokine production than KLRG1⁻ subset (Figure S16F). Splenic T-cell analysis revealed an increase in CD8⁺ Tcm and Tem cells in the CIT and CIT plus Zn groups (Figure S16G, H). KLRG expression increased in the CIT group, while PD-1 expression decreased in CIT and CIT plus Zn groups (Figure S16I).

The CIT and CIT plus Zn groups demonstrated significant antitumor effects in the subcutaneous lung cancer model, as observed after 13 days of treatment initiation (Figure S17B, C). The proportions of myeloid-derived cells, B cells, and NK cells remained unchanged across treatments (Figure S17D, E). However, T cells increased, with a notable shift toward increased cytotoxic CD8⁺ T cells and a reduction in CD4⁺ T cells proportions in the Zn plus CIT group (Figure S17F, G). Treatment influenced T-cell differentiation, with an increase in naïve and TEMRA subsets and decrease in Tem subset among both CD8⁺ and CD4⁺ T cells in the CIT group—changes that were more significance in the CIT plus Zn group (Figure S17H). CD8⁺ T cells showed no alterations in KLRG1, PD-1 and TIM3 expression in the subcutaneous model (Figure S17I). Splenic analysis revealed increased T-cell proportions in the CIT group (Figure S17J). Despite treatments, the composition of splenic T-cell differentiation remained stable, with Naïve T cells being the predominant subset (Figure S17K).

Lastly, we evaluated the antitumor activity of the splenocytes isolated from subcutaneous tumor-bearing mice. Splenocytes from the Zn plus CIT group exhibited the highest production of IFN-γ, TNF-α and IL-2 among the four groups (Fig. 8J). When re-exposed to LLC-OVA cells in vitro, splenocytes from the Zn plus CIT group exhibited enhanced antitumor responses, as indicated by increased secretion of antitumor cytokines (Fig. 8J). These results suggest that the systemic antitumor memory was strengthened in the Zn plus CIT group. To verify whether improved antitumor memory was antigen-specific, we evaluated the OVA-specific immune response in primed OT-1 splenocytes in vitro (Fig. 8K). OT-1 splenocytes primed with Zn supplementation displayed greater degranulation capacity (CD107a⁺) and enhanced cytokine production (IFN-γ and TNF-α) compared to those primed without Zn supplementation (Fig. 8L, M).

Taken together, our findings from orthotopic and subcutaneous lung cancer mouse models suggest that Zn supplementation enhances CIT treatment outcomes by improving local and the systemic antitumor immune response.

Discussion

As CIT treatment becomes the standard of first-line treatment in patients with advanced NSCLC, there has been more attention on predictive biomarkers for the treatment outcomes. In this present study, we report for the first time that the baseline CD28⁻KLRG1⁺CD57⁺ CD8⁺ T cells and on-treatment CD28⁻KLRG1⁺ CD8⁺ T cells in peripheral blood are predictive candidates for the first-line CIT treatment in advanced NSCLC. These cells are mainly in the TEMRA status and are characterized by activated antitumor capacity. More importantly, these cells share with pTRT clones and are likely to be reactivated by PD-1 inhibitors. The Zn-MT pathway is involved in regulating these cells. The addition of Zn to CIT resulted in an improved antitumor effect in both the orthotopic and subcutaneous lung cancer mouse models. Our findings help predict the therapeutic efficacy of CIT and provide a new understanding of the late-differentiated T cells participating in the antitumor immune response.

We found that the circulating CD28⁻KLRG1⁺ CD8⁺ T cells during CIT were positively associated with the treatment response and survival in patients with advanced NSCLC receiving first-line CIT treatment. However, on-treatment CD28⁻KLRG1⁻ CD8⁺ T cells were negatively correlated with the patient survival. Although the CD28⁻CD8⁺ T cells, regardless of KLRG1 expression, were mainly composed of the terminally differentiated TEMRA cells, the CD28⁻KLRG1⁺ CD8⁺ subset exhibited increased capacity for activation and proliferation, compared with the CD28⁻KLRG1⁻ CD8⁺ subset. Moderate checkpoint molecule expression in the CD28⁻KLRG1⁺ CD8⁺ subset indicated compensatory negative feedback during activated. Simultaneous expression of TCF1 and TOX may limit excessive T-cell proliferation and help prevent CD8⁺ T cells from entering terminal exhaustion, supporting the stem-like or Tpex traits of CD28⁻KLRG1⁺ subset. The subsets retained capacity to produce functional cytokines and responded strongly to TCR stimulation. TCR clonotype analysis revealed potential plasticity for CD28⁻ subsets to reversely differentiate into less-differentiated CD28⁺ subsets under NCIT treatment. Several studies have reported that the terminally differentiated CD8⁺ T cells in peripheral blood, which lacked the costimulatory receptors (CD28, ICOS, OX40), and highly expressed CD45RA, CD57, KLRG1, GMZB and perforins, demonstrated an improved response to PD-1/PD-L1 blockade in advanced NSCLC [13] and metastatic urothelial cancer [14]. CD27 and ICOS are recognized as important makers for delineating CD8⁺ T-cell differentiation and activation; however, they were not assessed in the present study; future work will incorporate these markers to provide a more comprehensive characterization. Nevertheless, another study showed that the patients with recurrent advanced NSCLC with high peripheral CD28⁻KLRG1⁺CD57⁺ CD8⁺ T cells did not benefit from the single-agent PD-1/PD-L1 inhibitor [10]. In these patients, circulating CD28⁻KLRG1⁺CD57⁺ CD8⁺ T cells showed decreased proliferation and IL-2 production compared to total CD28⁻ CD8⁺ T cells [10]. And the PD-1⁺ CD8⁺ T cells with CD28 co-expression in peripheral blood were associated with better response to PD-1/PD-L1 blockade after multiple-line treatments [44, 45]. In our study, all patients were treatment-naïve and received first-line CIT. These patients probably had a better antitumor immune status than the multiple-line treated patients, as the immune cells were always deteriorated under multiple cycles of chemotherapy and radiotherapy [10]. These divergent results suggested that immunotherapy applied in the early period might prevent the late-differentiated T cells from terminal exhaustion, thus being reinvigorated by immunotherapy and improving antitumor efficacy. Our results indicate that a paradigm where a “terminal-differentiated phenotype” does not necessarily equate to “functional exhaustion”. Since flow cytometry has been widely used in clinical laboratories. The detection of CD28⁻KLRG1⁺CD57⁺CD8⁺ or CD28⁻KLRG1⁺CD8⁺ T cells requires only a small number of markers, making the assay simple. And these molecules are expressed at sufficiently high levels to ensure sensitivity and reproducibility. Thus, this biomarker could be feasibly incorporated into routine peripheral blood monitoring at different treatment stages. Collectively, our findings suggest that assessment of the baseline CD28⁻KLRG1⁺CD57⁺CD8⁺ or on-treatment CD28⁻KLRG1⁺ CD8⁺ T cells in peripheral blood would be a reliable and convenient way to predict the treatment outcomes in advanced NSCLC patients with first-line CIT.

Our results revealed that different CD8⁺ T-cell subsets experienced various destinies under CIT. Clonal expansion observed in the CD28⁻KLRG1⁺ subset during CIT suggested that this subset was the primary group of cells that responded to the CIT treatment. As CD28 is a pre-requisite for priming and activation of T cells, the high occurrence of clonal replacement in the CD28⁺ subset in our findings implied that it might be an entry point for new emerged clones in the immune response. Meanwhile, some of the disappeared clones previous observed in the baseline CD28⁺ subset might transit to other differentiated subsets with distinct phenotypes. These conversions between CD28⁺ and CD28⁻ subsets were further proven in our conversion pattern analysis between the CD8⁺ T-cell subsets. In particular, in addition to the conventional differentiation from Tem (main differentiated status in CD28⁺ subset) to TEMRA (mainly in CD28⁻ subset), reverse differentiation from the terminal differentiated CD28⁻ subset to the less-differentiated CD28⁺ subset also occurred during CIT treatment. The favorable responsiveness to immunotherapy and flexible plasticity of the CD28⁻KLRG1⁺ CD8⁺ subset in the peripheral blood supports its crucial role in the antitumor immune response and prediction of the treatment outcomes. These results supported the predictive capacity of the CD28⁻KLRG1⁺ CD8⁺ subset. While the baseline CD57⁺ and CD57⁻ subpopulations within the CD28⁻KLRG1⁺ CD8⁺ subset displayed opposite associations with patient survival, both subsets were associated with improved survival during CIT. The CD57⁺ and CD57⁻ subpopulations within the CD28⁻KLRG1⁺CD8⁺ subset shared more phenotypes and functional characteristics compared to those in the CD28⁻KLRG1⁻CD8⁺ subset. Based on the TCR clonal analysis, the CD28⁻KLRG1⁺CD8⁺ subset both at baseline and during treatment displayed the highest composition of expanded clones, which indicated this subset was sensitive to anti-PD-1 treatment and could be reinvigorated by the treatment. CIT modulated the immune environment, effectively reprogramming these subsets and synchronizing their activation states mitigating. Consequently, their inherent differences were mitigated by the treatment-induced functional adaptations. These findings suggest that both CD57⁺ and CD57⁻ subpopulations within CD28⁻KLRG1⁺ CD8⁺ subset can actively respond to CIT and contribute to the antitumor immune response.

We found that the circulating CD28⁻KLRG1⁺ CD8⁺ subset was clonally shared with the T cells present in multiple tissues. The more clonotypes and clones shared with TILs, the better their survival. Consequently, our results revealed that these shared clones composed of a part of potentially tumor-reactive pTRT clones [34], and heterogeneously distributed in various TIL clusters identified in scRNA-seq analysis. Approximately half of the pTRT clones in the Teff, Trm and Tem clusters were shared with the circulating CD28⁻KLRG1⁺ CD8⁺ T cells, and less than 10% in the Tex cluster. The Teff TILs directly conducted the antitumor toxicity; Trm and Tem subsets supported the long-term antitumor surveillance [46]. Consistently, the pTRT cells enriched the Tem subset in the patients with long survival, whereas the Tex subset enriched in the pTRT cells was associated with poor survival. These results demonstrate that the functional status of pTRT cells is critical for improving the antitumor efficacy; preventing pTRT cells from developing exhaustion by the addition of immunotherapy in treatment, would benefit the clinical outcome. The shared clones of pTRT with circulating CD28⁻KLRG1⁺ CD8⁺ T cells in our study indicate that the systemic immune system supports the local antitumor immunity, partially by providing tumor-reactive T cells that migrate into tumors. Meanwhile, our findings of the shared clones between peripheral CD28⁻KLRG1⁺CD8⁺ T cells and tumor-draining LNs were supported by several studies that revealed that tumor-draining LNs supplied tumor-specific and stem-like CD8⁺ T cells into tumors [47, 48]. The vivid migration of pTRT throughout tumors and circulating required further studies.

Using scRNA transcription and trajectory analyses, we found that the Zn-MT signaling pathway was involved in the phenotype and functional regulation of CD28⁻KLRG1⁺ CD8⁺ T cells. These findings were subsequently confirmed by the results of in vitro experiments. MTs induced by Zn have not yet been reported in CD8⁺ T cells. We found that Zn supplementation increased the transcriptional expression of multiple MT isoforms in CD8⁺ T cells partially through enhancement of chromatin accessibility. MTs maintained metal homeostasis by binding both physiological (zinc, copper, selenium) and toxic (cadmium, mercy, silver, et al.) ions through their cysteine residues [49]. Zn was released from MTs under oxidative stress as well as T-cell activation. Besides the structural and functional role in proteins, Zn acted as a second messenger that prompted two MAPK pathways, the ERK and the JNK, and NF-κΒ signaling pathway in cytoplasm [50]. Consistently, the JNK pathway was enriched in the ATAC-seq analysis. Zn also translocated to the nucleus, where it combined with transcriptional factors, thus influencing the target gene structures and modifying gene transcription [51]. We found that the MTs translocated from the cytoplasm to the nucleus under Zn supplementation, which resulted in the accumulation of MTs and Zn in the nucleus and was closely associated with the alteration of chromatin accessibility. As a result, the Zn-MT signaling forms a feedback loop in CD8⁺ T cells. And with PD-1 co-expression, the CD28⁻KLRG1⁺CD8⁺ subset would benefit from PD-1/PD-L1 blockade for long-term maintenance of activation and functionality. While the mechanistic findings provide biological plausibility for the biomarker observed in the advanced cohort, direct extrapolation across treatment setting should be viewed as preliminary. The precise mechanism requires further investigation.

Lastly, both CIT and CIT plus Zn treatments exhibited excellent antitumor effects in the orthotopic and subcutaneous lung cancer mouse models. Zn has been reported to shape the tolerogenic DCs and alleviate the immunosuppressive polarization of macrophages [24, 52]. However, in our study, infiltration of myeloid-derived cells, neutrophils, B cells and NK cells, did not change with 5 mg/kg Zn supplementation. Tumor growth was also unaffected by supplementation alone. These results suggested that these immune populations are unlikely to be the major contributors to the observed antitumor effects. In the orthotopic model, KLRG1 expression was increased in tumor-infiltrating, circulating and splenic CD8⁺ T cells under the CIT and CIT plus Zn treatments, consistent with our findings in human sample. Importantly, addition of Zn to CIT greatly accelerated the antitumor functionality of KLRG1⁺ CD8⁺ T cells, CD8⁺ T cells and splenocytes, compared to CIT treatment, as evidenced by the in vivo and ex vivo results. These findings revealed that Zn combined with CIT was superior to CIT alone, not only profiling immune cells, but also enhancing the antitumor capacity of these cells. While Zn plus CIT exhibited a similar inhibitory effect on the tumor growth as CIT in short-term observations, we proposed that the construction of antitumor specific immune memory by the triple treatment benefit long-term antitumor efficacy. Consistent with this hypothesis, two patients who were included in the scRNA-seq analysis had high circulating CD28⁻KLRG1⁺ CD8⁺ Teff cells, and did not experience occurrence until the last follow-up, with a PFS of 45 and 60 months, respectively, although they were both assessed as having non-MPR after NCIT treatment. These results indicated that the identification of systemic antitumor immunity helped to provide comprehensive information to predict treatment outcome in immunotherapy. Mouse tumor models differ from human advanced NSCLC, and therefore results should be interpreted with caution. In summary, these in vivo findings suggest that Zn supplementation may promote the antitumor effect of CIT not only by accelerating cytotoxic efficacy but also by improving the long-term immune surveillance; however, clinical validation will be required to confirm these effects.

Our study has several limitations. Firstly, because of the nature of the phase 2 clinical trial, the peripheral blood samples assessed by flow cytometry were limited. This may increase statistical variability and limit generalizability. The prospective design of this clinical study may have reduced bias in the analyses. Meanwhile, we first applied unsupervised analysis to screen the potential circulating CD8⁺ subsets, then validated the subsets in a larger cohort from the same phase 2 study, and lastly integrated Kaplan–Meier survival analyses and univariate and multivariate analyses to evaluate the survival association. These the multi-step analyses support the validity of our findings. Secondly, TCR-based inference of tumor reactivity is indirect, functional validation is required to confirm tumor specificity. Finally, we focused on CIT because it is the standard first-line treatment for advanced NSCLC. While the chemotherapy group was not comparable here, previous studies have reported that neither the CD28⁻KLRG1⁺CD57⁺ nor the CD57⁺ CD8⁺ subset was correlated with the efficacy of chemotherapy in advanced NSCLC [10, 53]. Our present findings warrant validation in larger, independent, prospective cohorts.

Conclusions

In conclusion, we have identified peripheral CD28⁻KLRG1⁺ CD8⁺ T cells as a predictive candidate for outcomes following first-line CIT in advanced NSCLC. We demonstrate that these CD8⁺ T cells are functionally reinvigorated by CIT, share clonality with tumor potential active T cell, and are regulated by the Zn-MT signaling pathway. Our findings provide novel insight into how peripheral CD8⁺ T-cell subsets fuel local anti-tumor immunity.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

CIT

Chemoimmunotherapy

DCB

Durable Clinical Benefit

ICIs

Immune Checkpoint Inhibitors

MACS

Magnetic-Activated Cell Sorting

MPR

Major Pathological Response

NCIT

Neoadjuvant Chemoimmunotherapy

NSCLC

Non-Small Cell Lung Cancer

OS

Overall Survival

PBMCs

Peripheral Blood Mononuclear Cells

PFS

Progression-Free Survival

Tcm

Central Memory T Cell

Tem

Effector Memory T Cell

TEMRA

Effector Memory T Cell Re-expressing CD45RA

Teff

Effector T Cell

TILs

Tumor-Infiltrating Lymphocytes

TME

Tumor Microenvironment

Trm

Tissue-Resident Memory T Cell

Author contributions

Conceptualization: C.Y., X.R., and M.W. Methodology: C.Y., J.Z., H.B., Y.Y., J.L., S.L., L.Z., Z.H., Y.Q., Z.J., and W.Y. Investigation: J.Z., H.B., Y.Y, H.H, L.Z., W.Z., and W.Y. Analysis: C.Y., J.Z., W.Z., Y.M., H.L., and L.Z. Funding acquisition: C.Y., W.Z., and M.W. Project administration: C.Y., X.R., and M.W. Supervision: C.Y., and W.Z. Resources: J.Z., H.B., Y.Y. Writing—original draft: C.Y., J.Z., and W.Z. Writing—review and editing: C.Y., X.R., M.W., and W.Z.

Funding

This research was supported by the National Natural Science Foundation of China under Grants (number 82273083, 82272733), the National Key Laboratory of Druggability Evaluation and Systematic Translational Medicine (number QZ23-9), and the Tianjin Science and Technology Plan Project (24KPXMRC00140).

Data availability

scRNA-seq transcriptome data of the tumor tissues of patients receiving NCIT in the study has been deposited in the Gene expression Omnibus (GEO) dataset (GSE229353). Other datasets used are available from the corresponding author Xiubao Ren on reasonable request as another research is being conducted. This study did not generate any original coding; all utilized code and software tools adhered to official tutorials and were either freely or commercially available.

Declarations

Ethics approval and consent to participate

This study was approved by the Institutional Review Board and Ethics Committee of Tianjin Medical University Cancer Institute & Hospital (E20210014) and conformed to the Declaration of Helsinki and the Good Clinical Practice guidelines. Written informed consent was obtained from all participants. Animal experiments were approved by the Laboratory Animal Ethics Committee of Tianjin Medical University Cancer Institutes & Hospital (AE-2022040), and conducted following the Basel Declaration and institutional guidelines, with adherence to the International Council for Laboratory Animal Science (ICLAS) principles to ensure ethi cal standards.

Consent for publication

All participants provided written informed consent and agreed to the publication of anonymized data. No individual patient-identifiable information is included in this manuscript.

Competing interests

The authors declare that they have no competing interests.