M2 macrophages modulate the differentiation of CD8 + CD101-TIM3 + T cells via the SPP1‒CD44 pathway, influencing the immunotherapeutic response in NSCLC.

Introduction
Lung cancer is a representative malignant tumour and the major cause of cancer-related death, among which the most common type is the non-small cell lung cancer (NSCLC), which takes up about 85% of all lung cancer cases. Despite great advancements in traditional and complementary therapy research, clinical outcomes are still far from satisfactory. The practical application of immune checkpoint inhibitors (ICIs) such as those targeting programmed cell death protein 1 (PD-1) or programmed cell death-ligand 1 (PD-L1) has significantly improved the clinical efficacy and changed the treatment paradigm of NSCLC in recent years. However, owing to the heterogeneity of NSCLC, a nonnegligible proportion of NSCLC patients still respond poorly to immunotherapy even with the same treatment regimen [1–5]. This heterogeneity necessitates a deeper elucidation into the mechanisms of nonresponse to ICI therapy.
One potential way in which tumours evade the immune system is T-cell exhaustion, particularly of CD8 + T cells. Understanding the characteristics and mechanisms of T-cell exhaustion, one of the major mechanisms of nonresponse to immunotherapy, critically enables the checkpoint blockade therapy to succeed. Notably, according to relevant researches, T cells with high levels of inhibitory receptor expression should not be considered depleted but rather dysfunctional or differentiated [6], as indicated by the observation that some exhausted T cells still preserve the chemokine proliferation and production capabilities [7–9]. Thus, some researchers have proposed the concept of T-cell exhaustion as a dynamic progression from a long-lived “preexhausted stem-like progenitor” state to a “terminally exhausted” state [6, 10]. In the study by Hueso et al., stem-like CD8 + T cells differentiate into CD101-TIM3 + T cells, which retain their proliferation and effector capacities before eventually progressing to a depleted CD101 + TIM3 + state, and that blocking the PD-1 pathway increases the proliferation and differentiation of stem-like antigen-specific CD8 + T cells to the effector-like CD101-TIM3 + state, contributing to improved viral control [11]. Moreover, a high proportion of circulating CD8 + CD101hiTIM3 + T cells has been shown to be significantly associated with adverse clinical responses to immunotherapy [12]. Therefore, we investigated the potential mechanisms by which CD8 + CD101-TIM3 + T/CD8 + CD101loTIM3 + T (Pre-CCT T) cells differentiate into CD8 + CD101 + TIM3 + T/CD8 + CD101hiTIM3 + T (CCT T) cells in NSCLC and assessed the relevance of the proportion of these cells to clinical outcomes.
The tumour microenvironment (TME) is an important factor regulating biological tumour processes. In addition, there is a complex relationship between the TME and immunotherapy, where immunotherapy reshapes the TME, and the TME determines the outcome of immunotherapy to some extent [13]. M2 tumour-associated macrophage (TAM) serves as important immunosuppressive immune cell in the TME that excels in inhibiting CD8 + T-cell function and boosting Tregs infiltration [14, 15]. TAMs cause T-cell dysfunction and depletion by secreting cytokines and metabolites [16, 17]. Recent evidence suggests that the high infiltration of macrophages usually indicates unresponsive anti-PD-1/PD-L1 immunotherapy and poor clinical results [18–20]. Therefore, TAMs are considered important targets for reversing the unresponsiveness of ICI treatment [21, 22]. Consequently, we explored the role of TAMs in promoting CD101-TIM3 + T cell differentiation and the underlying mechanisms involved. Recent studies have identified a unique TAM subpopulation characterized by the exclusive expression of SPP1. These TAMs may activate fibroblasts and promote T-cell exhaustion through SPP1–CD44 and CD155–CD226 ligand–receptor interactions, thereby reshaping the metastatic lymph node microenvironment to facilitate disseminated tumour cell colonization and proliferation [23]. Further analyses revealed that SPP1 mediates T-cell functional impairment primarily via the CD44 signalling axis. Macro_SPP1high macrophages significantly reduce T-cell numbers and induce stress responses in both CD4 + and CD8 + T cells, thereby mediating colon cancer liver metastasis (CCLM) through the SPP1/CD44/PI3K/AKT signalling pathway [24]. Thus, this provides a foundation for investigating the role of the SPP1–CD44 pathway in TAM-promoted CD101-TIM3 + T-cell differentiation.
Major breakthroughs have been achieved in Single-cell RNA sequencing (scRNA-seq) technology in the last two decades, making it possible to identify the molecular characteristics of varying immune cell populations in the TME. According to prior studies, obtaining molecular profiles of immune cells utilizing scRNA-seq data and further using them to explore gene expression profiles may effectively assist in predicting immunotherapeutic response of cancer patients [25, 26]. The development of scRNA-seq technology alongside relevant data analysis methods profoundly helps to parse the complex interactions between tumour-infiltrating lymphocytes (TILs) and other cell types in the TME [25] and prompts researchers to focus their studies on crosstalk between more refined cellular subpopulations, such as TAMs and specific T-cell subpopulations, coupled with their impact on the efficacy of ICIs.
In this study, we elucidated the process and underlying signalling pathways by which M2-TAMs promote Pre-CCT T cells to be differentiated into CCT T cells and the impact of this biological behaviour on the response to immunotherapy in NSCLC through comprehensive bioinformatics analysis of single-cell RNA sequences and multiple experimental validations. The findings in the study theoretically assist in the elucidation of the underlying mechanisms of the immunotherapeutic response in NSCLC and the formulation of new strategies for improving the immunotherapeutic response.

Methods

Data acquisition and preprocessing
We obtained scRNA-seq data from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) for the GSE207422 dataset, which consists of 15 NSCLC samples. Three independent bulk RNA-seq datasets were from The Cancer Genome Atlas (TCGA) (https://portal.gdc.cancer.gov/) and GEO databases: TCGA-lung adenocarcinoma (LUAD) (n = 600, including 541 tumour samples and 59 normal samples), TCGA-lung squamous cell carcinoma (LUSC) (n = 553, including 502 tumour samples and 51 normal samples) and GSE135222 (n = 27, including 8 samples with response (R) after anti-PD-1 immunotherapy and 19 samples with immunotherapeutic nonresponse (NR)). Principal component analysis (PCA) served for testing the degree of correction. Finally, nonbiotechnological biases-induced batch effects were adapted to the “ComBat” method using the “sva” R package [27]. This study utilised preexisting ethically approved publicly available datasets from primary studies. Supplementary Table 1 lists the details of GSE207422 and GSE135222. Supplementary Fig. 1 illustrates the experimental procedures.

ScRNA-seq data processing
The “Seurat” R package (v.4.2.0) [28] served for scRNA-seq data analysis. High-quality scRNA-seq data were retained by applying two filtering measures to the raw matrix of each cell: including only genes expressed in ≥ 1 single cell and eliminating cells expressing < 200 genes. Data normalization was achieved via the “normalizedata” function in the Seurat R package, coupled with the confirmation of highly variable genes via the “FindVariableFeatures” function. Following PCA, the relevant PCs were subjected to cell clustering analysis. Single cells were clustered into 28 clusters via the “FindClusters” algorithm, with 50 PCs and a resolution of 0.8. Afterwards, the cell types were annotated with cell type-specific biomarkers.

Constructing single cell trajectories in NSCLC
Single-cell trajectories were analysed by using Monocle2, aiming at ascertaining the cell-state transitions [29]. For the cell lineage trajectory analysis of the CD8 + T-cell subsets, we sorted the cells in pseudotime order by selecting genes that met the thresholds of mean_expression ≥ 0.1 and dispersion_empirical ≥ 1, as identified by Monocle2 [30]. The “DDRTree” algorithm was used to reduce the number of dimensions. Branch expression analysis modelling (BEAM) was performed for the identification of genes with significant branch-dependent expression [29].

Cell − cell communication analysis
The “CellChat” R package (v.1.1.3) was used to assess cell‒cell communication between each cell type in NSCLC to examine molecular interaction networks [31]. Specifically, intercellular communication was simulated using gene expression data from annotated cells as input information and incorporating the interactions of ligands, receptors, and their cofactors. Ligand receptor pairs were estimated using “CellChat”. In this process, the outgoing signals were ligands and the incoming signals were receptors. For ligand–receptor interactions, P < 0.05 indicated statistical significance.

GO and KEGG analysis
GO [32] enrichment analysis revealed three classifications: molecular function (MF), biological process (BP), and cellular component (CC). The KEGG [33] is a well-known knowledge database used for comprehensively analysing gene function. GO and KEGG analyses relied on the “clusterProfiler” R package (v.4.2.2) using the significant DEGs (P < 0.05).

Weighted gene coexpression network analysis (WGCNA)
WGCNA helped to find clusters (modules) of highly correlated genes for relating modules to one another and to external sample traits [34]. With the use of the PickSoftThreshold function in the “WGCNA” R package (v.1.70-3), we increased the coexpression similarity to the power β = 6 to establish the weighted adjacency matrix and then converted it into a topological overlap matrix (TOM). Taking into account the variations in the TOM, we conducted average linkage hierarchical clustering. Pearson correlation analysis served for confirming the relationships between the gene consensus modules and Pre-CCT T cells and macrophages.

Identification of NSCLC subtypes and survival analysis
An unsupervised hierarchical cluster analysis was performed to characterize the different subtypes. We identified two NSCLC clusters using hub genes. For assessing their prognostic significance, we adopted the Kaplan‒Meier (KM) method for survival analysis, and determined the statistically significant differences via the log-rank test.

Gene set enrichment analysis (GSEA)
With the objective of more deeply elucidating the potential influencing mechanism of hub genes against the process of Pre-CCT T cells being differentiated into CCT T cells, we conducted GSEA to measure the enrichment of predefined gene sets at the top or bottom of the sorting table via the “clusterProfiler” R package (version 4.2.2). The predefined gene sets (c2.cp.kegg.v7.5.1. symbols) were screened from the Molecular Signatures Database (MSigDB) [35].

Immunotherapeutic response and immune cell infiltration analysis
The cell types from the single-cell analysis were deconvolved to the bulk RNA-seq dataset (GSE135222). In addition, the relative enrichment score of each immune cell from the gene expression profile of each NSCLC sample was calculated using ssGSEA. We presented the differences in immune cell infiltration levels among the different immunotherapeutic response groups under the assistance of the “ggplot2” R package (version 3.3.6).

Patients and samples
For patients initially diagnosed with NSCLC and requiring anti-PD-1 treatment at the Second Affiliated Hospital of Chongqing Medical University, tumour tissue specimens were collected at their first diagnosis. Peripheral blood samples were collected before receiving anti-PD-1 treatment for flow cytometry (FCM) detection. The pathology Department of the Second Affiliated Hospital of Chongqing Medical University took charge of confirming the pathological diagnosis. The treatment of patients was decided by their physician. The response to treatment was assessed following the Response Evaluation Criteria in Solid Tumours (RECIST). Patients who achieved complete response (CR) or a partial response (PR) after immunotherapy were classified into the immunotherapeutic R group. On the other hand, the NR group included patients who had progressive disease (PD) or stable disease (SD). A total of 16 immunotherapy-responsive NSCLC samples were collected and matched with 16 immunotherapy-unresponsive NSCLC samples with preimmunotherapy pathological tissue sections that were subjected to multiplexed immunofluorescence staining, and the results of their preimmunotherapy peripheral blood FCM assays were reviewed for subsequent statistical analysis. The Ethics Committee of the Second Affiliated Hospital of Chongqing Medical University granted the approval for the sample collection procedures in the study (No. 2021 − 547), and all patients’ written informed consent was obtained.

Peripheral blood mononuclear cell (PBMC) isolation
PBMCs were isolated via gradient centrifugation using Ficoll‒Paque Plus (17144002, Cytiva). Three millilitres of well-mixed whole blood was diluted with PBS at a ratio of 1:1. We pipetted three millilitres of Ficoll‒Paque gradient into 15 ml centrifuge tubes, and carefully layered the diluted blood over the Ficoll‒Paque gradient. The tubes then underwent 20 min of uninterrupted centrifugation at 800 × g. Subsequently, the isolated PBMCs were then stained with FCM antibodies.

Flow cytometry (FCM)
The cells were isolated from the peripheral blood or C57BL/6 mouse tumour tissues and subjected to live/dead cell staining (Fixable Viability Stain 780, BD Biosciences, 565388). Human TruStain FcX (BioLegend, 422301) or Mouse FcR Blocking Reagent (STARTER, S0B0599) was used to block nonspecific binding with Fc receptors before all surface staining. For cell membrane protein staining, the cells were stained at 4 °C for 30 min. For intracellular staining, the cells were fixed/permeabilized with a Fixation/Permeabilization Solution Kit (BD Biosciences) and then stained with antibodies against the indicated cytokines. After resuspension, data were collected using a BD FACSAria III and CytoFLEX and analysed under the assistance of FlowJo software (V.10.8.1). Supplementary Table 2 lists the specific antibodies for use in FCM.

Cell isolation and in vitro coculture
Antibody staining was followed by the sorting of the Pre-CCT T cells via a BD FACSAria III (BD Biosciences) to obtain T cells with greater than 90% purity. The sorted cells were cultured in OpTmizer T-Cell Expansion SFM (OpTmizer, Gibco). Pre-CCT T cells were randomly divided into different groups for coculture experiments. For the coculture experiments, 0.4 μm pore size transwell inserts in a 24-well plate were used. Pre-CCT T cells (5 × 104) were seeded into the upper well. Different macrophages were seeded into the lower wells containing 500 µL of RPMI 1640 medium. OpTmizer supplemented with either a CD44 inhibitor (1 × 10− 6 mM, AD 01, MCE) or a vehicle control and were used to treat the Pre-CCT T cells for 2 days. Pre-CCT T cells were collected for qPCR or cell immunofluorescence after 2 days of coculture with macrophages at a 1:1 ratio.

Multiplexed immunofluorescence staining
Formalin-fixed paraffin-embedded sections with a thickness of 5 μm were used for multiplexed immunofluorescence staining. After one night of incubation at 37 °C, the tissue sections underwent xylene dewaxing and ethanol rehydration at decreasing concentration in succession, followed by antigen retrieval in citrate buffer (pH 8.0). Nonspecific antigens were subjected to half an hour of blockage in 10% goat serum. The sections underwent one night of incubation using the primary antibody in blocking buffer at 4 °C, which, after PBST washing, received 50 min of polymer HRP-conjugated secondary antibody staining at room temperature (RT). Multiplexed immunofluorescence staining was performed on the tissue sections using the corresponding primary antibodies (Supplementary Table 3). The secondary antibodies and reagents used were HRP-conjugated anti-rabbit or anti-mouse antibodies and tyramine fluorescent dyes (TYR-480, TYR-520, TYR-570, TYR-620, and TYR-690). DAPI staining was applied to the nuclei. Finally, the slides were scanned with a KF-FL-020 scanner (KFBIO).

Cell line generation and culture
The Lewis lung carcinoma (LLC) cell line and Tohoku Hospital Pediatrics-1 (THP1) cell line were purchased from Pricella (Wuhan, China) and confirmed to be negative for mycoplasma contamination (Supplementary Fig. 2–3). The cells were cultured at 37 °C in a humidified atmosphere containing 5% CO2. Phorbol 12-myristate 13-acetate (PMA, 10 ng/ml, MCE) was added for 2 days to stimulate THP-1 cells to differentiate into macrophages. IL-4 and IL-13 (20 ng/mL, Gibco) were added for 2 days to induce the differentiation of M2 macrophages.

Viral transduction
For gene silencing, we prepared lentivirus particles through transfecting 293T cells using lentiviral vectors encoding specific shRNAs (shSPP1) or control shRNAs (shNC) along with the packaging plasmids PAX2 and PMD2, and then used the packaged lentiviruses to infect THP-1 cells and bone marrow-derived macrophages (BMDMs) for the knockdown. For infection, the lentiviral supernatant was added to cultured cells with 6 µg/mL polybrene (Beyotime, Shanghai, China). After incubation for 48 h, infected cells were selected for 5 to 7 days with puromycin (Beyotime, Shanghai, China).

Quantitative real-time PCR (qPCR)
An RNA-Quick Purification Kit (ES Science, #RN001) was employed to extract total RNA from Pre-CCT T cells, THP-1 cells, and BMDMs, and the isolated total RNA underwent reverse transcription by virtue of the PrimeScript RT reagent kit as per the producer’s protocol (TaKaRa, RRO37A). Quantitative RT‒PCR was proceeded via 2X SYBR Green qPCR Master Mix (No ROX) (MCE, HY-K0523). Supplementary Table 4 lists the primer sequences.

Western blotting
The harvested THP-1 cells and BMDMs underwent lysis treatment with RIPA buffer (Solarbio, #R0010) containing a protease and phosphatase inhibitor cocktail (Beyotime, P1045). A BCA kit (Beyotime, P0010S) was adopted for the quantification of protein concentration. Protein extracts underwent 10% SDS-PAGE separation, followed by being moved to 0.22 μm PVDF membranes (Millipore) to underwent TBS (Tris-buffered saline, containing 0.1% Tween 20 (TBST)) wash and 30 min of blockage in NcmBlot blocking buffer (NCM Biotech, P30500) in succession. The membranes underwent one night of incubation with specific primary antibodies at 4 °C, and another 1 h of incubation using HRP-conjugated secondary antibodies at RT. After washing, a Bio-Rad ChemiDoc MP Gel imaging system (Bio-Rad) was adopted for the detection of proteins in the membranes and ImageJ software (NIH, Bethesda, MD, USA) served for band intensity analysis. The Supplementary Table 5 lists the primary antibodies used for immunoblotting.

Cell immunofluorescence
Pre-CCT T cells were collected after coculture with macrophages. The cells were then fixed with 4% paraformaldehyde (PFA) for 15 min and incubated with primary antibodies against CD101 (rabbit, polyclonal, 26047-1-AP; Proteintech) overnight at 4 °C. Following washing with 1x PBS, the cells were incubated with secondary anti-rabbit CoraLite488 (goat, polyclonal, SA00013-2; Proteintech) for 90 min at RT. The sections were subsequently blocked with an anti-fluorescein quenching blocking agent (Solarbio, Cat: S2110). Scanning of the slides was conducted via a KF-FL-020 scanner (KFBIO).

Isolation and culture of BMDMs
BMDMs were prepared following a well-established protocol. Bone marrow cells were extracted from sample mice’ femur and tibia through fine dissection. After the lysis of red blood cells, the remaining bone marrow cells underwent 7 days of differentiation in DMEM supplemented with 10% FBS, 1% penicillin and streptomycin, and 10 ng/mL M-CSF (Gibco). IL-4 and IL-13 (20 ng/mL, Gibco) were employed to stimulate BMDMs for 48 h to make them further polarized to the M2 phenotype.

Mouse tumour model and in vivo research
All animal experiments adhered to protocols approved by the Animal Care and Use Committee of the Second Affiliated Hospital of Chongqing Medical University (IACUC-SAHCQMU-2024-00138). The mice were maintained in specific-pathogen-free (SPF) atmosphere in individually ventilated cages in the conditions of temperature of 21–22 °C, humidity of 39–50%, and a 12-h light‒dark cycles. All the animal experiments were carried out by using 6-8-week-old male C57BL/6 mice, which were injected subcutaneously with 1 × 106 LLC cells, with tumour growth and survival being monitored regularly. For PD-1 blockade therapy, samples were administered 200 µg of anti-mouse PD-1 antibody (clone RMP1-14 from MCE) or a rat IgG2a isotype control (clone 2A3 from BioXCell) i.p. every 3 days beginning on day 7 after tumour implantation. The tumour volume = 0.5×length×width2, where length and width are the longest and shortest diameter of the tumour, respectively. Twenty-one days after the subcutaneous tumourigenesis model was established, immune cells from the tumours of euthanized mice were arranged to be subjected to FCM analysis and multiplexed immunofluorescence staining.

Preparation of cell suspensions
The sectioned tumour tissues were digested in tissue dissociation solution (Absin). The centrifuge tube underwent 60 min of incubation on a constant temperature shaker at 37 °C/180 rpm to digest the tissue. The resulting cell suspensions were arranged to pass through a 70-µm nylon cell strainer for filtering for the preparation of single-cell suspensions, which were collected by centrifugation for subsequent experiments.

Statistical analysis
The statistical analysis of the bioinformatics data was performed in R software (v.4.1.2). Statistical analysis of the experimental data was performed with GraphPad Prism (v.10.1.2). Data that conformed to a normal distribution and homogeneity of variance are expressed as the means ± SDs. Pairwise comparisons were performed via unpaired two-tailed t tests. One-way ANOVA with Tukey’s multiple comparisons test and Dunn’s multiple comparisons test served for the comparison among ≥ 3 groups. Data failing to obey the normal distribution and homogeneity of variance are expressed as the median ± IQR. The Mann‒Whitney U test served for between-group comparisons, and the Kruskal‒Wallis test assisted in the multi-group comparison. P < 0.05 indicated statistical significance: *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001.

Results

Specific gene expression programmes determine the differentiation of Pre-CCT T cells into CCT T cells in NSCLC
The single-cell sequencing dataset GSE207422, consisting of tumour tissues from 15 NSCLC patients, was obtained from the GEO database. The samples included 3 samples from NSCLC patients before anti-PD-1 treatment and 12 samples after anti-PD-1 treatment (including 4 samples from patients with major pathologic response (MPR) and 8 samples from patients with nonmajor pathologic response (NMPR)). A final total of 92,330 cells were obtained for analysis by quality control and doublet exclusion criteria. Following dimensional reduction and visualization with UMAP, 28 cell clusters were identified (Fig. 1a-b). This was followed by annotation of the cellular identity of each cluster utilizing cell-specific biomarkers and a total of 10 cell types (Fig. 1c). Figure 1d illustrates the percentages of varying cell types in each sample. Dot plots visualize the marker genes of each cell type (Fig. 1e). The heatmap shows the expression of 5 characteristic genes in each cell population (Fig. 1f). To further determine the relationship between Pre-CCT T cells and CCT T cells, we reclustered T/NK cells and identified 7 subtypes (Fig. 2a-b). Next, according to the CD101 expression, we annotated CD8 + TIM3 + T cells as Pre-CCT T cells or CCT T cells (Fig. 2c-d, g-h). Myeloid cells were extracted, and macrophages and monocytes were identified by reclustering and annotation (Fig. 2e-f).

We constructed the pseudotime trajectory of CD8 + T cells. The normal differentiation paths were revealed through transcriptional states in the trajectory. Along the trajectory, CCT T cells were at the end of the developmental trajectory, and Pre-CCT T cells were located in separate branches of the trajectory (Fig. 2i-l). According to the pseudotime trajectory analysis results, the differentiation of CD8 + T cells accorded with a ‘Pre-CCT T cell-to-CCT T cell’ characteristic trajectory.
The reprogramming trajectory bifurcated into three branches and exhibited three different transcriptional states (Fig. 2i). We distinguished the molecular dynamics of the three branches to gain insight into the functional relevance of Pre-CCT and CCT T cells. The genes with a high expression in the prebranch stage were enriched in pathways, including T-cell activation and cytoplasmic translation. Genes enriched in pathways related to T-cell activation, lymphocyte differentiation, and mononuclear cell differentiation showed a high expression in cell fate2. Genes enriched in pathways associated with DNA replication, DNA-dependent DNA replication, and DNA recombination were highly expressed in cell fate1. These findings further supported the notion that Pre-CCT T cells can differentiate into CCT T cells (Fig. 2m). Supplementary Table 6 showed the genes and pathways enriched in pseudotime trajectory branches.

The differentiation fate of Pre-CCT T cells is heterogeneous in different response groups to immunotherapy
To characterize the impact of Pre-CCT T-cell terminal differentiation on the immunotherapeutic response, we compared patients in the MPR and NMPR groups in terms of the proportions of cell types (Fig. 3a). Deconvolution of the cell types from the single-cell analysis into the bulk-sequencing dataset (GSE135222) yielded consistent results (Fig. 3b). Specifically, MPR-group patients demonstrated a greater proportion of Pre-CCT T cells. To further verify this finding, pathological tissue sections at the initial diagnosis of 16 NSCLC patients in R group and 16 NSCLC patients in the NR group were subjected to multiplexed immunofluorescence staining. When CCT T cells were labelled according to the coexpression of CD8 and CD101, the results showed a greater abundance of CCT T-cell tumour infiltration in the NR group versus the R group (Fig. 3c-d). Moreover, the results of preimmunotherapy peripheral blood FCM assays were reviewed for statistical analysis. Also, a greater proportion of CCT T cells was associated with a poorer immunotherapeutic response (Fig. 3e-f; Table 1). The finding that Pre-CCT T cells differentiated into CCT T cells reflected the association of terminal differentiation of Pre-CCT T cells with a poorer response to immunotherapy.

The interaction between M2 macrophages and Pre-CCT T cells affects the immunotherapeutic response of NSCLC patients
Compared with that in the NMPR group, cell‒cell communication analysis revealed a decrease in the total number of interactions but an increase in interaction strength in the MPR group (Fig. 4a). These results suggested that complex cell crosstalk in the TME may influence the immunotherapeutic response in NSCLC. We further explored whether certain inhibitory cells in the TME affected the function and differentiation of Pre-CCT T cells. The network diagram revealed that among the various cells that interacted with Pre-CCT T cells, macrophages were the main signal providers (Fig. 4b). In addition, to characterize whether the interaction of macrophages with Pre-CCT T cells affected the immunotherapeutic response, we first clarified that the proportion of macrophages in the MPR and NMPR groups changed to the greatest extent, suggesting that macrophages in the TME may have a greater impact on the immunotherapeutic response than other types of cells (Fig. 4c). The specific signalling pairs between macrophages and Pre-CCT T cells were subsequently further analysed. Compared with that in the MPR group, the signal intensity of the SPP1‒CD44 pathway between macrophages and Pre-CCT T cells was significantly increased in the NMPR group (Fig. 4d). On these accounts, the interaction of macrophages with Pre-CCT T cells affected the immunotherapeutic response of NSCLC through the SPP1‒CD44 pathway.

To clarify which specific subtypes of macrophages interact with Pre-CCT T cells via the SPP1‒CD44 pathway and crucially influence the immunotherapeutic response, we extracted macrophages, and identified three classical macrophage subtypes, namely, M0, M1 and M2 macrophages, by clustering and annotation (Fig. 4e). The specific signalling pairs between the different macrophage subtypes and Pre-CCT T cells were subsequently analysed. Taken together, the interaction of M2 macrophages with Pre-CCT T cells via the SPP1‒CD44 pathway was most significantly enhanced in the NR group (Fig. 4f). In summary, we hypothesized that the interaction between M2 macrophages and Pre-CCT T cells affected the immunotherapeutic response through the SPP1‒CD44 pathway.

M2 macrophages promote the differentiation of Pre-CCT T cells into CCT T cells
We intended to further explore whether the interaction of M2 macrophages with Pre-CCT T cells could modulate the process of Pre-CCT T cells being differentiated into CCT T cells. WGCNA was applied to study the gene sets associated with Pre-CCT T cells and M2 macrophages. A total of 11 coexpression modules were identified (Fig. 5a-c). In the heatmap of module‒trait correlations, genes clustered in the green module the most strongly and positively related to Pre-CCT T cells as well as M2 macrophages (r = 0.8978, P < 0.05) (Fig. 5d). An obvious positive correlation was observed between the module membership (MM) of the green module and the gene significance (GS) of Pre-CCT T cells and M2 macrophages (cor = 0.87, p < 0.05) (Fig. 5e). Collectively, the most important (central) elements of the green module also tended to show a strong correlation with features of Pre-CCT T cells as well as M2 macrophages. Genes in the green module were then subjected to GO and KEGG enrichment analyses (Fig. 5f-g). The enriched signalling pathways directly or indirectly participated in the complex process of T-cell differentiation, involving the synergistic action of multiple signalling pathways, receptors and organelles. These findings supported the hypothesis that M2 macrophages promoted the differentiation of Pre-CCT T cells.

To validate this hypothesis, we performed in vitro coculture experiments. First, Pre-CCT T cells were obtained by flow cytometric sorting of PBMCs extracted from fresh peripheral blood. Then, we cocultured Pre-CCT T cells with M0 macrophages or M2 macrophages for 48 h according to the grouping (Fig. 5h). After coculture, Pre-CCT T cells were collected for qPCR and immunofluorescence staining experiments to detect CD101 expression. Compared with M0 macrophages, M2 macrophages significantly promoted CD101 expression on Pre-CCT T cells. Consistent results were obtained via immunofluorescence staining and qPCR (Fig. 5i-j). These findings suggested that M2 macrophages promoted Pre-CCT T cells to be differentiated into CCT T cells.

M2 macrophages affect the immunotherapeutic response of NSCLC patients by promoting the terminal differentiation of Pre-CCT T cells
On the basis of the distribution characteristics of varying CD8 + T-cell subtypes on the pseudotime trajectory, the three transcriptional states delineated by node 6 simulated the process of Pre-CCT T cells being differentiated into CCT T cells. We subsequently identified 8 hub genes participating in the promotion of Pre-CCT T-cell differentiation by M2 macrophages using genes characterized by node 6 on the pseudotime trajectory that intersected with genes related to the SPP1 signalling pathway in the communication between M2 macrophages and Pre-CCT T cells. The expression of these 8 hub genes showed an obvious difference between the NSCLC group and the control group, suggesting their important role in the TME of NSCLC (Fig. 6a).

We sought to further clarify whether the biological behaviour of M2 macrophages promoting the Pre-CCT T cell differentiation could impact the immunotherapy response. Using eight hub genes for unsupervised hierarchical cluster analysis, we identified two NSCLC subgroups (Fig. 6b). By comparing subgroup 1 and subgroup 2, a total of 444 DEGs between the two subgroups were identified (adjusted p value < 0.05, |Log2-fold change|>0.5). Volcano plots characterized all the DEGs (Fig. 6c). In addition, a heatmap was created for visualizing the expressions of the top-ranked genes (Fig. 6d). Biological functional differences between subgroups underwent GO and KEGG enrichment analyses, which demonstrated the enrichment of DEGs in multiple pathways (Fig. 6e-f). These pathways played important roles in T-cell differentiation by regulating redox status, extracellular matrix interactions, and metabolic pathways that affect T-cell activation, polarization, and function. Therefore, we defined subgroup 1 and subgroup 2 as M2-CCT-differentiated subgroup 1 and M2-CCT-differentiated subgroup 2, respectively.
We subsequently investigated differences in prognosis and immunotherapeutic response between the two M2-CCT-differentiated subgroups. Although the KM survival curves revealed no obviously different prognosis between the two subgroups (Fig. 6g), patients in M2-CCT-differentiated subgroup 1 had a worse immunotherapeutic response than did patients in M2-CCT-differentiated subgroup 2 (Fig. 6h). These findings supported our hypothesis that M2 macrophages drove the Pre-CCT T cell differentiation and could affect the immunotherapeutic response of NSCLC patients accordingly.
To investigate the importance of M2 macrophage regulation of Pre-CCT T-cell differentiation in NSCLC immunotherapy, we performed in vivo experiments. A subcutaneous tumour model was established in C57BL/6 mice by injecting LLC cells subcutaneously. Starting on day 7 posttumour implantation, the mice were treated with anti-mouse PD-1 antibody or a rat IgG2a isotype control. Simultaneously, we used clodronate liposomes (Clo-Lips) to eliminate bone marrow-derived M2-TAMs and to elucidate the role of M2 macrophages. The tumour-bearing mice were divided into 4 groups (n = 6): (1) the IgG + control liposome (Control-Lip) group; (2) the IgG + Clo-Lip group; (3) the anti-PD-1 + Control-Lip group; and (4) the anti-PD-1 + Clo-Lip group (Fig. 6i). The results of FCM demonstrated remarkably reduced M2-TAMs in the mice that received Clo-Lip versus those in the Control-Lip group, indicating that the Clo-Lip effectively removed M2-TAMs from the C57BL/6 mice (Fig. 6j-k; Table 2). Moreover, anti-PD-1 therapy could reduce the infiltration of M2-TAMs to some extent. We found that Clo-Lip or anti-PD-1 antibodies significantly restricted tumour growth and that Clo-Lip could promote the therapeutic response of patients to anti-PD-1 therapy (Fig. 6l-n). Besides, FCM and multiplexed immunofluorescence staining were used to detect the number and proportion of Pre-CCT and CCT T cells infiltrating the tumour tissue (Fig. 6o-s; Table 3). Accordingly, the IgG + Clo-Lip group exhibited remarkably lower abundance of tumour-infiltrating CCT T cells versus the IgG + Control-Lip group. Moreover, simultaneous treatment with both the anti-PD-1 agent and Clo-Lip further reduced the abundance of CCT T cells infiltrating the tumour. The results of in vivo experiments revealed that M2 macrophages may promote the differentiation of Pre-CCT into CCT T cells and that Clo-Lip inhibited the terminal differentiation of Pre-CCT T cells by lowering the number of M2 macrophages, thereby enhancing the response to immunotherapy in tumour-bearing mice.

M2 macrophages promote the Pre-CCT T cell differentiation into CCT T cells through the SPP1‒CD44 pathway
According to cell‒cell communication analysis of specific signalling pairs between M2 macrophages and Pre-CCT T cells, the interaction between M2 macrophages and Pre-CCT T cells affected the immunotherapeutic response of NSCLC through the SPP1‒CD44 pathway. Next, we performed single-gene GSEA, and ascertained the enrichment of genes with similar expression patterns to those of SPP1 and CD44 in pathways participating in T-cell signal transduction, migration and differentiation, namely KEGG_CYTOKINE_CYTOKINE_RECEPTOR_INTERACTION, KEGG_CELL_CYCLE, and KEGG_PENTOSE_PHOSPHATE_PATHWAY (Fig. 7a-b). Thus, the SPP1‒CD44 pathway is an important pathway through which M2 macrophages regulate the biological behaviour of Pre-CCT T cells, and SPP1 and CD44 also affect Pre-CCT T-cell differentiation.

To validate this hypothesis, we performed in vitro coculture experiments. Before the coculture experiments were performed, PBMCs extracted from fresh peripheral blood were sorted by FCM to obtain Pre-CCT T cells, and M2 macrophages with SPP1 knockdown were generated. We used 0.4 μm pore size transwell inserts in a 24-well plate and cocultured Pre-CCT T cells with M2 macrophages for 48 h under different conditions (M2 macrophages vs. shNC-M2 vs. shSPP1-M2 or CD44 inhibitor vs. vehicle control) according to grouping (Fig. 7c–f, i). After coculture, Pre-CCT T cells were collected for qPCR and immunofluorescence staining experiments to detect CD101 expression. After coculture, Pre-CCT T cells were collected for qPCR and immunofluorescence staining experiments to detect CD101 expression. According to the experiment results, the CD101 expression on Pre-CCT T cells was significantly attenuated after SPP1 was knocked down in M2 macrophages (Fig. 7g-h), and was also remarkably reduced after blocking CD44 of Pre-CCT T cells (Fig. 7j-k). The immunofluorescence staining results conformed to the qPCR results. In summary, these findings indicated that M2 macrophages relied on the SPP1‒CD44 pathway to facilitate Pre-CCT T-cell differentiation.

M2 macrophages promote the terminal differentiation of Pre-CCT T cells through the SPP1‒CD44 pathway to affect the immunotherapeutic response of NSCLC patients
The pathological tissue sections at the initial diagnosis of NSCLC patients in the R group (n = 16) and the NR group (n = 16) were subjected to multiplexed immunofluorescence staining (Fig. 8a-c). The results revealed the greater abundance of CD44-positive CCT T cell and SPP1-positive M2 macrophage tumour infiltration in the NR group versus the R group (Fig. 8d). Furthermore, correlation analysis uncovered an obvious positive association between the proportions of SPP1-positive M2 macrophages and CD44-positive CCT T cells in multiplexed immunofluorescence-stained sections (Fig. 8e). These results suggested the presence of an active interaction between M2 macrophages and CCT T cells through the SPP1‒CD44 pathway in the TME and that the SPP1‒CD44 pathway may benefit the promoting role of M2 macrophages in Pre-CCT T-cell differentiation, as well as suggesting that this biological behaviour influenced the immunotherapeutic response of NSCLC.

Next, we validated the role of the SPP1‒CD44 pathway in the promotion of Pre-CCT T-cell differentiation by M2 macrophages by blocking the SPP1‒CD44 pathway in vivo. BMDMs were extracted from C57BL/6 mice, and M2-BMDMs in which SPP1 was knocked down (shSPP1-M2-BMDMs) were constructed for subsequent in vivo experiments (Fig. 8f-h). A subcutaneous tumour model was established by subcutaneously inoculating C57BL/6 mice with shNC-M2-BMDMs or shSPP1-M2-BMDMs and LLC cells at a 1:1 ratio. Starting on day 7 posttumour implantation, the mice were treated with anti-mouse PD-1 antibody or a rat IgG2a isotype control. The tumour-bearing mice were divided into 4 groups (n = 6): 1) the IgG + shNC-M2-BMDMs group; 2) the IgG + shSPP1-M2-BMDMs group; 3) the anti-PD-1 + shNC-M2-BMDMs group; and 4) the anti-PD-1 + shSPP1-M2-BMDMs group (Fig. 8i). The results revealed the remarkably lower tumour volume and weight in anti-PD-1 group versus those receiving isotype control IgG. Relative to the shNC-M2-BMDMs group, the growth of tumours in the shSPP1-M2-BMDM group was significantly inhibited, and the therapeutic response to anti-PD-1 was improved (Fig. 8j-l). Single-cell suspensions prepared from tumour-bearing mice’ tumour tissues were analysed via FCM, and the results revealed no obvious difference between the shSPP1-M2-BMDMs group and the shNC-M2-BMDMs group in the number of tumour-infiltrating M2-TAMs (Fig. 8m-n; Table 4). These findings indicated that differences in the numbers of tumour-infiltrating T-cell subsets in the shSPP1-M2-BMDMs and shNC-M2-BMDMs groups in the in vivo experiments were not affected by the number of M2-TAMs. FCM analysis and multiplexed immunofluorescence staining of tumour tissues yielded consistent results. The IgG + shSPP1-M2-BMDMs and anti-PD-1 + shNC-M2-BMDMs groups presented dramatically lower abundance of tumour-infiltrating CCT T cells versus the IgG + shNC-M2-BMDMs group. Moreover, mice in the shSPP1-M2-BMDMs group receiving anti-PD-1 treatment could further reduce the abundance of tumour-infiltrating CCT T cells (Fig. 8o-r; Table 5). Taken together, SPP1 knockdown in tumour-infiltrating M2 macrophages effectively blocked the SPP1‒CD44 pathway, thereby inhibiting the promotion of Pre-CCT T-cell differentiation by M2 macrophages.

With the objective of more deeply confirming the significance of the SPP1‒CD44 pathway, we used intraperitoneal injection of CD44 inhibitors to achieve in vivo blockade of CD44. A subcutaneous tumour model was established in C57BL/6 mice by injecting LLC cells subcutaneously. Starting on day 7 posttumour implantation, the mice were treated with anti-mouse PD-1 antibody or a rat IgG2a isotype control. The tumour-bearing mice were divided into 4 groups (n = 6): (1) the IgG + vehicle control group; (2) the IgG + CD44 inhibitor group; (3) the anti-PD-1 + vehicle control group; and (4) the anti-PD-1 + CD44 inhibitor group (Fig. 9a). Relative to the vehicle control group, the CD44 inhibitor group presented significant inhibition of tumour growth and a better therapeutic response to anti-PD-1 therapy (Fig. 9b-d). The FCM results revealed no obvious difference in the number of tumour-infiltrating M2-TAMs between the CD44 inhibitor and vehicle control groups (Fig. 9e-f; Table 6). These findings indicated that the differences in tumour-infiltrating T-cell subsets in the CD44 inhibitor and vehicle control groups in the in vivo experiments were not affected by the number of M2-TAMs. FCM analysis and multiplexed immunofluorescence staining of tumour tissues yielded consistent results. The IgG + CD44 inhibitor group and the anti-PD-1 + vehicle control group exhibited remarkably lower number of tumour-infiltrating CCT T cells versus the IgG + vehicle control group. Moreover, mice in the CD44 inhibitor group receiving anti-PD-1 treatment could further reduce the abundance of tumour infiltration of CCT T cells (Fig. 9g-j; Table 7). Collectively, the CD44 inhibitor could well block the SPP1‒CD44 pathway, thereby inhibiting the promotion of Pre-CCT T-cell differentiation by M2 macrophages. On these accounts, M2 macrophages excel in promoting the differentiation of Pre-CCT T cells by virtue of the SPP1‒CD44 pathway. Moreover, blocking the SPP1‒CD44 pathway while receiving anti-PD-1 treatment inhibited Pre-CCT T-cell differentiation to a greater extent, which potentially explained why blocking the SPP1‒CD44 pathway promoted the response to anti-PD-1 therapy in tumour-bearing mice.

Discussion
The tumour immunosuppressive microenvironment (TISM) poses a great challenge for the treatment of NSCLC with immunotherapies such as anti-PD-1. This study is the first to conduct a comprehensive bioinformatics analysis of scRNA-seq and bulk RNA-seq data to investigate the following primary hypothesis: CCT T cells differentiate from Pre-CCT T cells via the M2 macrophage-mediated SPP1–CD44 pathway, which influences the immune response to NSCLC treatment. Subsequent analyses of peripheral blood FCM and tissue multiplexed immunofluorescence staining in NSCLC samples provided clinical evidence to support the scientific hypothesis. Finally, the above hypotheses were further validated through in vitro coculture experiments and in vivo experiments in a subcutaneous tumour model in C57BL/6 mice. Notably, Pre-CCT T-cell terminal differentiation is mediated by M2 TAMs through the SPP1‒CD44 pathway. These findings may contribute to the identification of potential treatment targets for NSCLC.
At present, ICIs are widely used to treat NSCLC [36–38]. Anti-PD-1 drugs are capable of releasing T cells from a dysfunctional state, i.e. T-cell exhaustion, and rescuing the cytotoxicity of CD8 + T cells. Although ICI therapy develops fast and achieves an obvious efficacy, immunotherapy response rates remain low, representing a major clinical challenge. T-cell exhaustion is a broad term that is used to describe the response of T cells to chronic antigenic stimulation [6]. T-cell exhaustion is also a differentiated state observed under chronic antigen exposure, in which some cells retain their proliferative and effector functions despite elevated expression of multiple immunosuppressive signals of PD-1 and TIM3 [39–42]. In recent studies, the confirmation of progenitor and terminally differentiated subpopulations of exhausted T cells as well as their importance in tumour patients’ response to anti-PD-1 therapy have been highlighted [43, 44]. In the study by Hudson et al., stem-like Tcf-1 + CD8 + T cells underwent a transition from CD101-TIM3 + T cells before eventually differentiating into CD101 + Tim3 + T cells. CD101-Tim3 + T cells are defined as newly generated, transitional cells that proliferate in vivo and are characterized by effector-like transcriptional features [11]. Leung et al. performed high-dimensional CyTOF and MSD multiplex cytokine longitudinal analyses of PBMCs from 25 Chinese patients with NSCLC treated with ICIs, confirming the presence of massive dysfunctional CCT T cells in nonresponders treated with ICIs [12]. On the basis of the current findings, considering the pivotal role of Pre-CCT T cells in anti-PD-1 immunotherapy and the close relationship between CCT T cells and the immunotherapeutic response, the present study provides further insights into Pre-CCT and CCT T cells. In this study, the trajectory from Pre-CCT T cells to CCT T cells was characterized, and the gene set that participates in the terminal differentiation of Pre-CCT T cells was identified. We further performed multiplexed immunofluorescence staining and peripheral blood FCM analyses of NSCLC tissue samples and peripheral blood samples collected at initial diagnosis, and the results demonstrated the association of a greater proportion of CCT T cells with a poorer immunotherapeutic response. Considering the possibility that Pre-CCT T cells differentiate into CCT T cells, the terminal differentiation of Pre-CCT T cells may contribute to the unresponsiveness of NSCLC to immunotherapy.
Previous researches have paid attention to characterizing TME remodelling following ICI treatment, suggesting that treatment with ICIs counteracts the dysfunction or exhaustion of T-cells and facilitates their clonal expansion [45]; simultaneously, the TME can influence the immunotherapeutic response [13]. The innate and adaptive immune systems being dysfunctionally interacted can promote tumour evasion from the immune system through the generation of the TISM [46, 47]. According to the cell‒cell communication analysis results, the strength of interactions between varying cell types differed significantly between the NMPR and MPR groups, which was consistent with the findings of previous studies, suggesting that complex cellular crosstalk in the TME may influence the immunotherapeutic response in NSCLC. Therefore, the presence of immunosuppressive cells in the TISM may critically help to modulate the terminal differentiation of Pre-CCT T cells, thus affecting the immunotherapeutic response.
Consequently, we performed further analyses, which revealed that among the various immune cells that interacted with Pre-CCT T cells, macrophages were the major signalling providers. Additionally, their proportions in different immunotherapy response groups were significantly different, with M2 macrophages being the most significantly different among the different immunotherapy response groups in terms of the key pathways interacting with Pre-CCT T cells. According to previous studies, TAMs constitute the main population of TISM, which nourishes tumour cells as well as benefits TISM, e.g. promoting cytotoxic CD8 + T cells to be releases and immunosuppressive cells (MDSCs and Tregs) to be recruited [48]. Notably, high infiltration levels of TAMs in diagnosed cancers indicate a nonresponse to anti-PD-1/PD-L1 therapy [18, 49–53], and are considered important targets for reversing the nonresponse to anti-PD-1/PD-L1 therapy [54, 55]. Among them, M2-TAMs can remarkably regulate tumour progression, angiogenesis, metastasis, nonresponse to therapy, and poor clinical outcomes [13, 19, 20, 56–59]. Therefore, we hypothesized that M2-TAMs well regulated the terminal differentiation of Pre-CCT T cells and thus affected the immunotherapeutic response. This study was validated in primary Pre-CCT T cells and revealed that coculture with M2 macrophages significantly increased CD101 expression in Pre-CCT T cells. Moreover, removing bone marrow-derived M2-TAMs significantly reduced the proportion of CCT T cells, restricted tumour growth in a subcutaneous tumour model in C57BL/6 mice, and enhanced the response to anti-PD-1 treatment. Thus, combining the in vitro and in vivo experiments, M2 macrophages facilitated Pre-CCT T-cell terminal differentiation, leading to an immunotherapeutic nonresponse. Notably, existing T lymphocyte cell lines were not available for the study of this specific effector T-cell subset, and it is difficult to obtain Pre-CCT T cells, a subpopulation of exhausted T cells, even from healthy volunteers. Therefore, in the present study, fresh peripheral blood from NSCLC patients was collected to extract PBMCs, and Pre-CCT T cells were obtained via a flow sorting technique for subsequent coculture experiments. Compared with T lymphocyte cell lines, primary Pre-CCT T cells avoid genetic and functional alterations resulting from long-term culture, and the results are more reliable. Moreover, these cells retain their natural properties, resulting in a more realistic reflection of the immune response in vivo. Furthermore, the experimental results can be more easily translated into clinical applications, which is especially valuable in immunotherapy.
Although M2-TAMs are involved in tumour immune escape and nonresponsive to ICI immunotherapy through multiple mechanisms, the potential mechanism through which M2-TAMs promote terminal differentiation of Pre-CCT T cells, a specific subtype of T cells, is not clear. Therefore, the potential mechanisms of this biological behaviour were further explored in this study. On the basis of analysis of specific signalling pairs between the two cell types and multiplexed immunofluorescence staining of tumour tissue samples from NSCLC patients, we determined that the SPP1–CD44 pathway may be the pathway through which M2 macrophages facilitate Pre-CCT T-cell differentiation to influence the response of NSCLC to immunotherapy. Next, we blocked the SPP1–CD44 pathway by knocking down SPP1 in macrophages and adding CD44 inhibitors. Blockade of the SPP1–CD44 pathway significantly reduced CD101 expression in Pre-CCT T cells in coculture experiments. In addition, it markedly inhibited tumour growth in C57BL/6 mice and promoted the response to anti-PD-1 therapy in vivo. Thus, the SPP1‒CD44 pathway was confirmed as a potential mechanism by which M2 macrophages promoted the differentiation of Pre-CCT T cells, confirming the SPP1‒CD44 pathway as a valuable target for enhancing the response of NSCLC to ICI immunotherapy. At present, some studies have focused on macrophages with high expression of SPP1, and Macro_SPP1 macrophages greatly impact the tumour angiogenesis [60] and facilitate immune escape by upregulating PD-L1 [61]. SPP1hi-TAMs have been identified in varying cancer types [60, 62, 63] and can drive the immunotherapeutic resistance [64]. In addition, a decreasing trend of SPP1 + TAMs in the TME has also been observed in NSCLC patients who responded well to immunotherapy [65]. While SPP1hi-TAMs have also been identified in these studies, and their contribution to tumour progression and immunotherapeutic nonresponsiveness has been elucidated in numerous ways, the present study highlights the first demonstration of how M2 macrophages affect NSCLC patient response to immunotherapy by promoting the differentiation of a specific T-cell subset, Pre-CCT T cells, through the SPP1‒CD44 axis. The present study focused on the immunosuppressive TME elements represented by M2 TAMs and their modulatory effects on more refined T-cell subsets. Currently, with the high clinical demand for adjuvant treatments that can increase the efficacy of anti-PD-1 therapy, intervention with exhausted T-cell subsets that are not terminally differentiated has potential therapeutic value and may provide a promising new strategy to improve the effect of immunotherapy against NSCLC in clinical practice. The discovery of the SPP1–CD44 axis may also aid in the identification of potential targets for combination immunotherapy in NSCLC. The heterogeneity of the TME increases the difficulty in predicting which patients can benefit from immunotherapy [66], and further studies should be conducted to identify effective predictors of response rates in patients with NSCLC [67–70]. In contrast to biopsy, evaluation of the proportions of Pre-CCT T cells and CCT T cells in the peripheral blood of NSCLC patients offers a feasible noninvasive approach for the early prediction of immunotherapy response. For future clinical applications, we recommend integrating our findings with widely used clinical parameters to provide more accurate information for predicting immunotherapy efficacy.
Interestingly, according to flow cytometric analysis of tumour tissues from tumour-bearing mice, tumour tissues from anti-PD-1-treated mice demonstrated lower proportion of M2-TAMs versus those from mice receiving isotype control IgG. This phenomenon might be the result of the correlation of PD-1 expression on macrophages with their polarization. According to previous studies, PD-1 overexpression on macrophages can downregulate the expression of M1-related markers (e.g., iNOS), upregulate the expression of M2-related markers (e.g., Arg-1) and reduce their release of cytokines (IL-1b and IL-12) [71], whereas anti-PD-1 treatment can inhibit M2 polarization of macrophages. Anti-PD-1 therapy achieves antitumour effects through multiple actions on various cells in the TME. Given the heterogeneous TME and complex mechanism of anti-PD-1 therapy, exploring strategies to improve the response to immunotherapy from a multifaceted and multidimensional perspective is highly important.
However, notably, for this study, we utilized datasets from only public databases for bioinformatics analysis, and in-house data and more diverse datasets could be combined in the future to further validate and extend the findings of this study. In addition, further studies are needed to analyse longitudinal changes before and after immunotherapy to characterize these T-cell subtypes and their function and dynamics in NSCLC.
Taken together, M2 macrophages promote the terminal differentiation of Pre-CCT T cells through the SPP1‒CD44 pathway, thereby affecting the immunotherapeutic response of NSCLC. The finding in the study theoretically elucidates the function of TISM in the immunotherapeutic nonresponse and assists in well comprehending the interactions between TAMs and T cells.

Supplementary Information
Below is the link to the electronic supplementary material.