CST7⁺ macrophages/monocytes in melanoma: single-cell insights into immunotherapy response

Introduction

The incidence of melanoma, an aggressively malignant tumor that mainly arises from melanocytes in the skin, increases annually despite it constituting only a small proportion all solid tumor cases (1). Prior to the advent of immunotherapy, clinical trials reported the poor prognosis of patients with late-stage tumors, and these patients often faced limited therapeutic options, with a median survival less than 1 year (2,3). Recent advances in the development of immunotherapy have demonstrated promising clinical activity across a wide variety of cancers. In particular, the survival of patients with advanced melanoma, a typical immune-responsive solid tumor, can be significantly improved by immune checkpoint inhibitor (ICI) treatment (4). Given the encouraging efficacy of ICIs in melanoma, approvals for antibodies against cytotoxic T-lymphocyte associated protein 4 (CTLA-4), programmed cell death protein 1 (PD-1), and programmed death-ligand 1 (PD-L1) were obtained between 2011 and 2017 (5).

Although immunotherapy has revolutionized the clinical management of metastatic diseases, approximately one-half of patients with melanoma do not experience long-term immunogenic response, suggesting an ongoing need of improved therapeutic strategies (6,7). There is also a lack of highly sensitive biomarkers to identify patients that will most likely benefit from ICI treatment. Studies investigating the immunological and molecular characteristics of melanoma have identified several prognostic markers for immunotherapy (8-10). Although some research suggests that PD-L1 can serve as an effective indicator for ICI treatment, its stratification accuracy is highly influenced by the cutoff of PD-L1 expression (11-15). It has also been shown that a substantial proportion of PD-L1-negative patients still respond well to immunotherapy (16). Studies employing multiomics have also identified several other molecular features, such as high tumor mutation burden (17-19) microsatellite instability-high status (20,21), and a specific gut microbiome profile (22,23) that have implicated in the immune responsiveness of patients with melanoma. However, due to the complex nature of tumorigenesis and tumor progression, these existing biomarkers often fail to fully address the fully diversity of clinical needs.

The tumor immune microenvironment (TIME) plays a significant role in regulating immune response. In melanoma, various biochemical and cellular components in the TIME contribute to either therapeutic sensitivity or resistance (24,25). For instance, sustained low levels of interferon-γ (IFN-γ) promote the upregulation of immunosuppressive molecules, including PD-L1, CTLA-4, and Indoleamine 2,3-dioxygenase 1 (IDO1), on melanoma cells, thereby enhancing immune evasion and contributing to ICI resistance (26). Tumor-associated macrophages (TAMs) and other immunosuppressive cells further diminish IFN-γ activity within the tumor microenvironment, impairing ICI efficacy. In metastatic melanoma, resistance to anti-PD-1 monotherapy is associated with an elevated expression of LAG-3 in peripheral blood lymphocytes, while improved therapeutic response and survival correlate with a higher proportion of unstable regulatory T cells (Tregs) (27). Additionally, MerTK (MER proto-oncogene, tyrosine kinase)-positive macrophages promote melanoma progression and immunotherapy resistance through the activation of the AhR.ALKAL1 pathway (28), while a distinct macrophage subset drives targeted therapy resistance via POSTN-mediated signaling and protection of melanoma cells (29). Furthermore, myeloid-derived suppressor cells contribute to immunotherapy resistance by suppressing antitumor immune responses and serve as indicators of poor prognosis (30). Despite the availability of these TIME biomarkers for the prognostic assessment of patients with melanoma, the mechanistic impact of these biomarkers remains to be elucidated.

The recent advancement of single-cell sequencing offers an opportunity to discover high-quality biomarkers and clinically relevant indicators of immunotherapy efficacy in melanoma and to identify novel therapeutic targets. To this end, we incorporated publicly available melanoma RNA-sequencing (RNA-seq) datasets at the both bulk and single-cell levels. We first employed single-sample gene set enrichment analysis (ssGSEA) to distinguish melanoma tumors with varying degrees of immune infiltration. Differentially expressed genes (DEGs) between the high- and low-immune-infiltration groups underwent downstream least absolute shrinkage and selection operator (LASSO) Cox regression analysis, which identified nine core genes with prognostic potential. Based on the signatures of these genes, we successfully established a prognostic model, and its performance was validated through external cohorts. Subsequently, the single-cell RNA-sequencing (scRNA-seq) data from patients with melanoma, along with their responsiveness to ICIs, was used to investigate the expression patterns of these prognostic genes in select immune populations and their associations with therapeutic outcomes. This led to the discovery of a CST7-expressing myeloid population that may contribute to treatment response. Our findings can further inform molecular classification and prognostic prediction and provide guidance in the development of combination therapies for patients with immune-resistant melanoma. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1943/rc).

Methods

Data source

A total of three independent melanoma datasets were used in the analysis: one included 189 skin cutaneous melanoma cases from The Cancer Genome Atlas (TCGA; dataset code: TCGA-SKCM; https://portal.gdc.cancer.gov/), one was a published dataset (Liu_Data; https://github.com/vanallenlab/schadendorf-pd1) (31) that included 102 cases, and one was GSE91061, available in the Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/) database, which included 51 cases. In the systematic analysis, the TCGA-SKCM, Liu_Data, and GSE91061 datasets served as the training cohort, internal validation cohort, and external validation cohort, respectively. To further validate the robustness of findings in melanoma, we used an additional seven multi-cancer cohorts from TCGA database: bladder carcinoma (BLCA; N=203), breast invasive carcinoma (BRCA; N=107), head and neck squamous cell carcinoma (HNSC; N=116), kidney renal papillary cell carcinoma (KIRP; N=173), liver hepatocellular carcinoma (LIHC; N=263), lung adenocarcinoma (LUAD), and lung squamous cell carcinoma (LUSC; N=257). Additionally, the scRNA-seq dataset GSE120575 (32) was retrieved from GEO database, containing 10,363 immune cells of 29 tumor tissues from patients with melanoma with treatment histories of checkpoint inhibitors. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Clustering of melanoma data via immune signatures

Using 29 immune-related gene sets (33), we calculated immune scores in samples from TCGA-SKCM and Liu_Data via the ssGSEA method through the “GSVA” (v.1.48.2) package in R (The R Foundation for Statistical Computing). Through the K-means clustering algorithm, we classified samples into two clusters: the low-immune-infiltration group (LIG) and the high-immune-infiltration group (HIG). The R package “factoextra” (v. 1.0.7) was employed to visualize the results of sample clustering. The abundance of infiltrating immune cells and stromal cells was determined via the ESTIMATE algorithm based on cells marker genes, and results were visualized with the R package “ggpubr” (v. 0.6.0). CIBERSORT (https://cibersort.stanford.edu/) was applied to estimate the proportions of 22 immune cell types, and the ratios of these subpopulations between HIG and LIG were compared via the Wilcoxon Mann-Whitney test. Furthermore, the expressions of human leukocyte antigen (HLA) genes were profiled with the R package “ggpubr” (v. 0.6.0).

Potential molecules and pathways involved in immune response

The DEGs between the HIG and LIG were determined via the Student t-test (P≤0.05) and a threshold of |log2 fold change| ≥0.58. To screen for the potentially prognostic molecules and to identify molecules with potential prognostic signatures, these DEGs underwent univariate Cox regression analysis via the R “survival” package (v. 3.5.5). In addition, prognostic DEGs shared by TCGA-SKCM and Liu_Data cohorts were extracted to characterize their downstream effects via Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses with the R package “clusterProfiler” (v. 4.8.1). A protein-protein interaction (PPI) network was constructed based on the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (https://string-db.org).

Construction and validation of a robust prognostic signature

LASSO regression analysis was conducted via the R package “glmnet” (v. 4.1.7) to screen for prognostic predictors. Ten-fold cross-validation was employed to determine the optimal tuning parameters to avoid overfitting, which was followed by multivariate Cox regression analysis to further identify the hub genes. A risk score based on the expression levels and corresponding coefficients of these genes was calculated with the following equation: risk score = 0.8903 × CST7 + 0.9324 × CSF2RB + 0.7497 × SPOCK2 + 1.1917 × PLA2G2D + 1.7198 × MYO1G + 0.5561 × NCF1 + 0.8860 × ASGR2 + 1.1036 × ICOS + 0.9741 × HSH2D. Using the R package “survminer” (v. 0.4.9), we confirmed the optimal cutoff value for the risk score, classifying patients into high- or low-risk groups. The R packages of “survival” (v. 3.5.5) and “survminer (v. 0.4.9)” were used to perform Kaplan-Meier survival analysis to compare the survival status between patients with high and low risk. Meanwhile, the candidate clinical variables were examined via multivariate Cox regression analyses. To further investigate the prognostic associations of important factors, including tumor stage, gender, age, and the risk score, we established a nomogram using the R packages “rms” (v. 6.7.1) and “survival” (v. 3.5.5).

Dimensional reduction and clustering for scRNA-seq data

The gene expression matrices of scRNA-seq data from the GSE120575 dataset were retrieved. The “Seurat” (v. 4.3.0) package (34) in R (v. 4.2.2) was used to filter out cells with fewer than 200 detected genes or those expressing more than 10% of mitochondrial genes; the “RunPCA” function was used to conduct principal component analysis (PCA) on genes identified as being highly variable with the “FindVariableFeatures” function. The top 13 principal components were used as input for “FindNeighbors” function to construct a shared nearest-neighbor graph, and the “FindClusters” function (resolution =0.7) was used to identify cell clusters. Furthermore, the top 13 principal components were selected to further reduce dimensionality via the uniform manifold approximation and projection (UMAP) method.

Annotation of cell types

To identify marker genes with differential expressions across cell types, we employed the “RunPrestoAll” function in the “Presto” package (v. 1.0.0) under default parameters. Genes that had a mean log fold change >0.25 and were expressed in over 25% cells in a given cluster were selected as marker genes. Manual cell type annotation was conducted based on previously published reference cell markers (32). Dot plots showing averaged expression levels of marker genes were implemented with the “ggplot2” package (v. 4.3.0) in R (v. 4.2.2).

Gene signature analysis

To characterize the function differences between the various clusters, the R package “clusterProfiler” (v. 4.6.2) was used to perform gene set enrichment analysis (GSEA) on several sets of genes, including cancer-associated hallmarks, M1 signature, and M2 signature. Pearson correlation coefficients were calculated to evaluate the relationships between CST7 (cystatin F) expression levels and the score of an M1 and M2 signature, with the results being visualized via the R package “ggplot2” (v. 3.5.0).

Cell-cell communication analysis

Cell-cell communication analysis was performed via the R package “CellChat” (v. 1.6.1) (35). Genes encoding ligands and receptors expressed by each cell were projected onto a manually selected reference communication network, and the probability of communication in each pathway was inferred by gene expression pattern. We statistically analyzed the communication probability among subclusters. Dot plots and circle plots were generated via the “ggplot2” (v. 3.5.0) and “CellChat” (v. 1.6.1) packages in R.

Statistical analysis

The Wilcoxon rank-sum test and Student t-test were employed for statistical comparisons of gene expression levels, infiltrating-immune cell proportions, clinical outcomes, and other relevant variables between groups. The statistical significance level was set at P<0.05.

Results

Sample clustering and stratification of patients with melanoma

The overview of this study is shown in Figure 1. Transcriptome and survival data from 189 skin cutaneous melanoma cases in TCGA (TCGA-SKCM) and 102 patients with melanoma from a published dataset (Liu_Data for short) (31) were initially collected. The ssGSEA signature scores of 29 immune-related hallmarks were calculated for 189 tumor samples of TCGA-SKCM. K-means clustering based on these scores partitioned samples into two distinct clusters: cluster 1 (N=111) and cluster 2 (N=78) (Figure 2A). Cluster 1 samples showed low stromal scores, immune scores, and ESTIMATE scores and high tumor purity scores, indicative of low immune infiltration; meanwhile, cluster 2 samples exhibited the opposite pattern. Therefore, cluster 1 and cluster 2 were defined as the LIG and HIG, respectively (Figure 2B). The expression levels of the 29 immune-related hallmarks were higher in the HIG than in the LIG, in line with their classification (Figure 2C). To deconvolute the degree of immune infiltration at the cellular level, the CIBERSORT algorithm was used to estimate immune cell frequencies in both groups. The HIG had elevated infiltration of various antitumor immune lineages, such as activated natural killer (NK) cells, activated memory CD4⁺ T cells, and CD8⁺ T cells, suggesting an active immune response. Conversely, the LIG were enriched for M2 macrophages, resting NK cells, and resting memory CD4 T cells, indicative of an immune-suppressive phenotype (Figure 2D). In line with these observations, the expression levels of several HLA genes in the HIG group were significantly upregulated as compared to the LIG (Figure 2E). To confirm the reliability of the identified clusters, a second melanoma dataset with 102 samples (Liu_Data) was analyzed in a similar manner (31). These samples were also clustered into a HIG (N=48) and a LIG (N=54) (Figure S1A). Characterizations of the HIG and LIG from this cohort led to similar molecular characteristic findings in terms of stromal score, immune score, ESTIMATE score, tumor purity score (Figure S1B,S1C), immune cell infiltration (Figure S1D), and HLA gene expression (Figure S1E).

Figure 1 Overview of the study (created with BioRender.com). BLCA, bladder carcinoma; BRCA, breast invasive carcinoma; DEGs, differentially expressed genes; HNSC, head and neck squamous cell carcinoma; ICOS, inducible T-cell costimulator; ICOSL, inducible T-cell costimulator ligand; KIRP, kidney renal papillary cell carcinoma; LASSO, least absolute shrinkage and selection operator; LIHC, liver hepatocellular carcinoma; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; NES, normalized enrichment score; scRNA-seq, single-cell RNA sequencing; SKCM, skin cutaneous melanoma; ssGSEA, single-sample gene set enrichment analysis; TCGA, The Cancer Genome Atlas.

Figure 2 Immune landscape characterization of lung adenocarcinoma samples. (A) Cluster 1 (LIG) and cluster 2 (HIG) divided by K-means clustering based on signatures of 29 immune-related hallmarks; (B) estimation of stromal score, immune score, ESTIMATE score, and tumor purity score; (C) landscapes of 29 immune-related hallmarks in the two groups; (D) the infiltration levels of 22 immune cells between the HIG and LIG samples; (E) expression profiles of HLA genes. *, P<0.05; ***, P<0.001. HIG, high-immune-infiltration group; LIG, low-immune-infiltration group; NK, natural killer.

Biological implications of DEGs

To identify signatures specific for the two melanoma clusters, we performed differential gene expression analysis between the two groups. A total of 1,556 DEGs and 1,445 DEGs were identified in TCGA-SKCM and Liu_Data datasets, respectively (Figure 3A, Figure S2A). In TCGA-SKCM dataset, there were 1,190 upregulated genes in the HIG and 366 in the LIG. Meanwhile, the Liu_Data had 1,352 highly expressed genes in the HIG and 93 in the LIG. Univariate Cox regression analysis of these DEGs indicated that 677 and 147 DEGs were significantly associated with the prognosis in TCGA-SKCM and Liu_Data, respectively (P≤0.05) (Figure 3B, Figure S2B). Among these genes, only 59 DEGs were shared by both groups (Table S1), and functional enrichment analyses were subsequently carried out. To clarify their biological roles, GO analysis on these 59 genes was performed, revealing a notable association with immune-related functions. The GO terms related to immune response, such as “T cell activation”, “regulation of T cell activation”, “regulation of leukocyte cell-cell adhesion”, “MHC protein complex”, and “immune receptor activity”, were significantly enriched (Figure 3C, Table S2). KEGG analysis also revealed the involvement of these DEGs in pathways mediating immune response, with the top pathway terms including “T cell receptor signaling pathway”, “Th17 cell differentiation”, “Th1 and Th2 cell differentiation”, “PD-L1 expression and PD-1 checkpoint pathway in cancer”, and “intestinal immune network for IgA production” (Figure 3D, Table S3). The PPI network analysis revealed a close association among several members of the 59 hub DEGs, including ITGB2, PLEK, FGR, WDFY4, HCK, and NCF4, indicating that they may collectively regulate the immune response during melanoma development (Figure S2C).

Figure 3 Differential expression and enrichment analysis of DEGs. (A) The volcano plots and (B) hazard ratios of DEGs. (C) GO terms and (D) KEGG pathways enriched by 59 hub DEGs. BP, biological process; CC, cellular component; DEGs, differentially expressed genes; GO, Gene Ontology; HR, hazard ratio; KEGG, Kyoto Encyclopedia of Genes and Genomes; MF, molecular function; not_sig, not significant.

Construction and validation of risk model based on hub immune-related genes

Having selected the 59 DEGs, we performed LASSO regression analysis to further screen for hub genes to develop a prognostic model for risk stratification. With the best λ value and output coefficients, nine core genes were retained for downstream analysis, including CST7, CSF2RB, SPOCK2, PLA2G2D, MYO1G, NCF1, ASGR2, ICOS, and HSH2D (Figure 4A,4B). A model was constructed based on these nine genes to calculate the risk score of each patient. Kaplan-Meier survival analysis showed that patients with a high risk score had a significantly shorter overall survival (OS) as compared to those with a low risk score in both TCGA-SKCM (Figure 4C) and Liu_Data cohorts (Figure 4D).

Figure 4 Construction and validation of the LASSO-based prognostic model. (A) LASSO coefficient characteristics of the variables; (B) the optimum number of the parameters λ in the LASSO regression model as determined by the 10-fold cross-validation method; (C) Kaplan-Meier survival curves of patients with high- and low-risk based on TCGA-SKCM and (D) Liu_Data. LASSO, least absolute shrinkage and selection operator; SKCM, skin cutaneous melanoma; TCGA, The Cancer Genome Atlas; λ, tuning parameter.

To validate the classification performance of our risk model, we collected transcriptomic and survival data from an additional melanoma cohort (GSE91061), together with other seven independent datasets from TCGA including those for BLCA, BRCA, HNSC, KIRP, LIHC, LUAD, and LUSC. According to our model, the OS of patients in the low-risk group was longer than that of patients in the high-risk group across eight external validation cohorts. The differences in survival probabilities between high- and low-risk groups were statistically significant in five datasets, including GSE91061, BLCA, HNSC, LIHC, and LUAD (P<0.05) (Figure 5A). These findings demonstrate the potential application of our nine-gene risk model in stratifying patients across a variety of cancers. The hazard ratios determined via multivariate Cox regression analysis identified two independent prognostic factors, age (P=0.04) and risk score (P<0.001) (Figure 5B). A nomogram, involving stage, gender, age, and risk signature as indicators, was developed to compare the survival differences among patients. The scores assigned to each variable in the nomogram were summed up and converted into a probability score for predicting the 1-, 2-, and 3-year OS (Figure 5C).

Figure 5 Prognostic evaluation and nomogram construction. (A) Kaplan-Meier survival curves of patients with high and low risk in a melanoma dataset (GSE91061) and BLCA, BRCA, HNSC, KIRP, LIHC, LUAD, and LUSC cohorts; (B) identification of independent prognostic factors according to multivariate Cox regression analyses; (C) the nomogram for patients with melanoma. BLCA, bladder urothelial carcinoma; BRCA, breast invasive carcinoma; HNSC, head and neck squamous cell carcinoma; KIRP, kidney renal papillary cell carcinoma; LIHC, liver hepatocellular carcinoma; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma.

Overall characteristics of immune cells in melanoma

As the nine prognostic genes within the risk model were found to be closely tied to immune activity, we next investigated their impact on immune cells in the tumor microenvironment. We used a scRNA-seq dataset of tumor samples from patients with melanoma who subsequently received immunotherapies and displayed varied treatment responses (21 non-responding and 8 responding) (32). A total of 10,163 immune cells were clustered into 11 clusters (Figure S3A). Based on the expressions of canonical marker genes, they were annotated as seven major immune cell types, including B cells, monocytes/macrophages, NK cells, plasmacytoid dendritic cells (pDCs), plasma cells, Tregs, and T cells (Figure 6A). The expressions of immunophenotype-specific genes on each cell type were confirmed, which corroborate the clustering results (Figure 6B). Cells from the responder group showed slight increase in the abundance of B cells and plasma cell and a decrease in that of Tregs and monocytes/macrophages, although these differences were not statistically significant (Figure S3B and Figure 6C). By examining the expression of nine prognostic genes across the immune cell types, we observed that the majority of these genes are readily detectable but exhibit varying degrees of enrichment across different cell types (Figure 6D and Figure S3C).

Figure 6 Single-cell immune landscape and prognostic gene expression. (A) UMAP plot of 10,163 immune cells showing seven cell types; (B) Dot plot indicating the expression levels of the marker genes for each cell type; (C) Box plots visualizing the proportions of cell populations in responders and non-responders; (D) UMAP plots showing the expression levels of prognostic genes in immune cell clusters. NK, natural killer; pDCs, plasmacytoid dendritic cells; UMAP, uniform manifold approximation and projection.

Selective expression of CST7 in myeloid cells correlated with therapy response

Since myeloid cells play a crucial role in the immunosuppressive tumor microenvironment, we focused on the expression landscape of the nine prognostic genes in myeloid cells (36). By reclustering a total of 1,227 monocytes/macrophages, six clusters were identified and designated as C0–C5 (Figure 7A). Among the nine prognostic genes, only CST7, CSF2RB, MYO1G, and NCF1 were readily detectable in myeloid cells. Of note, CST7 had the most drastic difference in expression level in monocytes/macrophages between responders and non-responders, while the expression levels of other genes were similar in both groups (Figure 7B). CST7 was highly expressed in four monocyte/macrophage clusters (C0, C3, C4, and C5) in the responder group but was only expressed in two clusters (C5 and C5) in the non-responder group (Figure 7C). In addition, C0 and C4 in the responder group had significantly upregulated CST7 expression as compared to the non-responder group (Figure 7C). Expression levels of the other eight genes across the six clusters also showed certain degrees of differences between responders and non-responders in certain subpopulations, although these were not as markedly different as was CST7 (Figure S3D). Based on CST7 expression (Figure S3E), we subsequently grouped monocytes/macrophages into CST7⁺ and CST7^– clusters (Figure 7D, Figure S3E). CST7⁺ cells were found to be more abundant in the responder group than in the non-responder group, suggesting its association with therapy response (P<0.05) (Figure 7E).

Figure 7 Characterization of monocytes/macrophages and CST7-associated immune features. (A) UMAP plots of 1,227 monocytes/macrophages, colored by cell cluster and immunotherapy response; (B) violin plots showing the expression profiles of 9 prognostic genes in monocytes/macrophages of immunotherapy responders and non-responders; (C) expression patterns of CST7 in six monocytes/macrophage subpopulations from immunotherapy responders and non-responders; (D) UMAP plot indicating CST7⁺ and CST7^– monocytes/macrophages; (E) bar plot showing proportions of CST7⁺ and CST7^– cells in immunotherapy responders and non-responders; (F) Hallmark pathway enrichment analysis between CST7⁺ and CST7^– monocytes/macrophages; (G) GSEA showing the enrichment for gene sets of the M1 signature and M2 signature in CST7⁺ and CST7^– monocytes/macrophages; (H) Kaplan-Meier survival curves for patients with melanoma stratified by different CST7 expression levels. *, P<0.05; ***, P<0.001; GSEA, gene set enrichment analysis; NES, normalized enrichment score; UMAP, uniform manifold approximation and projection.

To investigate the molecular mechanism by which these CST7-expressing cells influence the TME, we analyzed the DEGs between CST7⁺ and CST7^– cells. The upregulated genes in CST7⁺ cells were enriched for proimmune hallmarks (P<0.05), such as IFN-γ response, TNF-α signaling via NFKB, inflammatory response, IFN-α response, and IL-2.STAT5 signaling, while the upregulated genes in the CST7^– cluster were associated with protumor pathways, including hypoxia and angiogenesis (Figure 7F). Additionally, GSEA indicated that CST7⁺ monocytes/macrophages exhibited stronger M1 signature (normalized enrichment score =2.86, P<0.001), which is typically considered to indicate an antitumor phenotype (Figure 7G); moreover, the downregulation of genes was associated with protumorigenic M2 polarization. In line with this, Pearson correlation analysis indicated a significantly positive association between CST7 expression level and M1 signature score (P<0.05), but a negative correlation was observed with M2 signature score (P<0.05) (Figure S3F). To determine the clinical implications of antitumor CST7⁺ cells have, we analyzed the TCGA RNA-seq data of tumor samples from patients with melanoma and found that patients with high CST7 expression levels exhibited significantly better clinical outcomes (P<0.05) (Figure 7H). These findings support the notion that CST7⁺ monocytes/macrophages contribute to the formation of an immune-active melanoma tumor microenvironment and that its prevalence positively correlates with a favorable patient response to immunotherapies.

CST7⁺ monocytes/macrophage enhancement of antitumor response via C9T cells

To clarify how CST7⁺ cells impact immune activity in the tumor microenvironment, we examined on the crosstalk between these monocytes/macrophages and T cells, the major contributors to antitumor immunity. A total of 665 T cells were further reclassified into 15 clusters (Figure 8A). Based on the expressions of typical T cell markers, we defined nine subpopulations, C1T to C9T (Figure 8B,8C). The proportions of these subpopulations in each sample were subsequently evaluated, and a slightly higher percentage of C9T cluster was found in the responder group (Figure 8D). Cell-cell communication analysis showed a drastic increase in the communication intensity between CST7⁺ monocyte/macrophages and C9T cells in responders (Figure 8E). C9T cells expressed high levels of CD8A and CD8B, classical markers of CD8⁺ T cells; GZMH, a cytotoxicity-related molecule; and CCL4, a chemokine recruiting antitumor immune cell (Figure 8B). These results indicate that C9T cells may interact with CST7⁺ monocytes/macrophages in the tumor microenvironment of melanoma, subsequently upregulating T-cell cytotoxicity against cancer cells and improving immunotherapy efficacy. To identify the exact signals transmitted between these two cell types, we examined the ligand-receptor pairs between monocytes/macrophages and C9T cells and identified the ICOSL-ICOS axis as a pair that is specific to the transmission between CST7⁺ cells and C9T cells (Figure 8F). ICOSL is encoded by the ICOSG gene and expressed in antigen-presenting cells, whereas its receptor ICOS is a costimulatory molecule expressed on T cells and is crucial for T-cell activation (37). In line with the cell-cell communication analysis, in responders, ICOSG and ICOS were upregulated in CST7⁺ monocytes/macrophages and C9T cells, respectively (Figure 8G,8H). Our findings suggest that CST7⁺ monocytes/macrophages enhance the antitumor response by interacting with C9T cells through the ICOSL-ICOS axis in the TIME.

Figure 8 T-cell landscape and interactions with CST7⁺/CST7^– monocytes/macrophages in melanoma. (A) UMAP plot showing the clustering of T cells in melanoma; (B) dot plot displaying the expression levels of known marker genes across different T-cell subpopulations; (C) UMAP plot of 9 annotated clusters; (D) histograms presenting the proportions of T-cell subpopulations in each sample and groups of responders and non-responders; (E) chord diagrams showing the communication probability between T-cell subpopulations and CST7⁺ monocytes/macrophages or CST7^– monocytes/macrophages; (F) dot plots indicating the expression levels of ligand-receptor pairs involved in the interactions between macrophages/monocytes and C9T cells. A receptor-ligand pair specific to CST7⁺ macrophages/monocytes in the responder group has been highlighted in red; (G) violin plot showing the expression levels of ICOSLG in CST7⁺ and CST7^– monocytes/macrophages for the responder group; (H) violin plots showing the expression levels of ICOS across nine T-cell subpopulations, highlighting the overexpression in the C9T subpopulation in responders compared to that in non-responders. HLA, human leukocyte antigen; ICOS, inducible T-cell costimulator; ICOSL, inducible T-cell costimulator ligand; ICOSLG, also known as ICOSL, inducible T-cell costimulator ligand; UMAP, uniform manifold approximation and projection.

Discussion

Given the limited long-term response rate to ICIs and the lack of broadly applicable prognostic indicators in the clinical management of melanoma, conducting a comprehensive screening of TIME features involved in immune response or escape and clarifying the mechanisms underlying tumor-immunity interactions may contribute to identifying novel targets and robust efficacy biomarkers. In this study, we employed classical immune-related hallmarks to evaluate immunogenic traits of melanoma samples. A nine-gene risk model was established to stratify patients to an HIG and an LIG and showed a strong prognostic ability for melanoma. Subsequently, the robustness of this model was validated in multi-cancer cohorts. Integration of scRNA-seq data identified a CST7-expressing monocyte/macrophage population that strongly correlated with treatment efficacy and acted as a crucial regulator of T-cell function in the tumor microenvironment.

In patients with pancreatic ductal adenocarcinoma (38), cervical cancer (39), or liver cancer (40), elevated CST7 expression has been reported to be associated with lower risk and positive prognostic outcomes. Previous studies have indicated that CST7 expressed on CD8⁺ T cells can regulate cytotoxic functions (41,42). Additionally, cystatin F, encoded by CST7, has been revealed to modulate macrophage-mediated immune responses through inhibiting the cathepsins involved in antigen processing, cell adhesion, and phagocytosis (43-45). Compared to resting macrophages, M1 macrophages exhibit significant upregulation of cystatin F, suggesting reduced proteolytic activity that can preserve epitopes and potentially activate lymphocytes (46,47). Moreover, CST7 has been implicated in the regulation of neutrophil extracellular traps (NETs), with evidence indicating that targeting CST7 can effectively reduce NET formation (48). Furthermore, CST7 has been identified as a key gene involved in B cell maturation and correlated with improved immunotherapeutic efficacy in non-small cell lung cancer (49). However, the selective expression of CST7 in macrophages/monocytes and its correlation with ICI efficacy reported in our work have not been previously examined. We termed the macrophages/monocytes with high CST7 expression as CST7⁺ macrophages/monocytes and found that CST7⁺ monocytes/macrophages are enriched for M1 signature. This suggests that CST7 may contribute to polarization of monocytes/macrophages towards an M1 phenotype, forming an immune-activating microenvironment. Furthermore, some cysteine proteases, critical for protein degradation within the lysosome, may influence the release of pro-inflammatory factors. As CST7 encodes a cysteine protease inhibitor, it may thereby promote M1 polarization of macrophages by modulating the release of these pro-inflammatory factors (50). This provides a mechanistic basis for using the signature involving CST7 and increased CST7⁺ macrophages/monocytes in the TIME as a powerful tool to stratify patients with melanoma with different clinical outcomes.

Additionally, we found that CST7⁺ monocyte/macrophages enhance antitumor response by influencing C9T, a cytotoxic T-cell subpopulation. ICOS on C9T acts as a receptor to receive ICOSL, a ligand of ICOS, from CST7⁺ monocyte/macrophages. ICOS, an activating costimulatory immune checkpoint expressed on T cells, along with ICOSL, expressed on antigen-presenting cells, can exert distinct effects on immune response depending on the type of T-cell subset (37,51). When activated, ICOSL binds to Tregs, induces an immunosuppressive signal, and triggers tumor relapse (52). In a mouse model of prostate cancer, depletion of ICOS⁺ Treg cells led to remarkably reduced tumor sizes (53). In gastric cancer, tumor-infiltrating mast cells secrete IL-33 and IL-2, leading to the activation of ICOS⁺ Tregs, suppression of effector T-cell function, and promotion of tumor progression (54). Similarly, in ovarian and breast cancers, ICOS⁺ Tregs further dampen antitumor immune responses through interactions with ICOSL⁺ dendritic cell (55). In a retrospective study enrolling 14 patients with metastatic melanoma, longer OS was positively correlated with reduced proportions of CD4⁺ FoxP3⁺ ICOS⁺ T cells during ICI therapy (56). In contrast, the stimulated ICOS-ICOSL axis can enhance the antitumor immunity of CD4⁺ helper T cells, CD8⁺ cytotoxic T lymphocytes, and follicular helper T cells in the tumor microenvironment (37). A study using ICOS-knockout mice revealed the primary resistance against anti-CTLA-4, indicating the necessity of the ICOS/ICOSL pathway for therapeutic response activation (57). Increased ICOS⁺ CD8⁺ T cells were observed in ICI-treated clinical samples of breast cancer (58), lung cancer (59), and cervical cancer (60). Moreover, the interaction between ICOS and ICOSL activates downstream signaling pathways that regulate the activation and antitumor effects of cytotoxic T cells. These These signals are essential for cytotoxic T cell functionality, enhancing their cytotoxicity and persistence (37,61). For example, ICOS signaling significantly augments the cytotoxicity of engineered CAR-T cells, thereby improving their antitumor efficacy (55). Moreover, the synergistic effect of ICOS with costimulatory molecules, such as OX40, further potentiates the immune response of cytotoxic T cells (62). This mechanism underscores the pivotal role of the ICOSL-ICOS axis in tumor immunity.

In light of the dual roles of ICOS in the TIME, both ICOS/ICOSL agonistic and antagonistic antibodies may contribute to enhanced immunotherapy efficacy, as has been attested to in research over the past decades. In a melanoma mouse model, concurrent CTLA-4 blockade and ICOS activation not only selectively stimulated CD8⁺ T cells over Tregs but also reprogrammed TAMs toward a proinflammatory (M1-like) phenotype, collectively enhancing antitumor immunity (63,64). In other studies, patients receiving anti-CTLA-4 and cellular vaccine activating ICOS, the TIME was infiltrated, with an elevated effector T cell:Treg ratio, triggering increased therapeutic response (52,65). Given the synergistic effect of ICIs and ICOS agonistic antibodies in activating effector T cells reported in numerous preclinical and clinical studies, several ongoing trials are investigating this combination therapy in various solid-tumor cohorts. Despite being in the early phase, one study found promising clinical activity and good tolerability for the combination of ICOS agonistic antibody and ICIs (66). Another phase I study reported that an arm consisting of ICOS agonistic antibodies (JTX-2011) plus nivolumab displayed activated antitumor immune response in gastric cancer and triple-negative breast cancer (67). Meanwhile, ICOS antagonistic antibodies have been shown to improve the antitumor activity by silencing immunosuppressive Tregs and were well-tolerated when combined with atezolizumab in a phase I clinical data (68). However, clinical evidence for ICOS targeting in combination with ICIs is limited. Our findings, along with previous studies examining ICOS antibodies, support the potentially synergetic antitumor effects of ICIs and ICOS/ICOSL agonistic antibodies, providing the mechanistic basis for a novel combined therapeutic strategy than can maintain a long-term immunotherapy response.

Several limitations to this study should be acknowledged. All the data used in this study were collected from public datasets and retrospective cohorts, and our findings need to be further confirmed in prospective cohorts. Moreover, the expression profiles of hub genes in specific cell populations were assessed only at the transcriptional level, and their protein expression still needs to be validated in tumor samples. The identified cell–cell communication between immune cells was based on RNA-seq analysis, and further functional experiments are required to assess its biological and clinical significance. Furthermore, as the dataset analyzed only consisted of immune cells, the crosstalk between immune cells and malignant cells was not examined. A more comprehensive investigation of the malignant-immune cell interaction is warranted to garner novel insights into the underlying mechanisms of the tumor microenvironment of melanoma.

Conclusions

The identification of clusters with distinct immune characteristics and the development of a risk signature has deepened our understanding of the heterogeneous immune landscape in melanoma. Comprehensive analysis of multi-cancer cohorts demonstrated the reliability of our findings, supporting the application of this immune signature in stratifying patients. The discovery of key cellular components and their molecular communications may contribute to expanding the clinical options of combination immunotherapy in patients with advanced melanoma. This work may further inform personalized treatment strategies, patient management, and follow-up care.

Acknowledgments

We would like to express our sincere thanks to the providers of the public data utilized in this study and to express our deepest gratitude to all the researchers and participants involved in the construction of these data.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1943/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1943/prf

Funding: This work was financially supported by the Traditional Chinese Medicine Research Plan Project (2022), the Administration of Traditional Chinese Medicine of Hebei Province, China (No. 2022146 to Z.L.).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1943/coif). B.Z., Y.L., L.W., and L.Z. are current employees of Hangzhou Astrocyte Technology Co., Ltd. Z.L. reports funding support from the Traditional Chinese Medicine Research Plan Project (2022), the Administration of Traditional Chinese Medicine of Hebei Province, China (No. 2022146). The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Tasdogan A, Sullivan RJ, Katalinic A, et al. Cutaneous melanoma. Nat Rev Dis Primers 2025;11:23. [Crossref] [PubMed]
2. Boutros A, Croce E, Ferrari M, et al. The treatment of advanced melanoma: Current approaches and new challenges. Crit Rev Oncol Hematol 2024;196:104276. [Crossref] [PubMed]
3. Garbe C, Eigentler TK, Keilholz U, et al. Systematic review of medical treatment in melanoma: current status and future prospects. Oncologist 2011;16:5-24. [Crossref] [PubMed]
4. Zhang J, Joshua AM, Li Y, et al. Targeted therapy, immunotherapy, and small molecules and peptidomimetics as emerging immunoregulatory agents for melanoma. Cancer Lett 2024;586:216633. [Crossref] [PubMed]
5. Carlino MS, Larkin J, Long GV. Immune checkpoint inhibitors in melanoma. Lancet 2021;398:1002-14. [Crossref] [PubMed]
6. Long GV, Larkin J, Schadendorf D, et al. Pooled Long-Term Outcomes With Nivolumab Plus Ipilimumab or Nivolumab Alone in Patients With Advanced Melanoma. J Clin Oncol 2025;43:938-48. [Crossref] [PubMed]
7. Long GV, Atkinson V, Lo SN, et al. Ipilimumab plus nivolumab versus nivolumab alone in patients with melanoma brain metastases (ABC): 7-year follow-up of a multicentre, open-label, randomised, phase 2 study. Lancet Oncol 2025;26:320-30. [Crossref] [PubMed]
8. Gide TN, Mao Y, Scolyer RA, et al. Tissue-Based Profiling Techniques to Achieve Precision Medicine in Cancer: Opportunities and Challenges in Melanoma. Clin Cancer Res 2024;30:5270-80. [Crossref] [PubMed]
9. Pham JP, Staeger R, Joshua AM, et al. An updated review of immune checkpoint inhibitors in cutaneous oncology: Beyond melanoma. Eur J Cancer 2025;214:115121. [Crossref] [PubMed]
10. Reschke R, Enk AH, Hassel JC. Prognostic Biomarkers in Evolving Melanoma Immunotherapy. Am J Clin Dermatol 2025;26:213-23. [Crossref] [PubMed]
11. Espinosa E, Márquez-Rodas I, Soria A, et al. Predictive factors of response to immunotherapy-a review from the Spanish Melanoma Group (GEM). Ann Transl Med 2017;5:389. [Crossref] [PubMed]
12. Patel SP, Kurzrock R. PD-L1 Expression as a Predictive Biomarker in Cancer Immunotherapy. Mol Cancer Ther 2015;14:847-56. [Crossref] [PubMed]
13. Gibney GT, Weiner LM, Atkins MB. Predictive biomarkers for checkpoint inhibitor-based immunotherapy. Lancet Oncol 2016;17:e542-51. [Crossref] [PubMed]
14. Olbryt M, Rajczykowski M, Widłak W. Biological Factors behind Melanoma Response to Immune Checkpoint Inhibitors. Int J Mol Sci 2020;21:4071. [Crossref] [PubMed]
15. Luke JJ, Flaherty KT, Ribas A, et al. Targeted agents and immunotherapies: optimizing outcomes in melanoma. Nat Rev Clin Oncol 2017;14:463-82. [Crossref] [PubMed]
16. Wolchok JD, Chiarion-Sileni V, Gonzalez R, et al. Overall Survival with Combined Nivolumab and Ipilimumab in Advanced Melanoma. N Engl J Med 2017;377:1345-56. [Crossref] [PubMed]
17. Andrews MC, Li G, Graf RP, et al. Predictive Impact of Tumor Mutational Burden on Real-World Outcomes of First-Line Immune Checkpoint Inhibition in Metastatic Melanoma. JCO Precis Oncol 2024;8:e2300640. [Crossref] [PubMed]
18. Li G, Wu T, Li H, et al. Construction of a tumor mutational burden-derived LncRNA prognostic computational framework associated with therapy sensitivity in skin cutaneous melanoma. J Transl Med 2024;22:966. [Crossref] [PubMed]
19. Hanrahan GB, Giobbie-Hurder A, Allais B, et al. Melanoma Tumor Mutational Burden and Indoor Tanning Exposure. JAMA Dermatol 2025;161:198-202. [Crossref] [PubMed]
20. Holder AM, Dedeilia A, Sierra-Davidson K, et al. Defining clinically useful biomarkers of immune checkpoint inhibitors in solid tumours. Nat Rev Cancer 2024;24:498-512. [Crossref] [PubMed]
21. Le DT, Kim TW, Van Cutsem E, et al. Phase II Open-Label Study of Pembrolizumab in Treatment-Refractory, Microsatellite Instability-High/Mismatch Repair-Deficient Metastatic Colorectal Cancer: KEYNOTE-164. J Clin Oncol 2020;38:11-9. [Crossref] [PubMed]
22. Ballerini M, Galiè S, Tyagi P, et al. A gut-on-a-chip incorporating human faecal samples and peristalsis predicts responses to immune checkpoint inhibitors for melanoma. Nat Biomed Eng 2025;9:967-84. [Crossref] [PubMed]
23. Macandog ADG, Catozzi C, Capone M, et al. Longitudinal analysis of the gut microbiota during anti-PD-1 therapy reveals stable microbial features of response in melanoma patients. Cell Host Microbe 2024;32:2004-2018.e9. [Crossref] [PubMed]
24. Pozniak J, Pedri D, Landeloos E, et al. A TCF4-dependent gene regulatory network confers resistance to immunotherapy in melanoma. Cell 2024;187:166-183.e25. [Crossref] [PubMed]
25. Robertson BM, Fane ME, Weeraratna AT, et al. Determinants of resistance and response to melanoma therapy. Nat Cancer 2024;5:964-82. [Crossref] [PubMed]
26. Wawrzyniak P, Hartman ML. Dual role of interferon-gamma in the response of melanoma patients to immunotherapy with immune checkpoint inhibitors. Mol Cancer 2025;24:89. [Crossref] [PubMed]
27. Rolig AS, Peng X, Sturgill ER, et al. The response to anti-PD-1 and anti-LAG-3 checkpoint blockade is associated with regulatory T cell reprogramming. Sci Transl Med 2025;17:eadk3702. [Crossref] [PubMed]
28. Wu N, Li J, Li L, et al. MerTK(+) macrophages promote melanoma progression and immunotherapy resistance through AhR-ALKAL1 activation. Sci Adv 2024;10:eado8366. [Crossref] [PubMed]
29. Vasilevska J, Cheng PF, Lehmann J, et al. Monitoring melanoma patients on treatment reveals a distinct macrophage population driving targeted therapy resistance. Cell Rep Med 2024;5:101611. [Crossref] [PubMed]
30. Mukherjee N, Katsnelson E, Brunetti TM, et al. MCL1 inhibition targets Myeloid Derived Suppressors Cells, promotes antitumor immunity and enhances the efficacy of immune checkpoint blockade. Cell Death Dis 2024;15:198. [Crossref] [PubMed]
31. Liu D, Schilling B, Liu D, et al. Integrative molecular and clinical modeling of clinical outcomes to PD1 blockade in patients with metastatic melanoma. Nat Med 2019;25:1916-27. [Crossref] [PubMed]
32. Sade-Feldman M, Yizhak K, Bjorgaard SL, et al. Defining T Cell States Associated with Response to Checkpoint Immunotherapy in Melanoma. Cell 2018;175:998-1013.e20. [Crossref] [PubMed]
33. Shen Y, Peng X, Shen C. Identification and validation of immune-related lncRNA prognostic signature for breast cancer. Genomics 2020;112:2640-6. [Crossref] [PubMed]
34. Butler A, Hoffman P, Smibert P, et al. Integrating single-cell transcriptomic data across different conditions, technologies, and species. Nat Biotechnol 2018;36:411-20. [Crossref] [PubMed]
35. Jin S, Guerrero-Juarez CF, Zhang L, et al. Inference and analysis of cell-cell communication using CellChat. Nat Commun 2021;12:1088. [Crossref] [PubMed]
36. Barry ST, Gabrilovich DI, Sansom OJ, et al. Therapeutic targeting of tumour myeloid cells. Nat Rev Cancer 2023;23:216-37. [Crossref] [PubMed]
37. Solinas C, Gu-Trantien C, Willard-Gallo K. The rationale behind targeting the ICOS-ICOS ligand costimulatory pathway in cancer immunotherapy. ESMO Open 2020;5:e000544. [Crossref] [PubMed]
38. Yang C, Yu T, Liu Z, et al. Cystatin F as a key family 2 cystatin subunit and prognostic biomarker for early-stage pancreatic ductal adenocarcinoma. Oncol Rep 2019;42:79-90. [Crossref] [PubMed]
39. Yao S, Zhao L, Chen S, et al. Cervical cancer immune infiltration microenvironment identification, construction of immune scores, assisting patient prognosis and immunotherapy. Front Immunol 2023;14:1135657. [Crossref] [PubMed]
40. Yan Z, Lijuan Y, Yinhang W, et al. Screening and analysis of RNAs associated with activated memory CD4 and CD8 T cells in liver cancer. World J Surg Oncol 2022;20:2. [Crossref] [PubMed]
41. Qu J, Wu B, Chen L, et al. CXCR6-positive circulating mucosal-associated invariant T cells can identify patients with non-small cell lung cancer responding to anti-PD-1 immunotherapy. J Exp Clin Cancer Res 2024;43:134. [Crossref] [PubMed]
42. Gao H, Sun H, He A, et al. Single-cell combined bioinformatics analysis: construction of immune cluster and risk prognostic model in kidney renal clear cells based on CD8(+) T cell-associated genes. Eur J Med Res 2024;29:89. [Crossref] [PubMed]
43. Perišić Nanut M, Sabotič J, Jewett A, et al. Cysteine cathepsins as regulators of the cytotoxicity of NK and T cells. Front Immunol 2014;5:616. [Crossref] [PubMed]
44. Dautović E, Perišić Nanut M, Softić A, et al. The transcription factor C/EBP α controls the role of cystatin F during the differentiation of monocytes to macrophages. Eur J Cell Biol 2018;97:463-73. [Crossref] [PubMed]
45. Kopitar-Jerala N. The role of cystatins in cells of the immune system. FEBS Lett 2006;580:6295-301. [Crossref] [PubMed]
46. Pires D, Marques J, Pombo JP, et al. Role of Cathepsins in Mycobacterium tuberculosis Survival in Human Macrophages. Sci Rep 2016;6:32247. [Crossref] [PubMed]
47. Russell DG, Vanderven BC, Glennie S, et al. The macrophage marches on its phagosome: dynamic assays of phagosome function. Nat Rev Immunol 2009;9:594-600. [Crossref] [PubMed]
48. Hu S, Zhang F, Wang J, et al. MMP9(High) Neutrophils are Critical Mediators of Neutrophil Extracellular Traps Formation and Myocardial Ischemia/Reperfusion Injury. Adv Sci (Weinh) 2025;12:e2415205. [Crossref] [PubMed]
49. Shi H, Wang W, Luo J, et al. POU2F2(+) B cells enhance antitumor immunity and predict better survival in non small cell lung cancer. Sci Rep 2025;15:6549. [Crossref] [PubMed]
50. Oelschlaegel D, Weiss Sadan T, Salpeter S, et al. Cathepsin Inhibition Modulates Metabolism and Polarization of Tumor-Associated Macrophages. Cancers (Basel) 2020;12:2579. [Crossref] [PubMed]
51. Hutloff A, Dittrich AM, Beier KC, et al. ICOS is an inducible T-cell co-stimulator structurally and functionally related to CD28. Nature 1999;397:263-6. [Crossref] [PubMed]
52. Marinelli O, Nabissi M, Morelli MB, et al. ICOS-L as a Potential Therapeutic Target for Cancer Immunotherapy. Curr Protein Pept Sci 2018;19:1107-13. [Crossref] [PubMed]
53. Mo L, Chen Q, Zhang X, et al. Depletion of regulatory T cells by anti-ICOS antibody enhances anti-tumor immunity of tumor cell vaccine in prostate cancer. Vaccine 2017;35:5932-8. [Crossref] [PubMed]
54. Lv Y, Tian W, Teng Y, et al. Tumor-infiltrating mast cells stimulate ICOS(+) regulatory T cells through an IL-33 and IL-2 axis to promote gastric cancer progression. J Adv Res 2024;57:149-62. [Crossref] [PubMed]
55. Cao L, Peng H, Chen Y, et al. ICOS-expressing CAR-T cells mediate durable eradication of triple-negative breast cancer and metastasis. J Immunother Cancer 2024;12:e010028. [Crossref] [PubMed]
56. Carthon BC, Wolchok JD, Yuan J, et al. Preoperative CTLA-4 blockade: tolerability and immune monitoring in the setting of a presurgical clinical trial. Clin Cancer Res 2010;16:2861-71. [Crossref] [PubMed]
57. Fu T, He Q, Sharma P. The ICOS/ICOSL pathway is required for optimal antitumor responses mediated by anti-CTLA-4 therapy. Cancer Res 2011;71:5445-54. [Crossref] [PubMed]
58. Vonderheide RH, LoRusso PM, Khalil M, et al. Tremelimumab in combination with exemestane in patients with advanced breast cancer and treatment-associated modulation of inducible costimulator expression on patient T cells. Clin Cancer Res 2010;16:3485-94. [Crossref] [PubMed]
59. Kamphorst AO, Pillai RN, Yang S, et al. Proliferation of PD-1+ CD8 T cells in peripheral blood after PD-1-targeted therapy in lung cancer patients. Proc Natl Acad Sci U S A 2017;114:4993-8. [Crossref] [PubMed]
60. Heeren AM, Koster BD, Samuels S, et al. High and interrelated rates of PD-L1+CD14+ antigen-presenting cells and regulatory T cells mark the microenvironment of metastatic lymph nodes from patients with cervical cancer. Cancer Immunol Res 2015;3:48-58. [Crossref] [PubMed]
61. Amatore F, Gorvel L, Olive D. Role of Inducible Co-Stimulator (ICOS) in cancer immunotherapy. Expert Opin Biol Ther 2020;20:141-50. [Crossref] [PubMed]
62. Xiao Z, Mayer AT, Nobashi TW, et al. ICOS Is an Indicator of T-cell-Mediated Response to Cancer Immunotherapy. Cancer Res 2020;80:3023-32. [Crossref] [PubMed]
63. Fan X, Quezada SA, Sepulveda MA, et al. Engagement of the ICOS pathway markedly enhances efficacy of CTLA-4 blockade in cancer immunotherapy. J Exp Med 2014;211:715-25. [Crossref] [PubMed]
64. Sharma N, Fan X, Atolagbe OT, et al. ICOS costimulation in combination with CTLA-4 blockade remodels tumor-associated macrophages toward an antitumor phenotype. J Exp Med 2024;221:e20231263. [Crossref] [PubMed]
65. Liakou CI, Kamat A, Tang DN, et al. CTLA-4 blockade increases IFNgamma-producing CD4+ICOShi cells to shift the ratio of effector to regulatory T cells in cancer patients. Proc Natl Acad Sci U S A 2008;105:14987-92. [Crossref] [PubMed]
66. Burris HA, Callahan MK, Tolcher AW, et al. Phase 1 safety of ICOS agonist antibody JTX-2011 alone and with nivolumab (nivo) in advanced solid tumors; predicted vs observed pharmacokinetics (PK) in ICONIC. J Clin Oncol 2017;35:abstr 3033.
67. Yap TA, Burris HA, Kummar S, et al. ICONIC: Biologic and clinical activity of first in class ICOS agonist antibody JTX-2011+/-nivolumab (nivo) in patients (pts) with advanced cancers. J Clin Oncol 2018;36:abstr 3000.
68. Patel MR, Naing A, Burris III HA, et al. A phase 1/2 open-label study of KY1044, an anti-ICOS antibody with dual mechanism of action, as single agent and in combination with atezolizumab, in adult patients with advanced malignancies. J Clin Oncol 2021;39:abstr 2624.

(English Language Editor: J. Gray)