Introduction

Melanoma is a type of skin cancer that develops from melanocytes. It is one of the most aggressive forms of skin cancer and can quickly spread to other parts of the body if left untreated1. While surgery is effective in treating early-stage melanoma, advanced-stage melanoma has been proven to be difficult to manage with conventional therapies2. Currently, immunotherapy and targeted molecular therapy have emerged as promising treatment options for melanoma and immune checkpoint inhibitors have been approved in various settings, including adjuvant, first-line, and second-line treatments of metastatic melanoma3.

Immune checkpoint inhibitors, such as ipilimumab, nivolumab and pembrolizumab, block the signals that cancer cells use to evade the immune system4. Specifically, they target molecules on the surface of immune cells, such as cytotoxic T lymphocyte-associated antigen 4 (CTLA-4) and programmed death-1 (PD-1) on the surface of T cells, which act as “brakes” on the immune response. By blocking these checkpoints, immune cells are able to mount a stronger attack against cancer cells4. CTLA-4 inhibitors work by blocking the activity of CTLA-4, a protein that suppresses the ability of the immune system to attack cancer cells5. Clinical trials have shown that immune checkpoint inhibitors provide durable clinical benefit across different treatment settings, improving responses and survival in patients with advanced melanoma6 and also producing enhanced and durable responses in the neoadjuvant setting7,8.

Despite the promising results seen with these immunotherapies, there are several challenges and limitations remaining3. One major hurdle is that not all patients respond to these therapies, and even those who do often experience side effects, such as inflammation and autoimmune reactions. It is not yet clear which patients are most likely to benefit from immunotherapy. Current biomarkers for immunotherapy, such as PD-L1 expression, tumor mutational burden (TMB) and tertiary lymphoid structures (TLS), have limitations in sensitivity and practicality9. PD-L1 expression, analyzed by IHC, lacks approval as a predictive biomarker for melanoma due to inconsistent correlations with treatment outcomes10. TMB assessment also suffers from a lack of standardization, leading to variability across sequencing methods, complicating the establishment of universal thresholds for predicting immunotherapy responses10. Although intratumoral TLS has been associated with improved response and survival in melanoma patients treated with ICIs11, its routine clinical use as a predictive biomarker is still limited by the lack of standardized scoring criteria, marked spatial/temporal heterogeneity, technical requirements for robust quantification, and the paucity of prospective validation in homogeneous melanoma cohorts12. Identifying better biomarkers to predict therapeutic response may help to improve the effectiveness of checkpoint inhibitor therapy and provide better outcomes for patients with melanoma. Another challenge is the development of resistance to these therapies. While many patients initially respond well to immunotherapy, some may eventually develop resistance and experience disease progression. This highlights the need to develop new therapeutic strategies to overcome drug resistance.

CD27 is a transmembrane glycoprotein primarily expressed on various immune cells, including T cells, B cells and natural killer (NK) cells13. As a member of the TNF receptor superfamily, CD27 plays a key role in the regulation of immune responses. CD27 engagement with its ligand, CD70, on tumor-infiltrating lymphocytes (TILs) promotes T cell activation, differentiation and survival, leading to a stronger immune response against the tumor14. It has been reported that resistance to immune checkpoint inhibitors, such as PD-1, in melanoma is associated with the lack of CD8+ T cells in the tumor microenvironment, which is caused by the lack of recruitment of dendritic cells15,16. Another study demonstrated that agonizts targeting CD27 can bypass dendritic cell signaling and thus facilitate T cell activation for anti-tumor effects17. In recent years, the role of CD27 in tumors has garnered significant attention, as it has emerged as a promising target for cancer immunotherapy18.

Given its critical role in regulating anti-tumor immune responses, CD27 has emerged as an attractive target for cancer immunotherapy. Several approaches are being developed to target CD27 and enhance anti-tumor immune responses. One approach is to use agonistic antibodies that activate CD27 signaling and enhance T cell activation and survival. Preclinical studies have demonstrated that the efficacy of CD27 agonizts in promoting anti-tumor immune responses in various cancer models19. Another approach is to combine CD27 agonizts with other immunotherapies, such as immune checkpoint inhibitors or adoptive T cell therapy, to enhance their efficacy. Combination therapies targeting CD27 are currently being tested in clinical trials for various cancers19.

In conclusion, immune checkpoint inhibitors and immune co-stimulatory molecules have shown great promise in the treatment of melanoma. However, there are still many challenges and limitations that must be addressed in order to optimize the use of these drugs and improve patient outcomes. This study aims to investigate the ability of CD27 to predict the response to immunotherapy by analyzing the gene expression data of melanoma samples.

Results

High CD27 mRNA expression positively correlated with immune checkpoint genes and favorable prognosis

Immune checkpoint blockade-based therapy has become a major weapon in fighting cancer. Antibody drugs, such as anti-PD-1/PD-L1 and anti-CTLA4, have achieved broad applicability across cancer types and durable clinical response when treatment is effective, especially in melanoma20. CD27 is a member of the TNF receptor superfamily and is involved in T-cell activation by providing a co-stimulatory signal. To investigate the relationship between CD27 and immune checkpoint genes, we performed Pearson correlation analysis on mRNA expression of CD27 and immune checkpoint genes among 366 metastatic SKCM samples from TCGA21. We found that CD27 showed significantly positive correlations with immune checkpoint genes, including PDCD1 (R = 0.92, p < 0.001, Fig. 1A), TIGIT (R = 0.93, p < 0.001, Fig. 1B), LAG3 (R = 0.87, p < 0.001, Fig. 1C), HAVCR2 (R = 0.8, p < 0.001, Fig. 1D), IDO1 (R = 0.75, p < 0.001, Fig. 1E), BTLA (R = 0.75, p < 0.001, Fig. 1F), CD274 (R = 0.66, p < 0.001, Fig. 1G) and CTLA4 (R = 0.5, p < 0.001, Fig. 1H). Overall survival data were available for 352 patients. Overall survival analysis based on CD27 median expression value (FPKM 2.447049) indicated that higher CD27 mRNA expression levels were associated with significantly prolonged OS (p < 0.001, Fig. 1I). For the 346 patients with available disease-specific survival data, stratification by the median CD27 expression level(FPKM 2.461077) revealed that higher CD27 expression was associated with a prolonged disease-specific survival time (p < 0.001, Fig. 1J). For the 353 patients with available PFS data, stratification based on the median CD27 expression level (FPKM 2.452296) showed that patients with higher CD27 expression levels tend to have longer PFS time (p = 0.05, Fig. 1K). For the 165 metastatic patients for whom complete data on sex, age, and tumor stage were available, multivariate Cox regression analysis confirmed that CD27 mRNA expression remained an independent prognostic factor for OS (P < 0.001) in metastatic SKCM (Fig. 1L). Overall, high CD27 mRNA expression positively correlated with immune checkpoint genes and favorable prognosis in melanoma, which can be used as a predictive biomarker of favorable prognosis.

Fig. 1: Correlations between CD27 mRNA expression and immune checkpoint genes, as well as prognosis in melanoma.
Fig. 1: Correlations between CD27 mRNA expression and immune checkpoint genes, as well as prognosis in melanoma.The alternative text for this image may have been generated using AI.
Full size image

A-H CD27 mRNA expression was significantly positively correlated with the expression of immune checkpoint genes, including PDCD1, TIGIT, LAG3, HAVCR2, IDO1, BTLA, CD274, and CTLA4. I Kaplan–Meier survival curves for overall survival, grouped by median CD27 mRNA expression: High (FPKM ≥ 2.447049) and Low (FPKM < 2.447049). J Kaplan–Meier survival curves for disease-specific survival, grouped by median CD27 mRNA expression: High (FPKM ≥ 2.461077) and Low (FPKM < 2.461077). K Kaplan–Meier survival curves for progression-free survival, grouped by median CD27 mRNA expression: High (FPKM ≥ 2.452296) and Low (FPKM < 2.452296). L Multivariate risk regression analysis incorporating tumor stage, sex, age, and CD27 gene expression. *p < 0.05, **p < 0.01.

Associations of CD27 gene expression with immune cell infiltration in tumor microenvironment (TME)

Tumor-infiltrating immune cells have been reported to be independent predictors of prognosis22. Therefore, we investigated whether CD27 gene expression impacted the immune cell infiltration in TME of 366 metastatic SKCM samples from TCGA21. TIMER 2.0 web application analysis results indicated that increased CD27 mRNA expression was inversely correlated with tumor purity (Fig. 2A), suggesting that CD27 may play an important role in controlling tumor growth. When analyzed with immune cells, CD27 mRNA expression showed positive correlations with the infiltration levels of B cells, CD8+ T cells, CD4+ T cells, macrophages, neutrophils and dendritic cells (Fig. 2A), indicating that CD27 may impact lymphocyte infiltration. In addition, we also compared the infiltration levels of 22 immune cell types estimated by CIBERSORT23 between high and low CD27 mRNA expression groups of metastatic SKCM patients. We found that 16 immune cell types were significantly enriched in the high CD27 mRNA expression group (Fig. 2B), among which CD8+ T cells24, memory B cells and naïve B cells25 have been previously reported to be associated with a favorable prognosis of SKCM. Notably, immunosuppressive cell types, such as Tregs and M2 macrophages, were also enriched in the high CD27 expression group. Leukocyte fraction (Fig. 2C) imputed by methylation data and tumor-infiltrating lymphocyte percentage (Fig. 2D) evaluated by pathological images were also higher in the high CD27 gene expression group. These results indicated that high CD27 gene expression was associated with high immune cell infiltration levels and subsequently impacted prognosis.

Fig. 2: CD27 gene expression associated with immune cell infiltration levels and immune-related pathways.
Fig. 2: CD27 gene expression associated with immune cell infiltration levels and immune-related pathways.The alternative text for this image may have been generated using AI.
Full size image

A Correlations between CD27 mRNA expression and tumor purity, as well as immune cell infiltration levels, including B cells, CD8+ T cells, CD4+ T cells, macrophages, neutrophils and dendritic cells in metastatic SKCM. B Box plot showing differences in absolute proportions of immune cells in the tumor microenvironment between high and low CD27 mRNA expression groups. C Box plot displaying differences in estimated leukocyte proportions based on methylation data between high and low CD27 expression groups. D Box plot showing the proportion of TILs, evaluated from pathological images, in high and low CD27 mRNA expression groups. E GSEA plot illustrating enriched hallmark gene sets and pathways in the high CD27 mRNA expression group (FDR < 0.05). F Heatmap displaying the expression of four immune therapy-related signature gene sets, including IFNγ-6, IFNγ expanded immune 18, effector T-cell and IFNγ/effector T-cell. G–J Box plots of gene set enrichment scores of IFNγ-6 (G), IFNγ expanded immune 18 (H), effector T-cell (I) and IFNγ/effector T-cell (J) in high and low CD27 mRNA expression groups of metastatic melanoma. *p < 0.05, **p < 0.01, ****p < 0.0001; ns, no significance.

Alterations in immunological molecular characteristics related to CD27 mRNA expression

To reveal transcriptomic molecular characterizations associated with CD27 mRNA expression, differential expression analysis was performed between high and low expression groups of 366 metastatic SKCM from TCGA21. 2688 genes were identified to be significantly upregulated in the high expression group (Supplementary Table 2), including 8 immune checkpoint genes (TIGIT, PDCD1, LAG3, BTLA, HAVCR2, CD274, C10orf54, and SIGLEC7 [p < 0.01]) (Supplementary Fig. 1A). Unbiased gene set enrichment analysis demonstrated that 9 hallmark gene sets and 44 KEGG pathways (FDR < 0.05) were significantly enriched in the high CD27 mRNA expression group (Supplementary Table 3). Of note, the anti-tumor immune response-related hallmark gene sets and KEGG pathways including inflammatory response, interferon gamma, B cell receptor signaling pathway, chemokine signaling pathway, cytokine-cytokine receptor interaction, natural killer cell mediated cytotoxicity, T cell receptor signaling pathway, were significantly enriched in the high CD27 mRNA expression group (FDR < 0.001)(Fig. 2E). Furthermore, we explored the expression of four gene sets previously reported to be associated with immunotherapy response, and found that these genes were significantly up-regulated in the high CD27 mRNA expression group (Fig. 2F). Single sample gene set enrichment scores of these gene sets, including a six-gene IFNγ signature26(Fig. 2G), a related 18-gene IFNγ signature26(Fig. 2H), an effector T-cell signature27 (Fig. 2I), a combined IFNγ/effector T-cell signature28(Fig. 2J) were significantly higher in high CD27 mRNA expression group compared with low CD27 mRNA expression group.

Validation of correlations of CD27 gene expression with biological and immune features in an independent metastatic SKCM cohort

To validate the correlations of CD27 gene expression with biological and immune features in metastatic SKCM, we conducted the same analysis on 91 metastatic SKCM samples from the PRJEB23709. Consistent with the TCGA data, CD27 mRNA expression showed a significant positive correlation with the immune checkpoint genes PDCD1, TIGIT, LAG3, HAVCR2, IDO1, BTLA, CD274, and CTLA4 (Fig. 3A). We divided the 91 SKCM samples into high and low CD27 mRNA expression groups based on the median expression level. Consistent with findings in the TCGA SKCM cohort, we observed that the cell fractions of several immune cell types—including CD8 + T cells, activated memory CD4 + T cells, T follicular helper cells, activated NK cells, and M1 macrophages—were significantly elevated in the high CD27 expression group (Fig. 3B). Similarly, in these melanomas, immunosuppressive M2 macrophages were enriched in high CD27 expression group. Furthermore, we analyzed the expression patterns of four previously reported immunotherapy-responsive gene clusters and found a statistically significant upregulation of these genes in the high CD27 mRNA expression group (Fig. 3C). Single sample gene set enrichment scores of these gene sets, including a six-gene IFNγ signature (Fig. 3D), a related 18-gene IFNγ signature (Fig. 3E), an effector T-cell signature (Fig. 3F) and a combined IFNγ/effector T-cell signature (Fig. 3G), were significantly higher in the high expression group compared with the low expression group. The correlations of CD27 gene expression with biological and immune features were also found in 104 primary SKCM from TCGA, the detailed results can be found in Supplementary information.

Fig. 3: Validation of correlations of CD27 gene expression with biological and immune features in an independent cohort of 91 metastatic SKCMs from PRJEB23709.
Fig. 3: Validation of correlations of CD27 gene expression with biological and immune features in an independent cohort of 91 metastatic SKCMs from PRJEB23709.The alternative text for this image may have been generated using AI.
Full size image

A Correlation matrix plot shows correlations of CD27 mRNA expression with multiple checkpoint genes. B Box plot showing differences in the absolute proportions of immune cells in the tumor microenvironment between the high and low CD27 mRNA expression groups. C Heatmap displaying the expression of four immune therapy-related signature gene sets, including IFNγ-6, IFNγ expanded immune 18, effector T-cell and IFNγ/effector T-cell. D-G Box plots of gene set enrichment scores of IFNγ-6 (D), IFNγ expanded immune 18 (E), effector T-cell (F) and IFNγ/effector T-cell (G) in high and low CD27 groups of primary melanoma. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, no significance.

Validation of the associations of CD27 mRNA expression with prognosis and immunotherapy response in independent melanoma cohorts

To validate the prognostic value of CD27 mRNA expression, we performed overall survival analysis on a publicly independent dataset including 54 stage IV melanoma from GSE22153. As expected, patients with high CD27 mRNA expression levels had prolonged overall survival (p = 0.014, Fig. 4A). To further validate the correlation between CD27 mRNA expression and the response to immunotherapy, we performed survival analysis and area under the curve (AUC) analysis for the receiver operating characteristic (ROC) to compare sensitivity and specificity for prediction of therapy response on patients receiving melanoma immunotherapy in two public datasets. In 91 melanoma patients treated with PD-1/PD-L1 ICIs from PRJEB23709, patients were divided into high and low expression groups based on the median CD27 mRNA expression level (FPKM = 1.220795). The patients belonging to high expression group had significantly better overall survival (p = 0.00019, Fig. 4B) and progression-free survival (p = 0.0018, Fig. 4C) than patients with low CD27 mRNA expression. CD27 expression levels (FPKM) were also compared between poor responders and good responders among 91 melanoma patients treated with immunotherapy (PRJEB23709). Good responders showed significantly higher CD27 expression than poor responders, as indicated by the Wilcoxon test (Fig. 4D, p = 9.4e-06). ROC curves for CD27 expression, B-cell signature, and TLS signature in 91 melanoma patients treated with immunotherapy (PRJEB23709). CD27 showed an AUC of 0.7629, the B-cell signature showed an AUC of 0.7405, the TLS signature showed an AUC of 0.7677 and the IFNγ signature showed an AUC of 0.7473, indicating comparable predictive performance among the three biomarkers (Fig. 4E). In addition, correlation analysis indicates significant positive associations between CD27 expression and the B cell signature scores (Supplementary Fig. 3A, R = 0.64, p = 5.2e-12) and tertiary lymphoid structure signature scores (Supplementary Fig. 3B, R = 0.75, p = 2.2e-16) in patients from PRJEB23709. In 51 melanoma patients treated with PD-1 ICIs from GSE91061, due to the smaller sample size in GSE91061 and the presence of treatment factors, we employed a data-driven approach. ROC optimization was used to select the cut-off value, which was determined to be FPKM 7.738 of CD27 for this dataset (Supplementary Fig. 3C). The patients belonging to high expression group had significantly better overall survival (p = 0.038, Fig. 4F), Moreover, Good responders showed significantly higher CD27 expression than poor responders, as indicated by the Wilcoxon test (Fig. 4G, p = 0.023). ROC analysis of 51 melanoma patients treated with immunotherapy (GSE91061) shows that CD27 (AUC = 0.659), the B-cell signature (AUC = 0.6436), the TLS signature (AUC = 0.6795), and the IFNγ signature (AUC = 0.6377) have moderate predictive power, whereas tumor mutational burden (TMB) performs relatively poorly with the lowest AUC (0.5735). Similarly, CD27 expression also exhibits a significant positive correlation with the B cell (Supplementary Fig. 3D, R = 0.6, p = 5.5e-6) and tertiary lymphoid structure signature scores (Supplementary Fig. 3E, R = 0.6, p = 4.1e-6) in these patients.

Fig. 4: Associations of CD27 mRNA expression with prognosis and immunotherapy response in independent melanoma cohorts.
Fig. 4: Associations of CD27 mRNA expression with prognosis and immunotherapy response in independent melanoma cohorts.The alternative text for this image may have been generated using AI.
Full size image

A Kaplan–Meier survival curve depicting overall survival in 54 stage IV melanoma patients (GSE22153) stratified by high and low CD27 mRNA expression. B Overall Survival (OS) Kaplan-Meier curve for 91 immunotherapy-treated melanoma patients (PRJEB23709), stratified by CD27 mRNA expression (FPKM) at the median cut-off of 1.221 (High >= 1.220795; Low < 1.220795). C Progression-free survival (PFS) Kaplan–Meier curve for 91 immunotherapy-treated melanoma patients (PRJEB23709), stratified by CD27 mRNA expression (FPKM) using the median cut-off of 1.220795 (High >= 1.220795; Low < 1.220795). D Box plot shows the mRNA expression level of CD27 between poor and good responders (PRJEB23709). E The ROC curve demonstrating the ability of CD27 mRNA expression, B cell, TLS and IFNγ signatures to predict immunotherapy response in 91 melanoma patients (PRJEB23709). F OS Kaplan–Meier curve for 51 immunotherapy-treated melanoma patients (GSE91061), stratified by gene expression (FPKM) using the ROC-optimized cut-off of 7.738 (High > = 7.738; Low < 7.738). G Box plot shows the mRNA expression level of CD27 between poor and good responders (GSE91061). H The ROC curve demonstrating the ability of CD27 mRNA expression, B cell, TLS and IFNγ signatures, and TMB to predict immunotherapy response in 51 melanoma patients treated with immunotherapy (GSE91061). I Kaplan-Meier survival curve depicting progression-free survival in 102 melanoma patients treated with immunotherapy (FAHZZU) stratified by high and low CD27 mRNA expression by qRT-PCR. J Box plot showing differential expression of CD27 (qRT-PCR) between poor and good responders. K The ROC curve demonstrating the ability of CD27 mRNA expression (qRT-PCR) to predict immunotherapy response in the FAHZZU cohort. *p < 0.05, **p < 0.01, ****p < 0.0001.

Validation of association of CD27 mRNA expression with immunotherapy response in the FAHZZU cohort

To further validate the potential clinical application of CD27 mRNA expression in predicting the response to immunotherapy, we retrospectively collected clinical prognostic information and FFPE tumor tissues from 102 melanoma patients who received immunotherapy from FAHZZU. The mRNA expression levels of CD27 were quantitatively analyzed using qRT-PCR, and the patients were divided into high expression and low expression groups based on the median mRNA expression levels (0.006245 normalized ratio of PCR) (Supplementary Table 1). Survival analysis showed that melanoma patients in the high CD27 mRNA expression group had a significantly prolonged PFS compared with the low CD27 mRNA expression group (p = 0.033, Fig. 4I). Comparative analysis showed that good responders exhibited significantly higher mRNA expression levels of CD27 analyzed by qRT-PCR than those of poor responders (p = 0.0019, Fig. 4J). The mRNA expression level of CD27 analyzed by qRT-PCR achieved an AUC of 0.688 for predicting immunotherapy response in the FAHZZU cohort (Fig. 4K).

CD27 protein expression was associated with immune cell infiltration and immunotherapy response in the FAHZZU cohort

In accordance with established immunological characterization, CD27 has been documented as a surface receptor expressed across T lymphocyte, B lymphocyte populations and a subset of NK cells29, while CD45RO identifies human memory T cells30 and characterizes anti-tumor NK cells31. To validate this expression profile within the melanoma microenvironment, we conducted multiplex immunofluorescence staining assays targeting CD27 and CD45RO on three independent FFPE melanoma specimens. Imaging analysis revealed spatial co-localization of CD27 (pseudo-colored yellow) and CD45RO (pseudo-colored red) in specific cellular subsets (Fig. 5A and Supplementary Fig. 4A, B). This co-expression pattern provided histological confirmation that CD27-positive lymphocytes within melanoma tissues corresponded to either T cell or B cell lineages or NK cells, corroborating previous hematological findings in the context of solid tumor immunobiology. The patients were divided into high and low CD27 expression groups based on the H-scores analyzed by IHC (Supplementary Table 1). Hematoxylin and eosin (HE) and IHC staining revealed that tumors in the high CD27 expression group had more lymphocyte infiltration than those in the low CD27 expression group (Fig. 5B and Supplementary Fig. 4C). IHC staining of CD27 and PD-L1 was performed on tumor samples of 95 melanoma patients from the FAHZZU cohort. Patients with positive CD27 IHC staining had significantly better PFS than patients with negative CD27 IHC staining (p = 0.023, Fig. 5C). Comparative analysis showed that good responders exhibited significantly higher H-scores of CD27 analyzed by IHC than those of poor responders (p = 0.0067, Fig. 5D). The AUC of the expression level of CD27 analyzed by IHC for predicting PFS was much higher than that of PD-L1 (0.656 vs. 0.460, Fig. 5E), indicating that CD27 may be a better biomarker for predicting immunotherapy response in melanoma patients. Among CD27-positive patients (n = 54), the response rate was 55.56% (30/54), significantly higher than the 31.71% (13/41) observed in CD27-negative patients. While PD-L1-positive patients (n = 23) had a response rate of 52.17% (12/23), PD-L1-negative patients (n = 72) still exhibited a moderate response rate of 43.01% (31/72), indicating limited discriminative power. CD27 demonstrated a sensitivity of 69.8% in identifying good responders, outperforming PD-L1, which showed a sensitivity of only 27.9% (Fig. 5F), suggesting that CD27 may be a better biomarker for stratifying good responders in melanoma patients. For the 95 patients for whom complete data on tumor stage, sex, age, brain metastasis, ECOG score, LDH level and CD27 IHC were available, multivariate Cox regression analysis confirmed that CD27 IHC positive remained an independent prognostic factor for PFS (p = 0.035) in the FAHZZU cohort (Supplementary Fig. 4D).

Fig. 5: CD27 protein expression associated with immune cell infiltration and immunotherapy response.
Fig. 5: CD27 protein expression associated with immune cell infiltration and immunotherapy response.The alternative text for this image may have been generated using AI.
Full size image

A CD27/CD45RO dual fluorescence immunostaining confirming the co-localization of CD27 (yellow signal) and CD45RO (red signal) within T cells or NK cells. B Representative images showing H&E staining of tissue and cell sections, CD27 and PD-L1 IHC staining for the high and low CD27 expression groups. ×20, 20-fold original magnification; ×80, 80-fold original magnification. C Kaplan–Meier survival curve depicting progression-free survival in 102 melanoma patients treated with immunotherapy (FAHZZU) stratified by the high and low CD27 expression groups. D Box plot showing differential H-scores of CD27 (IHC) between the poor and good responders (FAHZZU cohort). E The ROC curve demonstrating the ability of the CD27 and PD-L1 H-scores (IHC) to predict immunotherapy response in the FAHZZU cohort. F Comparison of sensitivity and response rates based on CD27 and PD-L1 H-scores in melanoma patients treated with ICIs. **p < 0.01.

Discussion

Melanoma, a malignancy notorious for its aggressive behavior and propensity for immune evasion, has witnessed transformative advances through immunotherapy, particularly ICIs targeting PD-1, CTLA-4, and LAG31. Despite these strides, a substantial proportion of patients exhibit primary or acquired resistance to ICIs, underscoring the urgent need for biomarkers that predict therapeutic efficacy and guide personalized treatment3. CD27, a co-stimulatory molecule within the TNF receptor superfamily, has emerged as a pivotal regulator of T-cell activation and memory formation13. This study elucidated the prognostic and predictive value of CD27 expression in melanoma, demonstrating its association with an immunologically active TME, prolonged survival, and enhanced response to ICIs.

Our analysis revealed robust co-expression of CD27 with immune checkpoint genes, including PDCD1, TIGIT, LAG3, HAVCR2, IDO1, BTLA, CD274, and CTLA4, indicating that high CD27 mRNA expression may reflect an immunologically active TME conducive to better responses to ICIs. Furthermore, recent studies demonstrated that TIGIT32, PDCD133, HAVCR234, CTLA435, BTLA36, IDO137, and LAG338 expression is predictive for favorable prognosis and immunotherapy response in cancer. These genes are hallmark features of T-cell exhaustion but paradoxically signify an immunoreactive TME amenable to checkpoint blockade20. For instance, PDCD1 upregulation is linked to T-cell dysfunction, yet its presence correlates with clinical response to anti-PD-1 therapy, as PD-1+ T cells retain proliferative potential upon receptor blockade33. Similarly, co-inhibition of TIGIT and LAG3 synergizes with PD-1 blockade to reinvigorate exhausted T cells, a strategy under evaluation in clinical trials (NCT04370704, NCT03470922)39. The co-regulation of CD27 with these inhibitory receptors suggested a dual role: while chronic antigen exposure drives exhaustion markers, CD27 may sustain T-cell survival and effector function through CD70-mediated costimulation17. Preclinical models support this hypothesis, where CD27 agonism enhances CD8+ T-cell persistence and synergizes with anti-PD-1 therapy to suppress tumor growth19. These findings align with our observation that high CD27 mRNA expression correlated with prolonged OS and PFS in multiple cohorts.

High CD27 mRNA expression was associated with reduced tumor purity and increased infiltration of B cells, CD8+ T cells, CD4+ T cells, macrophages, neutrophils, and dendritic cells. CD8+ T cells are central to anti-tumor immunity, and their abundance predicts response to ICIs across cancer types40. B cells, particularly those within tertiary lymphoid structures (TLS), contribute to antigen presentation and antibody production, fostering durable immune responses11. In melanoma, TLS presence correlates with improved overall survival and response to PD-1 inhibitors, as demonstrated in the CheckMate-069 trial41. Dendritic cells, which exhibit a specialized antigen-presenting function and critical role for priming naïve T cells42, were enriched in CD27-high tumors, likely due to CD27-CD70 interactions that promote DC maturation and cross-presentation43. Macrophages, characterized by pro-inflammatory cytokine secretion (e.g., IL-12, TNF-α), further amplify anti-tumor immunity44. Interestingly, we observed a significant enrichment of immunosuppressive Tregs and M1 macrophages in the high CD27 expression group. This suggests a nuanced immune response in the tumor microenvironment, where the presence of immune-activating cells, such as CD8 + T cells, is counterbalanced by immunosuppressive cells. Tregs are known to play a pivotal role in immune evasion by suppressing the activity of effector T cells, thereby promoting tumor progression45. Similarly, M2 macrophages, which are typically associated with immunosuppressive functions, can promote tumor growth by fostering an immunosuppressive microenvironment46. These findings underscore the complexity of the immune response in SKCM and suggest that high CD27 expression may reflect a dynamic interaction between immune activation and immune suppression, which could influence prognosis and therapeutic responses. Notably, CD27’s role in sustaining memory T-cell immunity may explain the observed infiltration of CD45RO+ memory T cells in our multiplex immunofluorescence assays47. This immune-rich milieu, driven by CD27, may create a permissive environment for ICIs, as evidenced by the superior PFS in CD27-high patients. The association between high CD27 mRNA expression and increased immune cell infiltration suggested that CD27 may play a critical role in orchestrating immune responses within the TME, potentially enhancing the efficacy of ICIs. Interestingly, these findings appear to contrast with recent data showing that elevated baseline plasma levels of soluble CD27 (sCD27) are associated with primary resistance to anti-PD-1 monotherapy, but not to combined anti-PD-1/anti–CTLA-4 therapy, in melanoma patients48,49. A plausible explanation is that tumoral CD27 mRNA and circulating sCD27 capture distinct aspects of the CD27–CD70 axis: high intratumoral CD27 likely reflects the presence of activated and memory T cells, TLS-associated B cells, and professional antigen-presenting cells within an immunologically “hot” TME, whereas increased sCD27 may instead indicate chronic systemic CD27-CD70 engagement, T-cell exhaustion or apoptosis, and more advanced, treatment-resistant disease. In addition, differences in biomarker compartment (tumor tissue vs. plasma), timing of sampling, treatment regimens (anti–PD-1 alone vs combination immunotherapy), and patient selection between studies may further contribute to these apparently divergent associations. In our retrospective cohort, plasma was not systematically collected, and we were therefore unable to directly assess the relationship between sCD27 and tumoral CD27 expression. Future studies integrating both tissue and circulating CD27 measurements will be important to delineate how these complementary biomarkers can be combined to optimize immunotherapy stratification.

Current biomarkers for melanoma immunotherapy, including PD-L1 expression, tumor mutational burden (TMB), IFNγ signatures, B cell signature and TLS signature, have limitations in sensitivity or practicality9. PD-L1 protein expression analyzed by IHC has not been approved as a predictive biomarker for melanoma immunotherapy due to inconsistent correlation with treatment outcomes and lack of standardized testing methods10. TMB assessment lacks standardization, leading to variability in measurement and interpretation across different sequencing methods, such as whole exome sequencing (WES) and targeted next-generation sequencing (NGS) panels. This inconsistency complicates the establishment of universal TMB thresholds for predicting immunotherapy responses. Additionally, TMB testing can be costly, posing challenges for its widespread implementation in clinical practice50. Our comparative analysis showed that CD27 expression has superior predictive power for immunotherapy response compared with TMB in the GSE91061 cohort (51 melanomas). IFNγ signatures are utilized to predict responses to immunotherapy based on the mRNA expression of multiple genes26. Our comparative analysis also demonstrated that CD27 expression has predictive performance comparable to IFNγ signatures for immunotherapy response across two independent datasets. Previous studies have demonstrated that B cells and tertiary lymphoid structures (TLS) are strongly associated with immunotherapy responses41,51,52. In our analysis, we compared the predictive performance of CD27 gene expression with that of the B cell and TLS signatures. We found that CD27 exhibited similar predictive power to both signatures, with a high degree of correlation (R = 0.64 for B cells, R = 0.75 for TLS; p < 0.001 for both), suggesting a significant relationship. Based on these findings, we hypothesize that CD27 expression may serve as a marker reflecting the activity of B cells and TLS in the tumor microenvironment. However, in clinical practice, RNA degradation in formalin-fixed, paraffin-embedded (FFPE) tissues poses significant challenges for accurate quantification. The fixation and embedding processes can lead to RNA fragmentation and chemical modifications, complicating reliable mRNA expression analysis53. In contrast, CD27 IHC demonstrated robust prognostic performance in our cohort (AUC = 0.656 for PFS prediction), outperforming PD-L1 (AUC = 0.460). CD27 also demonstrated a sensitivity of 69.8% in identifying responders, outperforming PD-L1, which showed a sensitivity of only 27.9%. Recent studies have revealed significant inconsistencies in the predictive value of PD-L1 immunohistochemistry for immunotherapy response in melanoma. A study from Denmark suggested that PD-L1 serves as a biomarker of real-world clinical outcomes for anti-CTLA-4 in combination with anti-PD-1 or anti-PD-1 monotherapy in metastatic melanoma54. In contrast, a German study found that tumor PD-L1 expression in skin or subcutaneous metastases showed no correlation with immunotherapy outcome. Notably, both studies were based on European populations55. In our study, the FAHZZU cohort, which consists of Asian patients, served as an independent external validation cohort and similarly showed that PD-L1 expression did not effectively predict immunotherapy response, consistent with the findings from the German cohort. This may be attributed to the fact that all melanoma cases in our study were derived from cutaneous origins. Interestingly, CD27 immunohistochemistry demonstrated superior predictive performance compared to PD-L1, suggesting its potential as a more sensitive biomarker for identifying responders in melanoma immunotherapy. This superiority stems from CD27’s stability in FFPE tissues and its reflection of both immune activation and suppressive signals within the TME53. Furthermore, CD27’s association with IFNγ signatures and effector T-cell signatures provided mechanistic insight into its predictive value. These attributes may position CD27 as a pragmatic biomarker for routine clinical use, particularly in resource-limited settings where advanced genomic profiling is unavailable. Additionally, due to the protein stability in FFPE tissues, CD27 could be more practical than RNA-based biomarkers, which are prone to degradation.

Beyond its prognostic utility, CD27 represents a compelling therapeutic target. Agonistic anti-CD27 antibodies (e.g., varlilumab) have shown promise in early-phase trials, enhancing T-cell activation and synergizing with PD-1 blockade in preclinical models18. In a phase I trial (NCT02335918), varlilumab combined with nivolumab induced durable responses in PD-1-naïve melanoma patients, with a disease control rate of 33%56. Our data supported this approach, as CD27-high tumors exhibited elevated expression of co-stimulatory molecules (e.g., GZMA, PRF1) and chemokines (CXCL9, CXCL10) that recruit effector T cells. Our findings provided a strong rationale for the development of combination therapies targeting CD27, which may enhance the efficacy of existing immunotherapies and overcome resistance mechanisms.

Despite the promising findings, there are several limitations to this study that warrant consideration. First, the retrospective nature of the study design may introduce biases, such as selection bias, which could affect the generalizability of the results. Second, while we used multiple independent datasets to validate our findings, prospective studies with larger sample sizes are needed to further confirm the prognostic and predictive value of CD27 expression in melanoma. Third, the functional mechanisms underlying CD27-mediated immune regulation in the TME remain to be fully elucidated. Future studies should focus on understanding the molecular pathways regulated by CD27 and how they interact with other immune checkpoints and co-stimulatory molecules.

In conclusion, CD27 expression is a robust and practical biomarker for predicting immunotherapy response and prognosis in melanoma. Its integration into clinical workflows, via IHC or mRNA assays, could refine patient stratification and identify candidates for CD27-targeted therapies. By fostering an immunologically active TME and synergizing with existing ICIs, CD27 may serve as a potential therapeutic target, and combination strategies involving CD27 agonizts and ICIs may offer new opportunities to improve outcomes of melanoma patients. Future prospective studies and clinical trials are needed to validate these findings and explore the therapeutic potential of targeting CD27 in combination with conventional immunotherapies.

Methods

Sample collection and resources

The analyzed results in this study are partly based on skin cutaneous tumor (SKCM) data generated by The Cancer Genome Atlas Research Network (TCGA [RRID:SCR_003193], http://cancergenome.nih.gov/). RNA-seq and clinicopathological data of 470 SKCM samples were collected from TCGA, comprising 366 metastatic and 104 primary SKCM tissues. These metastatic and primary tumors were analyzed as two independent cohorts to explore whether the observed associations were shared or distinct across different disease stages. Among 366 metastatic melanomas, 352 had overall survival (OS) data, 346 had disease-specific survival (DSS) data, and 353 had progression-free survival (PFS) data. Our study utilizes the TCGA Pan-Cancer Clinical Data Resource (TCGA-CDR)57 to standardize survival outcome analysis. OS is the period from the date of diagnosis until the date of death from any cause. The censored time is from date of initial diagnosis until the date of last contact (largest number of days) from all the clinical data files (including both enrollment and follow-up forms). DSS is defined as the time from initial diagnosis until death from the diagnosed cancer, which is considered more relevant to cancer biology and therapeutic impact. However, due to TCGA dataset limitations, DSS events were approximated by including patients classified as “Dead” and “With Tumor”. The Progression-Free Interval (PFI) is used as a surrogate for Progression-Free Survival (PFS); it is defined as the time from diagnosis until the first event of disease recurrence, progression, new primary tumor development, or death with tumor. Since these events typically occur earlier, PFS/PFI is often favored in clinical studies requiring shorter follow-up times.

In this study, another three different public datasets were utilized, and the clinical-pathological characteristics of the patients in these datasets were available in Supplementary Table 1.

Expression profiles generated by the Illumina human-6 v2.0 expression beadchip platform and clinical data of 54 stage IV melanomas were required from Gene Expression Omnibus (GEO) (RRID:SCR_005012) database with the accession number of GSE2215358. In this dataset, melanoma patients with distant metastases were enrolled in a prospective study from year 2000 and onward, including 29 males and 25 females with a median age of 65 years (range 25–86 years), and the definition of OS is consistent with that in the TCGA dataset.

RNA-seq data of tumor tissues derived from 91 melanoma patients treated with immunotherapy were obtained from the European Nucleotide Archive by accession number PRJEB23709, and the clinical summary of patients was available in Table S2 of the original paper59. In this dataset, melanoma patients were retrospectively identified based on formalin-fixed paraffin-embedded (FFPE) tissue availability, including 60 males and 31 females with a median age of 61 years (range 24–90 years), and the OS was defined as the time from the initiation of immunotherapy to death from any cause, while the PFS was defined as the time from immunotherapy initiation to the first documented disease progression (as per RECIST 1.1) or death from any cause, whichever occurred first.

The RNA-seq, tumor mutation burden (TMB) and clinical data of 51 pretreatment melanoma patients treated with anti-PD-1 inhibitors were collected from GEO with the accession number of GSE91061. In this dataset, melanoma patients were accrued to a multi-arm, multi-institutional, institutional-reviewboard-approved, prospective study (CA209-038; NCT01621490) with a median age of 55 years (range 22–89 years), and the OS was defined as the time from the initiation of immunotherapy to death from any cause60.

Patients’ collection at the First Affiliated Hospital of Zhengzhou University

For the in-house validation cohort, we retrospectively collected formalin-fixed, paraffin-embedded (FFPE) blocks of tumors and clinical information derived from 102 melanoma patients, who had biopsy or surgery tumor samples and were treated with PD-1/PD-L1 ICIs at the First Affiliated Hospital of Zhengzhou University (FAHZZU) from January 2017 to December 2023. The 8th edition of the AJCC staging system was used for classifying the stages of melanoma. Response assessment was performed retrospectively by Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST 1.1)61. Good responders were defined as those with a RECIST CR, PR, or SD for more than 6 months; poor responders had RECIST PD as best response or SD for 6 months or less. The detailed information about patient demographics and clinical characteristics, including gender, age, cancer stage, PFS time and immunotherapy response, was provided in Supplementary Table 1. PFS was defined as the time from immunotherapy initiation to the first documented disease progression (as per RECIST 1.1) or death from any cause, whichever occurred first. Patient inclusion and sample analyses at FAHZZU were approved by the Institutional Review Board (IRB) of the FAHZZU (Project ID 2024-KY-0801-002). All procedures performed in this study were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki Declaration and its later amendments.

Whole transcriptome comparison analysis

The raw gene counts of 470 SKCM samples were downloaded from TCGA. Then we performed differential expression analysis between high and low CD27 expression groups with DESeq2 (RRID:SCR_015687)62. Genes with the Benjamini-Hochberg adjusted p value less than 0.05 and log2 fold-change more than 1.0 were considered differentially expressed.

Gene set enrichment analysis (RRID:SCR_003199)

We collected 50 hallmark gene sets and 186 KEGG pathway signatures from the Molecular Signatures Database (MSigDB v2022.1.Hs, https://www.gsea-msigdb.org/gsea/msigdb)63. Based on high and low expression groups, gene set enrichment analysis was carried out to identify CD27 expression-related gene sets with clusterProfiler (RRID:SCR_016884)64. The predictive signatures of immunotherapy, such as a six-gene interferon-γ (IFNγ) signature (IFNγ-6), a related 18-gene expanded immune signature (IFNγ expanded immune 18)26, an effector T-cell signature (effector T-cell)27 and a combined IFNγ/effector T-cell signature28, B cell signature11 and tertiary lymphoid structure(TLS)41 were collected and used to calculate single-sample gene set enrichment scores of each sample with the GSVA (RRID:SCR_021058) R package65.

Immune cell infiltration estimation

TIMER (RRID:SCR_018737) 2.0 web application66 was used to estimate the correlations of CD27 gene expression with the tumor purity and infiltration levels of B cells, CD8+ T cells, CD4+ T cells, macrophages, neutrophils and dendritic cells. The absolute fraction data of 22 infiltrating immune cell types, which were inferred by the CIBERSORT (RRID:SCR_016955) algorithm23 based on gene expression profiles, were downloaded from the TIMER database66 (http://timer.cistrome.org/infiltration_estimation_for_tcga.csv.gz). And the leukocyte fraction data (TCGA_all_leuk_estimate.masked.20170107.tsv), which was estimated based on DNA methylation in Thorsson’s study67, was retrieved from https://gdc.cancer.gov/about-data/publications/panimmune. The TIL percentage evaluated by pathological images of TCGA tumors via a deep learning method is available in the supplementary table in Saltz’s study68. Then we compared immune cell infiltration, leukocyte fraction and TILs between high and low gene expression groups.

Quantitative reverse transcription-PCR (qRT-PCR)

qRT-PCR analysis was employed to assess CD27 mRNA expression levels across 102 FFPE melanoma specimens from the FAHZZU cohort. Total RNA was isolated using the FFPE DNA/RNA Extraction Kit (AmoyDx, Xiamen, China) following the manufacturer’s protocols. Subsequently, complementary DNA (cDNA) synthesis was performed by reverse-transcribing 500 ng of purified RNA using HiScript III RT SuperMix for qPCR (+gDNA wiper) (Vazyme, Nanjing, China) following the manufacturer’s protocols. Amplification reactions were conducted in 20 µL volumes containing Taq Pro Universal SYBR qPCR Master Mix (Vazyme, Nanjing, China), with each reaction comprising 20 ng of cDNA template and 0.2 µM each primer (CD27 forward primer: actactgggctcagggaaagct, CD27 reverse primer: ggatcacactgagcagcctttc). The housekeeping genes ACTB and GAPDH were used as references for normalization (ACTB forward primer: atgtggccgaggactttgatt, ACTB reverse primer: agtggggtggcttttaggatg; GAPDH forward primer: tgcaccaccaactgcttagc, GAPDH reverse primer: ggcatggactgtggtcatgag). qRT-PCR was carried out using a Bio-Rad CFX96 real-time PCR detection system (Bio-Rad Laboratories [RRID:SCR_008426], Hercules, CA) with the following temperature profile: 30 s at 95 °C and 40 cycles with 10 s at 95 °C, and 30 s at 60 °C. Relative CD27 expression levels were calculated using the ΔCT method.

Multiplex immunofluorescence and image acquisition

The multiplex immunofluorescence staining procedure was conducted utilizing a 3-color fluorescence immunohistochemistry kit (PhenoVision Bio Co., Ltd.) in accordance with the manufacturer’s protocol. Briefly, tissue sections underwent a series of preparatory steps, including deparaffinization, rehydration, and immersion in citrate buffer (pH 6.0) for antigen retrieval. Subsequently, the sections were subjected to washing steps followed by sequential incubation with H2O2 and blocking reagents, each for a duration of 10 min. Primary antibodies (CD27 from Proteintech and CD45RO from ZSbio) were then applied to the sections and incubated at room temperature for 30 min. The detection of bound primary antibodies was achieved through the application of PVB anti-Rb/Mm-HRP detection reagent, followed by signal amplification using PVB tyramide-conjugated fluorophores (PVB 520 and PVB 570) for 10 min. Nuclear counterstaining was performed with 4′,6-diamidino-2-phenylindole (PhenoVision Bio Co., Ltd.), and the sections were subsequently coverslipped. Finally, the prepared slides were imaged using the PhenoImage HT system (Akoya Biosciences) at a magnification of ×200.

Immunohistochemical (IHC) Analysis for CD27

FFPE tissue sections of 4-μm thickness were stained for CD27 using an anti-human CD27 rabbit monoclonal antibody (anti-CD27, 66308-1-Ig, Proteintech). The staining procedure was carried out on the Roche (Ventana BenchMark ULTRA) immunohistochemistry/ISH Automatic Staining Module. The Roche Ventana UV AP-Red kit was employed as the detection system. The staining protocol included the following sequential steps: (1) EZ deparaffinization, 12 min; (2) CC1, 64 min; (3) incubation with primary antibody (CD27), 48 min; (4) UV Red Univ Mult, 12 min; (5) UV Red Enhancer, 4 min; (6) UV Fast Red A/UV Red Naphthol, 8 min; (7) UV Fast Red B, 8 min; (8) counterstaining with Mayer’s hematoxylin, 12 min; and (9) post-counterstaining with bluing reagent, 4 min. For the evaluation of CD27 protein expression, a semi-quantitative H-score method was applied for IHC staining images. The H-score was calculated using the formula: (percentage of weakly staining cells) + (percentage of moderately staining cells × 2) + (percentage of strongly staining cells × 3)69.

Statistical analysis

All statistical analyses were carried out in R software (RRID:SCR_001905). Correlations between two variables were calculated by Pearson correlation analysis with cor.test function in R. Continuous value comparisons between two groups were tested with Wilcoxon–Mann–Whitney U test. Kaplan–Meier estimate and log-rank testing were used to conduct the survival analysis for OS, PFS, and DSS. The multivariate Cox proportional hazard model was used to investigate the association of the combination of CD27 gene expression, clinical tumor stage, age, and gender with survival.