Multi-cohort validation of BRAF-PTPRT co-mutation as a predictor of favorable immunotherapy response in melanoma

Abstract

Background

Immune checkpoint blockade (ICB) has transformed melanoma therapy, but predictive biomarkers remain suboptimal. BRAF is the most common driver mutation, yet its single-gene association with ICB response remains inconsistent. PTPRT loss activates STAT3-driven immunity. We hypothesized that concurrent BRAF and PTPRT mutations (co-mutation) synergistically enhance immunogenicity and predict ICB benefit.

Methods

We analyzed three independent cohorts: Memorial Sloan Kettering Cancer Center (MSK) (n=320, ICB-treated), The Cancer Genome Atlas (TCGA) (n=448, mixed treatments), and a combined ICB cohort (CS, n=436). Patients were categorized into four groups: co-mutation, BRAF+/PTPRT−, BRAF−/PTPRT+, double wild-type (WT). Overall survival (OS), tumor mutational burden (TMB), neo-antigen load, clinical benefit, immune infiltration, and transcriptomic pathways were evaluated. A simplified risk score (0–2) was constructed by assigning 1 point for co-mutation and 1 point for high TMB (above median) to predict clinical benefit.

Results

In ICB-treated cohorts (MSK and CS), only co-mutation consistently predicted superior OS, while single mutations showed no consistent benefit. Co-mutated tumors exhibited significantly higher TMB and neo-antigen loads, an immune-inflamed microenvironment with elevated M1 macrophages, reduced Tregs, and upregulation of programmed death-ligand 1 (PD-L1), CTLA-4, TIM-3, as well as enrichment of IFN-γ and IL6-JAK-STAT3 pathways. The combined risk score (co-mutation + TMB) predicted clinical benefit with an area under the curve (AUC) of 0.659. Patients with high-risk score (2 points) demonstrated significantly higher clinical benefit rates compared to low-risk score (0 points).

Conclusions

BRAF-PTPRT co-mutation is a robust, cross-cohort predictive biomarker for ICB response in melanoma. A simple risk score integrating co-mutation and TMB further improves patient stratification. Routine testing of both genes may guide immunotherapy decisions.

Highlight box.

Key findings

• BRAF-PTPRT co-mutation defines a melanoma subtype with enhanced immune checkpoint blockade (ICB) sensitivity via synergistic tumor mutational burden (TMB)-driven immunogenicity and STAT3-mediated immune activation.

What is known and what is new?

• BRAF single mutation alone has inconsistent predictive value.

• Concurrent PTPRT loss and STAT3 activation, synergizing with TMB, establishes a novel predictive biomarker.

What is the implication, and what should change now?

• Our findings advocate for incorporating PTPRT testing into routine sequencing panels for BRAF-mutant melanoma, suggesting that co-mutant patients may achieve superior outcomes with frontline ICB rather than targeted therapy alone.

Introduction

Immune checkpoint blockade (ICB) has transformed the treatment of advanced melanoma, yet durable clinical benefit is limited to a subset of patients, and reliable predictive biomarkers remain an unmet need (1-3). Tumor mutational burden (TMB) and PD-L1 expression have limited predictive accuracy in melanoma, underscoring the importance of discovering more robust biomarkers (4-6).

BRAF is the most common driver mutation in melanoma (7,8). Although BRAF-targeted therapy achieves high response rates, acquired resistance inevitably develops (9,10). The role of mutation alone as a predictor of ICB outcome is controversial: some studies report higher TMB and potential benefit, while others find no significant association (11). This inconsistency reflects the complex, context-dependent impact of BRAF mutation on the tumor immune microenvironment, which may be modified by co-occurring genetic alterations. PTPRT encodes a transmembrane phosphatase that negatively regulates STAT3 signaling (12). Loss-of-function mutations in PTPRT lead to constitutive STAT3 activation, upregulating T-cell-recruiting chemokines (e.g., CXCL9/10/11) and promoting an immune-inflamed microenvironment (13,14). Recent studies have associated PTPRT mutation with improved ICB response in several cancers (15,16), but its clinical value as an independent biomarker remains unvalidated.

Importantly, our preliminary three-cohort analysis revealed that single-gene associations with overall survival (OS) are inconsistent and context-dependent: BRAF mutation alone was associated with better survival only in the non-immunotherapy The Cancer Genome Atlas (TCGA) cohort, whereas PTPRT mutation alone showed benefit only in immunotherapy-treated cohorts. This heterogeneity indicates that neither single mutation serves as a robust biomarker. Moreover, BRAF and PTPRT are the two most frequently mutated coding genes in the Memorial Sloan Kettering Cancer Center (MSK) cohort (36% and 27%, respectively) (17). Biologically, BRAF provides moderately elevated TMB (“fuel” for immunogenicity), while PTPRT loss drives STAT3-mediated immune activation (“engine” for T-cell recruitment). We therefore hypothesized that concurrent BRAF and PTPRT mutations (co-mutation) synergistically combine both features, defining a melanoma subset with exceptional ICB sensitivity and a stable, cross-cohort predictive signature. Validating this hypothesis fills several critical gaps: it explains the heterogeneous ICB responses of BRAF-mutant melanomas, establishes PTPRT mutation as a clinically actionable biomarker only in the context of BRAF co-mutation, and provides a mechanistic framework linking TMB to immune activation. Importantly, this co-mutation signature can be easily obtained from routine next-generation sequencing panels, offering an immediate clinical application.

To test this hypothesis, we integrated three independent cohorts and assessed survival, TMB, neo-antigen load, clinical benefit, immune microenvironment, and transcriptomic pathways. Our results demonstrate that BRAF-PTPRT co-mutation is a robust, treatment-specific predictive biomarker, and a simple risk score combining co-mutation and TMB further improves patient stratification. We present this article in accordance with the REMARK reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1745/rc).

Methods

Patient cohorts and data sources

We analyzed three independent melanoma cohorts.

MSK immunotherapy cohort

Genomic and clinical data were obtained from the “TMB and immunotherapy (MSK, Nat Genet2019)” study available through the immunogenomics module of cBioPortal for Cancer Genomics (https://www.cbioportal.org/). This study originally included 1,661 pan-cancer patients treated with immune checkpoint inhibitors (ICIs), with tumor-normal sequencing performed using the MSK-IMPACT panel. From this pan-cancer cohort, we selected 320 melanoma patients who received anti-PD-1, anti-CTLA-4, or combination therapy. All selected patients had available somatic mutation data, TMB, and OS information.

TCGA cohort

RNA-sequencing, somatic mutation, TMB, and clinical data of 448 skin cutaneous melanoma (SKCM) patients were downloaded from TCGA via cBioPortal. This cohort included patients who received various therapies (surgery, radiotherapy, chemotherapy, and a small fraction receiving immunotherapy). A subset of patients treated with radiotherapy alone (TCGA-RT) was used as a negative control to evaluate treatment-specificity.

Combined study cohort (CS cohort)

We constructed a combined immunotherapy cohort by integrating six independent melanoma datasets from the cBioPortal immunogenomics module: Metastatic Melanoma (DFCI, Science 2015; n=110), Melanoma (MSK, NEJM 2014; n=64), Metastatic Melanoma (UCLA, Cell 2016; n=38), Metastatic Melanoma (DFCI, Nat Med. 2019)—iAtlas Harmonized (n=122), Metastatic Melanoma (Immunotherapy Trial, Cancer Cell 2019)—iAtlas (n=91), and Metastatic Melanoma (UCLA, Cell 2016)—iAtlas Harmonized (n=30). All patients received ICIs. After harmonizing mutation calls and clinical variables, 436 patients with complete data on TMB, survival, response assessment, and BRAF/PTPRT mutation status were retained. Additional variables included neo-antigen load, clinical response labels [complete response (CR), partial response (PR), stable disease (SD), long-term benefit (LB), no benefit (NB), progressive disease (PD)], serum lactate dehydrogenase, and metastasis status. All data were de-identified and used in accordance with the Declaration of Helsinki and its subsequent amendments; institutional review board approval was waived due to the retrospective, publicly available nature of the data.

Mutation classification and co-mutation definition

Somatic non-synonymous mutations (missense, nonsense, frameshift, and splice-site variants) were extracted for each patient. Based on BRAF and PTPRT mutation status, patients were divided into four mutually exclusive groups: Co-mutation: presence of both BRAF and PTPRT mutations; BRAF+/PTPRT−: BRAF mutation with PTPRT wild-type (WT); BRAF−/PTPRT+: PTPRT mutation with BRAF WT; double WT: no mutation in either gene.

For single-gene analyses, patients carrying the respective mutation (regardless of the other gene) were compared to those without that mutation.

TMB and neo-antigen load

TMB was defined as the total number of somatic non-synonymous mutations per megabase (mutations/Mb) of the sequenced coding region. For the MSK and TCGA cohorts, TMB values were directly obtained from cBioPortal. For the CS cohort, TMB was recalculated using the same definition. Neo-antigen load (number of predicted neo-epitopes) was available for the CS cohort and was log-transformed when necessary to reduce skewness.

Immune cell deconvolution and transcriptomic analysis (TCGA)

For the TCGA cohort, gene expression data (RNA-seq, FPKM normalized) were retrieved. The immune cell composition was estimated using the CIBERSORT and single-sample gene set enrichment analysis (ssGSEA) algorithm with the immune cell subsets. Differentially expressed genes (DEGs) between co-mutant and WT tumors were identified using the DESeq2 package (version 1.36.0) in R. Genes with |log₂ fold change| >0.58 and adjusted P value (FDR) <0.05 were considered significantly different. Unsupervised hierarchical clustering of DEGs was performed on Z-scored expression values using the pheatmap package. Gene set enrichment analysis (GSEA) was performed using the clusterProfiler package (version 4.4.0) with the Hallmark gene sets (MSigDB v7.4). Statistical significance was set at FDR <0.05, and pathways with |normalized enrichment score (NES)| >1.2 were reported.

Clinical response definitions (CS cohort)

The original studies in the CS cohort provided categorical labels for each patient’s best response or clinical benefit status. These labels included: CR, PR, SD, LB, NB, and PD. The specific time points of response assessment were not uniformly reported across all contributing datasets; therefore, we did not apply any additional time-based criteria. To evaluate the predictive value of the co-mutation and TMB risk score, we defined clinical benefit as CR, PR, SD, or LB (i.e., “effective”), and no benefit as NB or PD (i.e., “ineffective”). This binary classification was used as the endpoint for receiver operating characteristic (ROC) analysis and for comparing outcome distributions between groups. To generate a clinically interpretable predictive tool, we constructed a simplified risk score in the CS cohort using co-mutation status and TMB level. One point was assigned for the presence of BRAF-PTPRT co-mutation (score 1, else 0) and one point for high TMB, defined as TMB ≥ cohort median (score 1, else 0). The combined score ranged from 0 to 2. The performance of this score for predicting clinical benefit (effective vs. ineffective) was assessed by ROC curve analysis using the pROC package (version 1.18.0) in R. The area under the curve (AUC) was calculated with 95% confidence intervals (CIs). Patients were then stratified into three groups: low-risk (score 0), intermediate-risk (score 1), and high-risk (score 2).

Statistical analysis

Continuous variables were compared using the Wilcoxon rank-sum test (two groups) or Kruskal-Wallis test (multiple groups). Categorical variables were compared using Fisher’s exact test or the Chi-squared test, as appropriate. OS was defined as the time from start of immunotherapy (for MSK and CS cohorts) or from diagnosis (for TCGA) to death from any cause. Patients alive at last follow-up were censored. Survival curves were estimated by the Kaplan-Meier method and compared by the log-rank test. Univariable and multivariable Cox proportional hazards regression models were used to evaluate prognostic factors. Multivariable models were adjusted for four groups, age, sex, TMB. Results were expressed as hazard ratios (HRs) with 95% CIs. All tests were two-sided, and P<0.05 was considered statistically significant. Statistical analyses were performed using R software (version 4.2.2) with packages survival, survminer, ggplot2, pheatmap, clusterProfiler, DESeq2, pROC. Figures were generated using ggplot2 and pheatmap.

Results

Landscape of genomic alterations in the MSK immunotherapy cohort

To characterize the genomic mutational landscape of the MSK immunotherapy cohort, we analyzed somatic non-synonymous mutations from targeted sequencing data (Figure 1). Among 313 patients, 262 (83.7%) harbored at least one mutation. The most frequently mutated genes were established melanoma drivers, with BRAF (36%) and PTPRT (27%) being the top two, followed by NRAS (24%) and NF1 (23%). Missense mutations predominated across these genes, consistent with their roles as driver events in melanoma pathogenesis. TMB varied widely across the cohort, as shown in the upper panel of the waterfall plot. These data provide a comprehensive overview of the genomic background of the MSK cohort, forming the basis for subsequent analyses of the prognostic impact of BRAF and PTPRT alterations.

Figure 1.

Mutational landscape of the MSK immunotherapy cohort. (A) Waterfall plot of non-synonymous somatic mutations in the MSK cohort (n=313). (B) Mutation subtype distribution of BRAF and PTPRT at the patient level. MSK, Memorial Sloan Kettering Cancer Center; TMB, tumor mutational burden.

We next dissected the mutation subtypes of BRAF and PTPRT at the patient level in the MSK cohort (Figure 1B). For BRAF, V600E accounted for 44.2% (80/181) of BRAF-mutant patients, V600K for 7.2% (13/181), other V600 for 1.7% (3/181), and non-V600 mutations for 47.0% (85/181). Thirteen patients carried multiple BRAF mutations; after excluding these, V600E represented 46.0% (75/163) and V600-series mutations comprised 51.5% (84/163). The relatively high non-V600 proportion likely reflects passenger mutations captured by the MSK-IMPACT full-gene sequencing. For PTPRT, loss-of-function mutations were observed in 12.3% (30/244), while missense mutations accounted for 87.7% (214/244). Among the latter, 29.9% (64/214) localized to the phosphatase domain (amino acids 800–1200) and may have functional implications, whereas 70.1% (150/214) fell outside this domain and are more likely passenger events. These findings confirm that the co-mutation signature is primarily driven by functionally relevant alterations, with passenger mutations only diluting, rather than inflating, the observed effect.

Prognostic impact of BRAF/PTPRT co-mutation in immunotherapy-treated patients

To determine whether BRAF/PTPRT co-mutation predicts survival in patients receiving ICB, we analyzed three independent cohorts, including two immunotherapy-treated cohorts (MSK and CS) and the TCGA cohort. Univariate analysis of single-gene mutations revealed context-dependent associations (Figure 2A). In the MSK cohort, PTPRT single mutation was associated with better OS (HR =0.59, P=0.02), while BRAF single mutation was not (HR =0.72, P=0.08). In TCGA, BRAF single mutation showed a significant survival benefit (HR =0.74, P=0.03), whereas PTPRT single mutation did not (HR =0.83, P=0.22). In the CS cohort, only PTPRT single mutation was significant (HR =0.71, P=0.04). When patients were stratified into four groups (co-mutation, BRAF+/PTPRT−, BRAF−/PTPRT+, WT), only co-mutation consistently predicted better OS across all three cohorts (Figure 2B). In MSK, co-mutation yielded HR =0.40 (P=0.02); in TCGA, HR =0.62 (P=0.02); and in CS, HR =0.49 (P=0.02). Kaplan-Meier curves confirmed these survival advantages (Figure 2C). Multivariable Cox regression adjusted for age, sex, and TMB showed that co-mutation remained independently associated with improved OS in the two immunotherapy cohorts: MSK (HR =0.438, P=0.046) and CS (HR =0.445, P=0.039). In TCGA, a consistent trend was observed (HR =0.655, P=0.051) (Figure 2D-2F). Taken together, BRAF/PTPRT co-mutation is a robust and independent prognostic biomarker for favorable survival in patients receiving immunotherapy.

Figure 2.

Prognostic impact of BRAF/PTPRT co-mutation across three cohorts. (A) Forest plot of univariable Cox regression for single-gene mutations (BRAF or PTPRT) versus wild-type in each cohort. (B) Forest plot of univariable Cox regression for co-mutation versus wild-type in each cohort. (C) Kaplan-Meier curves comparing overall survival between co-mutant and wild-type patients in each cohort. (D-F) Univariable and multivariable Cox regression forest plots for the MSK (D), TCGA (E), and CS (F) cohorts, adjusted for age, sex, and TMB. CS, combined study; MSK, Memorial Sloan Kettering Cancer Center; TCGA, The Cancer Genome Atlas; TMB, tumor mutational burden.

Association of BRAF/PTPRT co-mutation with TMB and its synergistic prognostic effect

We next examined the relationship between BRAF/PTPRT co-mutation and TMB across three cohorts. Co-mutated tumors consistently showed significantly higher TMB than WT tumors in all three cohorts (all P<0.001, Figure 3A). In immunotherapy-treated cohorts (MSK and CS), high TMB was associated with significantly better OS (P<0.001; Figure 3B). However, this survival benefit was not observed in the TCGA cohort, even when restricting to radiotherapy-treated patients (TCGA-RT) or comparing co-mutant versus WT patients (Figure 3C). To assess synergy, patients were stratified by co-mutation status and TMB level (Figure 3D). Those with both co-mutation and high TMB had the most favorable OS, whereas WT/low-TMB patients had the worst prognosis (P<0.001). To explore the potential synergistic effect of BRAF/PTPRT co-mutation and TMB, we stratified immunotherapy-treated patients (MSK + CS) into four subgroups by co-mutation status and TMB level (Figure 3D). Patients with both co-mutation and high TMB had the most favorable OS, whereas WT/low-TMB patients had the worst prognosis (P<0.001). Subgroup analyses further dissected this combined effect (Figure 3E-3H): in the TMB-high subgroup, co-mutant patients survived significantly longer than WT patients (HR =0.54, P=0.04); however, no survival difference was observed between co-mutant and WT patients in the TMB-low subgroup (P=0.22). Among co-mutant patients, TMB-high and TMB-low groups showed comparable survival (P=0.70), while in the WT subgroup, high TMB was associated with a trend toward better survival (P=0.08). Together, these results demonstrated that BRAF/PTPRT co-mutation was strongly associated with high TMB, and their combination identifies a subgroup with particularly favorable immunotherapy outcomes.

Figure 3.

Association of BRAF/PTPRT co-mutation with TMB and synergistic prognostic effect. (A) Violin plots comparing TMB levels between co-mutant and wild-type patients in the three cohorts. (B) Kaplan-Meier curves comparing overall survival between TMB-high and TMB-low patients in the MSK and CS cohorts. (C) Survival analysis stratified by TMB or co-mutation status in the TCGA and TCGA-RT cohorts. (D) Kaplan-Meier curves of four subgroups defined by co-mutation and TMB status in the combined immunotherapy cohort (MSK + CS). (E-H) Subgroup survival analyses: (E) co-mutant vs. WT in TMB-high; (F) co-mutant vs. WT in TMB-low; (G) TMB-high vs. TMB-low in co-mutant; (H) TMB-high vs. TMB-low in WT. ***, P<0.001. CS, combined study; MSK, Memorial Sloan Kettering Cancer Center; RT, radiotherapy; TCGA, The Cancer Genome Atlas; TMB, tumor mutational burden; WT, wild type.

Molecular, clinical, and immunological characteristics of BRAF/PTPRT co-mutant patients in the CS immunotherapy cohort

To dissect the mechanisms underlying superior immunotherapy outcomes in co-mutant patients, we performed multi-dimensional analyses in the CS cohort (Figure 4). TMB and neo-antigen load exhibited a clear stepwise gradient across the four mutation subtypes, with co-mutant tumors showing the highest levels, followed by the single-mutation groups (BRAF+/PTPRT− and BRAF−/PTPRT+), and WT tumors displaying the lowest values (Figure 4A,4B), with TMB strongly correlated with neo-antigen load (R=0.953, P<0.001; Figure 4C). In terms of clinical outcomes, the co-mutation subtype accounted for a higher proportion in the clinical benefit group compared with the non-benefit group (Figure 4D). Co-mutant patients achieved significantly better durable clinical benefit (DCB) rates (P<0.001; Figure 4E) and higher rates of complete/partial response and long-term benefit, with fewer progressive diseases and no benefit. ROC analysis showed that combining co-mutation and TMB improved predictive accuracy for immunotherapy response (AUC =0.659) over either alone (co-mutation: 0.594; TMB: 0.628; Figure 4F). In the multivariate logistic regression analysis (Figure S1), high TMB was identified as an independent predictor for clinical benefit (OR =2.33, 95% CI: 1.21–4.56, P=0.01). The BRAF-PTPRT co-mutation also demonstrated a strong clinical trend toward improved benefit, independent of TMB (OR =2.04, 95% CI: 0.90–4.77, P =0.09), likely due to the limited sample size of co-mutant cases. These findings supported the complementary value of the co-mutation in patient stratification. Immune cell deconvolution revealed significantly lower infiltration of immunosuppressive regulatory T cells (Tregs, P<0.05), eosinophils (P<0.01), and resting NK cells (P<0.01) in co-mutant tumors (Figure 4G-4I), indicating a less immunosuppressive, more “inflamed” microenvironment. Collectively, co-mutant tumors are characterized by high immunogenicity, favorable clinical response, and an immune-active microenvironment, providing a mechanistic basis for their improved outcomes with ICB.

Figure 4.

Molecular, clinical, and immunological characteristics of co-mutant patients in the CS cohort. (A,B) Violin plots of TMB and neo-antigen load in the four groups. (C) Scatter plot showing correlation between TMB and neo-antigen load. (D) Comparison of the distribution of the four mutation subtypes between the clinical benefit group and the non-benefit group. (E) Stacked bar plot of durable clinical benefit categories. (F) ROC curves for predicting immunotherapy response using co-mutation, TMB, or their combination. (G-I) Violin plots of immune cell infiltration for Tregs (G), eosinophils (H), and resting NK cells (I). *, P<0.05; **, P<0.01; ***, P<0.001. CR, complete response; CS, combined study; FPR, false positive rate; LB, long-term benefit; NB, no benefit; NK, natural killer; PD, progressive disease; PR, partial response; ROC, receiver operating characteristic; SD, stable disease; TMB, tumor mutational burden; TPR, true positive rate.

Transcriptomic and pathway analyses of BRAF/PTPRT co-mutant tumors in the TCGA cohort

To explore the molecular basis of the immunogenic phenotype, we performed transcriptomic analyses in the TCGA cohort (Figure 5). Differential expression analysis identified 481 upregulated and 68 downregulated genes in co-mutant tumors compared to WT (Figure 5A). Upregulated genes included cytotoxic effectors (GZMA, GZMB), antigen presentation components (B2M), and T-cell-recruiting chemokines (CXCL9, CXCL10, CXCL11). Unsupervised clustering of these DEGs clearly separated co-mutant from WT tumors (Figure 5B). KEGG pathway enrichment analysis showed that co-mutant tumors were enriched in immune- and inflammation-related pathways, including PI3K-Akt signaling, NOD-like receptor signaling, and phagosome (Figure 5C). GSEA further revealed significant enrichment of hallmark immune gene sets: Interferon Gamma Response (NES =1.700), Interferon Alpha Response (NES =1.782), Inflammatory Response (NES =1.555), and IL6-JAK-STAT3 Signaling (NES =1.456) (all FDR <0.001; Figure 5D). Conversely, co-mutant tumors also enriched metabolic and proliferative pathways, such as MYC Targets V1, fatty acid metabolism, and glycolysis (Figure 5E). Collectively, these transcriptomic data revealed that BRAF/PTPRT co-mutant tumors exhibited a distinct expression profile with upregulation of immune activation and inflammatory pathways, further supporting their immunogenic, immune-inflamed phenotype.

Figure 5.

Transcriptomic and pathway analyses of co-mutant tumors in the TCGA cohort. (A) Volcano plot of differentially expressed genes between co-mutant and wild-type tumors. (B) Heatmap of differentially expressed genes with unsupervised clustering. (C) KEGG pathway enrichment analysis of differentially expressed genes. (D,E) GSEA hallmark enrichment plots for immune-related (D) and metabolic/proliferative (E) pathways. FDR, false discovery rate; GSEA, Gene Set Enrichment Analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; NES, normalized enrichment score; TCGA, The Cancer Genome Atlas.

Validation of the immune microenvironment and immune checkpoint expression in the TCGA cohort

To validate the immune microenvironment features observed in the CS cohort, we analyzed immune cell infiltration and checkpoint gene expression in TCGA using CIBERSORT (Figure 6A). To ensure the robustness of our immune infiltration analysis, we cross-validated the CIBERSORT results using the ssGSEA algorithm. As shown in Figure 6B, both methods yielded consistent results, confirming that co-mutant tumors were characterized by significantly higher M1 macrophages and lower regulatory T cells (Tregs). Correlation analysis among immune subsets (Figure 6C) showed that activated effector cells (CD8+ T cells, activated CD4+ memory T cells, M1 macrophages, activated dendritic cells) formed a positively correlated cluster, which was negatively associated with immunosuppressive cell types, including Tregs and M2 macrophages. This pattern suggested a coordinated anti-tumor immune response in co-mutant tumors. Finally, co-mutant tumors expressed significantly higher levels of immune checkpoint molecules (Figure 6D), including programmed death-ligand 1(PD-L1/CD274) (P<0.01), CTLA4 (P<0.05), and TIM-3 (HAVCR2) (P<0.05), with a trend toward higher PD-1 (PDCD1). This elevated checkpoint expression provided a mechanistic basis for the enhanced sensitivity of co-mutant tumors to ICB. Together, these results validated the immune-inflamed, checkpoint-high phenotype of BRAF/PTPRT co-mutant tumors in an independent TCGA cohort and reinforced the robustness of our findings.

Figure 6.

Immune microenvironment and immune checkpoint expression in the TCGA cohort. (A) Stacked bar plot of relative proportions of immune cell subsets estimated by CIBERSORT and ssGSEA. (B) Violin plots comparing key immune cell subsets between co-mutant and wild-type tumors. (C) Correlation heatmap of pairwise correlations among immune cell subsets. (D) Violin plots comparing expression levels of immune checkpoint molecules (PDCD1, CD274, CTLA4, HAVCR2, LAG3). *, P<0.05; **, P<0.01; ***, P<0.001. aDC, activated dendritic cells; DC, dendritic cells; iDC, immature dendritic cells; pDC, plasmacytoid dendritic cells; ssGSEA, single-sample gene set enrichment analysis; TCGA, The Cancer Genome Atlas; Tcm, T central memory; Tem, T effector memory; TFH, T follicular helper; Tgd, T gamma delta; Treg, regulatory T cells.

Discussion

In this multi-cohort study, we identified BRAF-PTPRT co-mutation as a robust, treatment-specific predictive biomarker for favorable response to ICB in melanoma. Through integrative analysis of three independent cohorts (MSK, TCGA, and CS), we demonstrated that patients with concurrent mutations in BRAF and PTPRT consistently exhibited significantly longer OS and higher rates of clinical benefit (CR/PR/SD/LB) compared to those with single mutations or double WT. Notably, this survival benefit was observed exclusively in ICB-treated cohorts (MSK and CS) and was absent in radiotherapy-treated patients from TCGA, confirming that the prognostic value of co-mutation is context-dependent and reflects true predictive power for ICB rather than a general favorable outcome. A simplified risk score integrating co-mutation and TMB predicted clinical benefit with modest but significant discriminative ability (AUC =0.659), offering a clinically actionable tool for patient stratification.

The role of BRAF mutation alone in melanoma immunotherapy has been controversial. Some studies reported that BRAF-mutant melanomas harbor higher TMB and may benefit from ICB, while others found no significant association or even worse outcomes (18,19). Our single-gene analyses recapitulated this inconsistency: PTPRT single mutation was associated with better OS only in ICB-treated cohorts (MSK and CS), whereas BRAF single mutation showed benefit only in the non-immunotherapy TCGA cohort (13). This context-dependent heterogeneity underscores the pitfall of using single-gene alterations as predictive markers. PTPRT loss has been linked to STAT3 activation and enhanced chemokine expression, yet its clinical relevance alone appears insufficient (20). Here we showed that co-mutation of BRAF and PTPRT synergistically combined high TMB (from BRAF) with STAT3-driven immune activation (from PTPRT loss), thereby producing a stable, cross-cohort favorable signature that overcame the limitations of either single mutation. Our findings aligned with recent reports that co-occurring mutations in driver genes and immune-modulatory genes can define distinct melanoma subtypes with unique immunotherapy responses. To our knowledge, this is the first study to systematically validate BRAF-PTPRT co-mutation across multiple independent cohorts and to provide multi-omic mechanistic evidence linking this co-mutation to an immune-inflamed, checkpoint-high tumor microenvironment.

The TCGA transcriptomic analysis revealed that co-mutated tumors exhibit an “immune-inflamed” phenotype with upregulation of T-cell-recruiting chemokines (CXCL9/10/11), antigen-presentation machinery (B2M), and cytotoxic effectors (GZMA, GZMB) (21). GSEA further identified enrichment of IFN-γ response, inflammatory response, and IL6-JAK-STAT3 signaling—the latter being a direct consequence of PTPRT inactivation, as PTPRT normally dephosphorylates STAT3 (20,22). Conversely, immune infiltration analysis (CIBERSORT) showed that co-mutated tumors have higher proportions of M1-like (pro-inflammatory) macrophages and lower levels of immunosuppressive regulatory T cells (Tregs) and M2 macrophages (23). This M1-skewed, Treg-low microenvironment is known to be permissive for ICB efficacy. Mechanistically, PTPRT loss activates the IL6-JAK-STAT3 pathway, leading to the upregulation of T-cell recruiting chemokines CXCL9, CXCL10, and CXCL11. This STAT3-driven inflammatory signaling cascade appears sufficiently robust to override the classical immunosuppressive effects of BRAF activation, thereby promoting the observed M1-skewed, Treg-low inflamed microenvironment. Moreover, co-mutated tumors expressed significantly higher levels of immune checkpoint molecules including PD-L1, CTLA-4, and TIM-3, suggesting a pre-existing checkpoint-engaged state that renders these tumors particularly vulnerable to PD-1/CTLA-4 blockade (24). The strong correlation between TMB and neo-antigen load further supports that the genomic instability driven by co-mutation translates into higher immunogenicity (25). In the CS cohort, despite having higher serum LDH-a traditional marker of poor prognosis (26), co-mutated patients achieved significantly better clinical benefit, indicating that the co-mutation effect overrides the negative impact of high LDH.

An important observation is that neither co-mutation nor high TMB was associated with survival in TCGA patients treated with radiotherapy alone. This negative finding, together with the positive associations exclusively in ICB-treated cohorts, confirms that co-mutation and TMB are predictive biomarkers for ICB response rather than general prognostic factors. A growing body of evidence supports this context-dependent role. For instance, in cervical cancer treated with definitive radiotherapy, high TMB has been identified as an independent risk factor for worse OS (27). Similarly, a high TMB combined with specific genomic mutation signatures predicts a higher risk of relapse in patients with head and neck cancer undergoing chemoradiotherapy (28). These observations are in stark contrast to the well-established benefit of high TMB in immunotherapy. This treatment-context dependency reinforces the clinical utility of the co-mutation signature for guiding immunotherapy decisions and avoids the common pitfall of claiming a biomarker that is merely prognostic regardless of therapy.

In current clinical practice, BRAF-mutant melanoma patients are often prioritized for targeted therapy (BRAFi + MEKi) due to high response rates, albeit with inevitable acquired resistance (29). More than half of advanced melanoma patients will still die due to melanoma despite current treatment advances (30). The pathways that lead to acquired resistance to targeted therapy, which may confer resistance to immunotherapy, remain incompletely understood. Resistance is driven by multiple mechanisms, including secondary mutations, pathway reactivation, and microenvironmental interactions (29). These challenges have motivated ongoing research into combining BRAF inhibitors with immunotherapies to enhance and prolong treatment effectiveness (31). Our results suggested that among BRAF-mutant patients, those with concurrent PTPRT mutations may derive exceptional benefit from ICB alone or in combination, potentially achieving durable responses without the rapid resistance seen with targeted agents. Because BRAF and PTPRT status can be obtained from routine next-generation sequencing panels (e.g., MSK-IMPACT, FoundationOne), this co-mutation signature is readily implementable in clinical laboratories. The simplified risk score combining co-mutation (1 point) and high TMB (1 point) provided an AUC of 0.659 for predicting clinical benefit. While this discriminative ability is modest, it is comparable to or better than currently used single biomarkers (e.g., TMB alone) in many solid tumors. More importantly, the score can be easily calculated from standard sequencing reports and may be combined with other clinical variables to further improve accuracy. In our analysis, patients with a score of 2 (co-mutation and high TMB) had significantly higher clinical benefit rates and longer OS compared to those with score 0, suggesting that this high-risk group could be prioritized for ICB-based regimens.

Several limitations should be acknowledged. First, the sample size of co-mutated patients remained modest (25–36 per cohort), and the multivariable significance in the MSK cohort was borderline after adjusting for drug type. Small sample sizes in rare biomarker studies can increase the risk of type II errors and limit statistical power. However, the consistent direction across cohorts and the significant results in the larger CS cohort support robustness. Second, the absence of transcriptomic data in the MSK and CS cohorts precluded direct validation of the immune pathways identified in TCGA; we relied on TCGA for mechanistic exploration. Third, the retrospective design may introduce selection bias, and prospective validation is warranted (32). Fourth, the exact molecular crosstalk between BRAF and PTPRT at the protein level requires functional studies, including in vitro and in vivo models. Fifth, the definition of “long-term benefit (LB)” in the CS cohort was not accompanied by a prespecified minimum duration; we used the original database annotations to avoid introducing arbitrary thresholds (33). We acknowledge that the lack of functional experimental validation is a major limitation of the present study. While our transcriptomic and bioinformatics analyses provide strong evidence for the STAT3-driven inflammatory phenotype, these findings are based on retrospective cohort data and indirect inferences. Future mechanistic studies utilizing CRISPR-engineered melanoma cell lines and syngeneic mouse models are urgently required to functionally dissect the crosstalk between BRAF activation and PTPRT loss, and to verify the STAT3/IFN-γ regulatory axis. Our ongoing work is dedicated to addressing these functional gaps. Moreover, from a clinical translation perspective, the optimal cut-off for TMB to combine with the co-mutation remains to be standardized across different sequencing platforms, as TMB calculation is known to be influenced by panel design, sequencing depth, and algorithm selection (34,35).

Prospective clinical trials stratifying melanoma patients by BRAF/PTPRT co-mutation status are needed to validate its predictive value for ICB versus targeted therapy (36). Functional investigations using CRISPR-engineered cell lines and syngeneic mouse models will elucidate the cooperative mechanism between MAPK activation (BRAF) and STAT3-dependent immune modulation (PTPRT loss) (37). Single-cell RNA sequencing and spatial transcriptomics could further dissect the cellular composition and ligand-receptor interactions in co-mutated tumors. Additionally, the risk score could be refined by incorporating other genomic or clinical variables (e.g., LDH, tumor stage) and validated in larger, multi-institutional cohorts.

Conclusions

BRAF-PTPRT co-mutation defines a distinct melanoma subtype characterized by high TMB, high neo-antigen load, an Treg-low immune microenvironment, and enhanced sensitivity to ICB. This co-mutation signature is a robust, cross-cohort, and treatment-specific predictive biomarker that can be easily integrated into clinical sequencing workflows. A simplified risk score combining co-mutation and TMB offers additional predictive value. Our findings supported the inclusion of PTPRT testing in routine panels for BRAF-mutant melanoma to guide immunotherapy decisions.

Supplementary

The article’s supplementary files as

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.