Scientific illustration showing blue immune T cells attacking a reddish melanoma tumor, with antibody molecules representing immune checkpoint inhibition and a pigmented skin lesion in the foreground.
Illustration depicting immune cells targeting a melanoma tumor during immunotherapy

Immune checkpoint inhibitors have transformed the treatment of advanced melanoma, but as many as 40% to 60% of patients either fail to respond or eventually develop resistance. Two new studies suggest that examining the proteins within melanoma tumors could help explain these differences and eventually guide more personalized treatment strategies.

Researchers at the Icahn School of Medicine at Mount Sinai and collaborators in the National Cancer Institute’s Clinical Proteomic Tumor Analysis Consortium identified three distinct molecular subtypes of melanoma associated with different responses to combination immunotherapy in a study published in Clinical Cancer Research. A companion study published in Cancer Discovery created a detailed molecular atlas of melanoma, uncovering potential mechanisms of immune resistance and more than 1,100 candidate tumor antigens that could inform future immunotherapies.

“Together, these two studies provide important new insights into melanoma heterogeneity, both in terms of tumor biology and response to immunotherapy, bringing us one step closer to precision medicine for melanoma patients,” Pei Wang, PhD, professor of genetics and genomic sciences at Mount Sinai and a corresponding author of both studies, told Inside Precision Medicine.

Three melanoma subtypes, different treatment outcomes

In the Clinical Cancer Research study, investigators retrospectively analyzed preserved tumor samples collected before immunotherapy from patients with metastatic cutaneous melanoma treated at 12 institutions. The analysis included 164 patients who received combination PD-1 and CTLA-4 checkpoint blockade and 44 who received PD-1 monotherapy.

By examining protein expression alongside genomic and RNA data, researchers identified three reproducible melanoma subtypes: ICI-sensitive, characterized by increased immune signaling and immune cell infiltration; Ribo-high, marked by increased protein synthesis and tumor growth-related activity; and EMT-like, characterized by increased stromal activity and extracellular matrix remodeling that may limit immune cell access to tumors.

Among patients receiving combination immunotherapy, approximately 73% to 74% of those with ICI-sensitive tumors responded to treatment, compared with 43% to 44% of those with EMT-like tumors. Response rates in the Ribo-high group ranged from 60% to 67%.

The subtypes also distinguished patients according to progression-free survival.

Wang explained that protein profiling captures biological activity that conventional genomic biomarkers may overlook. “In our study, tumor mutational burden was not significantly associated with response to combined immune checkpoint inhibitor therapy,” she said. “Although PD-L1 expression was elevated in the ICI-sensitive subtype, it did not distinguish between the Ribo-high and EMT-like subtypes, which exhibit distinct biological characteristics and potentially require different therapeutic strategies.”

The investigators also developed a machine-learning classifier using 127 proteins to assign tumors to the three subtypes. In an independent validation cohort, the classifier identified subtype membership with an area under the receiver operating characteristic curve (AUC) ranging from 0.89 to 0.97. It also predicted progression or death within three years with an AUC of 0.79. Adding proteomic subtype information to clinical variables improved prediction of three-year progression or death compared with clinical variables alone, achieving an AUC of 0.80 versus 0.64.

Importantly, the classifier was developed using routinely collected tumor biopsy formalin-fixed, paraffin-embedded specimens, supporting the possibility of developing a clinically practical test. “If validated in larger prospective studies, this approach could help identify patients more likely to benefit from combined immunotherapy and those who may need alternative or additional treatments,” Wang said.

A molecular atlas reveals potential therapeutic targets

The companion Cancer Discovery study examined 183 primary and metastatic melanomas using 13 layers of molecular analysis, including proteins, metabolites, protein modifications, and peptides displayed on tumor cells for immune recognition.

“One of the major contributions of the study is the generation of a large-scale immunopeptidomic dataset in melanoma, enabling us to identify more than 1,100 candidate tumor antigens, many of which were previously uncharacterized,” Wang explained. “Unlike approaches that predict antigens from DNA or RNA sequences, this technique directly measures peptides presented by human leukocyte antigen (HLA) molecules.”

The researchers found that greater tumor antigen presentation was associated with increased infiltration by CD8-positive T cells, particularly for antigens derived from MAGEA family proteins. “Antigens shared across patients could potentially serve as targets for broadly applicable cancer vaccines or T-cell therapies, while patient-specific antigens could inform personalized approaches,” Wang added.

The molecular atlas also revealed potential mechanisms underlying resistance. In EMT-like tumors, investigators identified a signaling network involving fibronectin (FN1), TGF-β, and cancer-associated fibroblasts that may create an environment that excludes immune cells.” These observations raise the possibility that targeting TGF-β signaling or fibroblast-associated pathways could help overcome resistance and improve responses to immune checkpoint inhibitors,” Wang explained, emphasizing that functional validation is still needed.

Other findings implicated altered ubiquitination of interferon signaling proteins and AKT1 activation in tumors associated with poor survival. They also identified elevated CDK12 activity in a subset of tumors associated with less favorable immunotherapy responses. “CDK12 is particularly interesting because inhibitors targeting CDK12/13 are already under clinical investigation in other cancers,” Wang said.

Another unexpected finding involved glycosylation of GPNMB, a protein already being investigated as a therapeutic target. The modification showed different associations with survival and immune-related features in male and female patients, suggesting that protein modifications, rather than protein abundance alone, may influence tumor behavior.

Moving toward clinical application

Although the studies identify promising biomarkers and potential therapeutic targets, the findings remain exploratory. Neither study establishes that the identified molecular pathways cause resistance or that targeting them will improve treatment outcomes.

The immediate priority is validating the 127-protein classifier in larger prospective cohorts and developing a reproducible assay suitable for routine clinical use.

“Ultimately, we hope to evaluate whether incorporating this classifier into clinical decision-making can improve treatment selection and patient outcomes,” Wang said. “However, additional validation and prospective clinical studies will be necessary before it can be used to guide melanoma treatment decisions.”

Also of Interest