A new metabolic signature to predict melanoma recurrence

Melanoma is one of the most prevalent cancers in the United States, accounting for 5% of all new cancer cases, and has been increasing worldwide.¹ Melanoma arises from the malignant proliferation of melanocytes, cells that produce melanin that provides skin pigmentation. The high metastatic potential of melanoma drastically affects the prognosis of patients diagnosed at later stages. In particular, 50%–75% of melanoma cases metastasize to the brain, accounting for 54% of all melanoma-related deaths.^{2, 3} It is thus critical to diagnose melanoma at earlier stages to limit metastasis and more widespread invasion, allowing for complete surgical removal of the tumour.

While excision of early-stage melanoma results in a favourable prognosis, recurrence can occur in many of these patients, being responsible for a large proportion of melanoma-related deaths. The mechanisms underlying melanoma recurrence remain incompletely understood. Recent evidence points toward the critical influence of the tumour microenvironment (TME) on melanoma recurrence. The TME is composed of the extracellular matrix, immune cells, mesenchymal cells, and blood vessels that surround the tumour cells, communicating their behaviour and response to treatment. Dissecting the melanoma TME therefore may give rise to diagnostic markers and mechanistic insights into its recurrence patterns, opening avenues for the development of targeted therapies that can lower the risk of melanoma recurrence.

Recent development in artificial intelligence (AI)-driven histopathology introduces important tools that can be leveraged to investigate the spatial organization and molecular markers in the TME. The application of AI technology has been proven to objectively and consistently produce valuable clinical information that provides molecular markers and predicts cancer outcomes. The utilization of AI in cancer histopathology is not just limited to its predictive and prognostic usage. This technology can be used to profile and map tumour immune cell networks and dissect molecular pathways within the TME, revealing critical mechanistic insights into response to treatments, recurrence patterns, and metastatic potential.

In a recent study, Szadai et al.⁴ employed AI-powered histopathology and spatial proteomics to explore and compare the tumour cells and TME interplay between recurrent and non-recurrent primary melanoma samples. This AI-driven methodology proved to be effective in distinguishing normal tissue from stromal and tumour regions across both groups. Notably, these authors integrated laser-microdissection technology and quantitative proteomic analysis with digital pathology readouts. This unique approach produced critical spatial proteomics data that sheds light on the molecular interplay between mitochondrial functions and immune response in the tumour and stromal components of recurrent melanoma.

In particular, Szadai et al. found that tumour cells in recurrent melanoma showed higher expression of components of the DNA synthesis pathway and mitochondrial translation as compared to stromal cells. Vice versa, recurrent melanoma stromal cells displayed higher expression of mediators of the epithelial-mesenchymal transition and programmed death 1 signalling pathways. These findings contrast with the enriched pathways between the tumour and stromal components from non-recurrent melanoma, with higher expression of keratinization and mitophagy pathways in tumour cells and interleukin signalling and collagen degradation in stromal cells. The dichotomous findings between the tumour and stromal compartments suggest that elevated mitochondrial functions in tumour cells may alter the TME toward a pro-recurrent and immune-evasive phenotype. To bolster the evidence of the interplay between mitochondrial activity and immune dysregulation in melanoma recurrence, Szadai et al. compared the proteome differences between recurrent and non-recurrent melanoma in both tumour and stromal compartments. In both cell groups, the recurrent melanoma specimens exhibited an upregulation in mitochondrial pathways and cellular proliferation, while also displaying a downregulation in immune response pathways.

The findings of Szadai et al. pose yet another paradox to the constitutional Warburg effect theory, positing that glycolytic metabolism is the main driver for cancer cell proliferation and metastasis.⁵ Accumulating evidence suggests that mitochondrial function is essential in all aspects of tumour progression, from cellular growth to immune evasion and invasion.^{6, 7} In the context of melanoma, numerous studies have identified the critical role of oxidative phosphorylation (OxPHOS) in promoting tumour cell proliferation, survival, chemo- and immunotherapy resistance, and metastasis.⁸ These findings allude to the metabolic plasticity of melanoma cells, readily shifting between glycolysis and OxPHOS in response to changes in the TME that include biofuel availability, immune response, and the presence of anti-tumour molecules. Furthermore, mitochondria can be transferred from the stromal cells of the TME to the tumour cells to meet the demand for mitochondrial OxPHOS,⁹ depicting a dynamic landscape of the metabolic interactions between stromal and tumour cells in melanoma. The findings from Szadai et al. thus provide a spatial visualization of this mitochondrial network and open an avenue for the development of therapy pertinent to the molecular pathways involved in melanoma recurrence. Indeed, recent studies have begun to explore the therapeutic potential of mitochondrial targeting agents and uncouplers in treating treatment-naive and resistant melanomas.¹⁰

The integration of quantitative proteomics with AI-powered histopathology by Szadai et al. is a unique and powerful tool for spatial analyses of molecular networks in recurrent melanoma and other types of cancer. There are still, however, limitations regarding the incorporation of this methodology with other omics approaches such as genomic and transcriptomic analyses. Indeed, Szadai et al. attempted to assess the protein-level upregulation of the two mitochondrial pathways that are highly associated with lower survival rates at the transcriptomic level. Interestingly, their proteome analysis did not reveal upregulation in these two pathways. This finding suggests a dissociation between transcriptomic and proteomic analysis, thus emphasizing the convention that RNA expression is not equivalent to protein expression and activity. Indeed, multiple post-transcriptional and post-translational modifications can precipitate the disconnect between the two types of omics analyses, adding complexity to deciphering the mechanisms of melanoma recurrence. This underscores the need for a multi-omics approach that integrates spatial proteomics data with other forms of spatial and/or single-cell level omics such as spatial transcriptomics and single-cell RNA sequencing. This multi-level approach can help unravel the convoluted molecular networks pertinent to the recurrence of melanoma across different levels of molecular products and modifications.

In conclusion, Szadai et al. demonstrated a proof of concept for the combination of AI-driven histopathology and quantitative proteomics to produce a spatial proteomics approach that depicts the interplay between mitochondrial functions and immune evasion in melanoma recurrence. Future applications of this methodology can be integrated with other high-powered omics tools to dissect these molecular networks in melanoma and other cancer types, allowing for risk stratification of disease recurrence and metastasis and the development of more effective targeted therapies.

N.K.V. and R.J.P. drafted and edited the piece.

Not Applicable