Spatial transcriptomics reveals molecular determinants of mortality in thin melanoma

Summary: Abstract Body: Despite their excellent prognosis, thin cutaneous melanomas (<1 mm) account for one-quarter of total melanoma deaths. Current prognostic factors demonstrate limited accuracy in thin melanoma, precluding identification of high-risk individuals who could benefit from adjuvant therapeutic strategies. Here, we apply spatial transcriptomics to profile a nested case-case series comprising 15 fatal and 15 non-fatal thin melanomas, individually matched by age, sex, tumour thickness, and follow-up duration. Relative to non-fatal cases, fatal cases demonstrated differential upregulation of multi-gene signatures associated with two independent tumour phenotypes: an invasive-state cluster driven by epithelial-mesenchymal transition, and an immune-cold cluster characterized by immunosuppressive ligand-receptor interaction and reduced lymphocyte proportions. On prognostic analysis, we identified a five-feature combination that demonstrated a sensitivity of 75% and specificity of 91% for predicting mortality in a validation cohort of 23 additional patients. Broadly, we present a spatial transcriptomics framework for the discovery of prognostic features in thin melanoma, which could enhance stratification and inform clinical management of early-stage disease. Chenhao Zhou¹, Samuel X. Tan¹, Yung-Ching Kao¹, Magdalena Claeson^{2, 1}, Susan Brown¹, Duncan Lambie³, David C. Whiteman⁴, Nirmala Pandeya⁴, Catherine M. Olsen⁴, Andrew Barbour^{1, 5}, Stuart MacGregor⁶, H P. Soyer^{1, 7}, Mitchell S. Stark¹, Nicholas Hayward⁴, B M. Smithers⁸, Quan Nguyen⁹, Kiarash Khosrotehrani^{1, 7} 1. Frazer Institute, The University of Queensland, Brisbane, QLD, Australia. 2. Department of Dermatology and Venereology, Goteborgs universitet Institutionen for kliniska vetenskaper, Gothenburg, Sweden. 3. Pathology Queensland, Princess Alexandra Hospital, Woolloongabba, QLD, Australia. 4. Department of Population Health, QIMR Berghofer Medical Research Institute, Herston, QLD, Australia. 5. Department of Surgery, Princess Alexandra Hospital, Woolloongabba, QLD, Australia. 6. Statistical Genetics, QIMR Berghofer Medical Research Institute, Herston, QLD, Australia. 7. Department of Dermatology, Princess Alexandra Hospital, Woolloongabba, QLD, Australia. 8. Queensland Melanoma Project, The University of Queensland, Brisbane, QLD, Australia. 9. Institute for Molecular Bioscience, The University of Queensland, Brisbane, QLD, Australia. Bioinformatics, Computational Biology, and Imaging