Advanced melanoma is an aggressive cancer with a high metastatic potential often resulting in melanoma brain metastases (MBM). The complexity of the MBM tumor microenvironment (TME) impacts tumor progression and therapy response, leading to poor outcomes. Our study integrates spatial transcriptomics profiles with tissue morphology annotations of MBM and clinical data to investigate the TME landscape in relation to treatment and patient outcomes. We performed spatial transcriptomics on 21 MBM samples collected after various treatment strategies. All samples were preserved in formalin-fixed paraffin-embedded (FFPE) blocks and profiled using the Visium Spatial Gene Expression kit (10X Genomics) with a 6.5x6.5mm Visium Capture slide. All sequencing was performed on a NovaSeq6000 platform (Illumina) with an average of 71,391 reads per spots. H&E slides were annotated by our study pathologist (DL) to enable pathology-guided cluster analysis. Spatial transcriptomics data analysis was conducted using Seurat R pipeline which involved data quality check, normalization, integration, cell type annotation and deconvolution, and clustering. Further differential gene expression analyses were performed between different identified tumor clusters or between clinical groups, leading to the identification of enriched pathways and cell-cell interaction patterns. Our phenotyping of 25,208 spots revealed the presence of diverse cell populations, including neurons, melanoma, plasma, endothelial, stromal, and myeloid cells. We identified 8 distinct functional clusters characterized by specific gene expression and activated pathways. These include 5 tumor clusters with different immune infiltration or expression profiles, identifying differences in cell metabolism and heterogeneity in the TME. Importantly, we found MBM from patients exposed to radiation were significantly enriched for epithelial-to-mesenchymal transition pathway and genes related to cell adhesion. We further showed a reduced immune response in patients received radiation compared those who did not, and we noted that an active immune response was generally positively correlated with survival. Our study is the largest spatial analysis of MBM reported to date. Advances in spatial technologies provide novel insights into the structure and composition of MBM TME. Our findings highlight the importance of spatial context in understanding MBM biology and the potential of spatial transcriptomics in advancing precision oncology. Clemence J. Belle, Sandra Brosda, Vanessa F. Bonazzi, Victor Bulteau, Thomas Stuart, Zherui Xiong, Duncan Lambie, Peter A. Johansson, Lauren G. Aoude, Kalpana Patel, Samantha J. Stehbens, Mitchell S. Stark, Nikolas K. Haass, Wen Xu, Mark B. Pinkham, Matthew C. Foote, Sarah Olson, Victoria Atkinson, Arutha Kulasinghe, Quan H. Nguyen, Andrew Barbour. Unravelling melanoma brain metastasis tumor microenvironment characteristics using spatial transcriptomics [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 5076.

Thin melanomas (<1mm Breslow), despite their excellent prognosis, account for the majority of melanoma deaths. There is a lack of predictors to identify thin melanomas at the highest risk of progression. We conducted a nested case-case analysis of 16 fatal thin melanomas versus 16 non-fatal thin melanoma controls matched for age, sex, year of diagnosis, thickness and length of follow-up, randomly selected from a population cohort in Queensland, Australia. We examined spatial transcriptome profiles of archived formalin-fixed paraffin-embedded tissues to identify features associated with fatal cases. Within tumors and their adjacent regions, we observed decreased proportions of activated immune cells (Treg, NKT, dendritic cells and recruited monocytes) and increased proportions of naive CD4+T cells and M2 macrophages in cases when compared to controls. Furthermore, we identified a list of differentially expressed genes overrepresented in Epithelial to Mesenchymal Transition (EMT) and E2F transcription factor targets pathways. Upregulation of a set of 28 EMT- and E2F target-associated genes allowed to distinguish case from control tumors (AUC = 0.95). These differences were confirmed in a separate set of 12 fatal and 12 control thin melanomas. Moreover, T cell activation-related genes were increasingly expressed in control tumors, mainly in keratinocyte- and fibroblast-enriched regions, implying stroma-immune cross-talks. Our study identifies key cellular and molecular characteristics associated with patient survival. The findings provide new strategies in predicting prognosis and developing adjuvant therapies for patients at high-risk of disease progression.

While population screening for melanoma is not recommended, screening those at high risk could lead to earlier detection and improved patient outcomes. Current melanoma risk stratification is largely based on subjective clinical assessment and self-report. This is time-consuming and dependent on the experience and understanding of the person reporting. Therefore, improved, standardized methods for melanoma risk assessment are required. We developed an automated skin surface phenotype which assesses skin phenotypes associated with an increased risk of melanoma from three-dimensional (3D) total body imaging. We focused on four risk phenotypes: i) skin color, ii) naevus count and their distribution, iii) photodamage, and iv) freckling density. 3D total body imaging provides a high-resolution capture of almost the entire skin surface within minutes. Images of Queensland adults from both the general (n=156) and high-risk populations (n=300) were annotated for the four risk skin phenotypes. Skin color was assessed via image processing techniques, while convolution neural networks (CNNs) were used to develop classification algorithms for naevus counts, photodamage and freckling density. Results are promising, with overall accuracies > 85% for naevus count (k=0.56) and photodamage CNNs, suggesting melanoma risk phenotypes can be accurately and objectively extracted from 3D total body images. Validated against traditional risk calculators, the combined phenotypes can stratify people into appropriate prevention and early detection protocols, as well as to monitor changes over time, providing earlier indications of progression along the entire naevus- and melanoma-genesis pathway. This novel approach combining CNNs and total body imaging overcomes the current limitations of melanoma risk assessment and could improve clinical workflow prioritizing those at highest risk of developing melanoma.

radiation oncology collaboration G. Fogarty, M. Jobbins,M. Fay, A. Kaminski, D. Schlect, D. Christie, L. Spelman, R. Sinclair, S. Shumack Genesis Cancer Care, St Vincent’s Hospital, Sydney, New South Wales, Australia Genesis Cancer Care, Lake Macquarie Private Hospital, Newcastle, New South Wales, Australia Genesis Cancer Care, Wesley Hospital, Brisbane, New South Wales, Australia Genesis Cancer Care, Nambour Hospital, Gold Coast, Queensland, Australia Dermatology, Brisbane, Queensland, Australia Northern Clinical School, University of Sydney, Sydney, New South Wales, Australia