Spatial proteomics enables the concurrent assessment of over 50 proteins at single-cell resolution, facilitating unprecedented evaluation of the intra-tumoural interactions that drive cancer progression and therapy resistance. This review examines the translational potential of spatial proteomics to stratify metastasis, treatment response, and survival in cancer. We then explore key spatial findings with reference to the pathophysiology and molecular heterogeneity of melanoma.
Methods
This systematic review screened the PubMed, Embase, ACM Digital Library, and IEEE Xplore databases (inception to 1 Jan, 2022) for studies explicitly applying spatial proteomic parameters as predictive biomarkers in cancer. Data procedures were formalized in a prospectively registered protocol (PROSPERO; CRD42021283721), and cross-checking was performed against the REMARK and QUIPS criteria.
Results
From 11398 records, 64 studies were deemed eligible for review, comprising 17 cancers and 9502 patients. From these, we identified a comprehensive taxonomy of nine geometric features: co-expression, density, organization, subcellularity, distance, adjacency, area, communities, and heterogeneity. Each spatial feature was represented across multiple cell types, and all nine exhibited predictive value.
Ten studies examined melanoma across 731 patients, encompassing the spectrum of locally invasive, nodal, and distantly metastatic disease. Several studies found that closer median distances between CD8+ CTLs and SOX10+ tumour cells correlated with disease-specific survival beyond raw CD8 expression alone, highlighting the importance of weighting cell density by likelihood of activity. Another promising strategy was the use of spatial PD-1/PD-L1 autocorrelation as a surrogate for interaction frequency; indeed, high pre-treatment adjacency was consistently a favourable predictor of response to anti-PD-1 blockade
Conclusion
In this review, we demonstrate early evidence of spatial proteomic advantage over single-plex techniques for outcome prediction and treatment selection in cancer. We illustrate these topography-informed findings through melanoma, the archetypal model of a phenotypically heterogeneous cancer with prognostic molecular hallmarks that can be precisely elucidated through spatial proteomics