Primary cutaneous melanoma is a highly heterogeneous tumour. My hypothesis is that the histopathological heterogeneity reflects biological variation, which is likely to have prognostic and predictive significance.
A histopathological review of 798 primary melanomas from the Leeds Melanoma Cohort Study was recorded using virtual pathology. The tumour blocks had previously been sampled using a tissue microarray needle, yielding a core from which RNA was extracted and assayed using Illumina® WG-DASL. This provided the opportunity to compare histopathological characteristics with gene expression data derived from consistently sampled regions. RandomSpot© was used to estimate the percentage of stroma (POS) within cored regions. Statistical analyses were performed using STATA v14.2. Inter- and intraobserver agreement were analysed and robust measures were retained.
Histopathological characteristics were analysed with respect to germline BAP1 mutation status to assess whether they could predict germline BAP1 mutation status. BAP-like histopathology was not significantly associated with germline BAP1 mutation status (deleterious versus none, Fisher’s exact test, p=0.1). A personal history of mesothelioma (Fisher’s exact test, p=0.005), or a family history of meningioma (Fisher’s exact test, p=0.005) or BCC (Fisher’s exact test, p=0.02) was associated with deleterious, germline BAP1 mutations. Cancer history appeared to be a better indicator of germline BAP1 mutation status than BAP-like histopathology.
Several histopathological factors were predictive of melanoma-specific survival, including the POS. The area under the curve increased by 3% when the POS and AJCC stage were combined in ROC curve analysis. The POS was an independent predictor of melanoma-specific survival (HR 0.99, 95% CI 0.98-0.99, Cox proportional hazards model, p=0.005), even adjusting for known prognostic factors. SDF1 gene expression was significantly associated with the POS and was independently protective for melanoma-specific death (HR 0.8, 95% CI 0.68-0.94, Cox proportional hazards model, p=0.005) in adjusted analyses. The POS and SDF1 could represent novel predictive and prognostic biomarkers.
