An immunogenomic analysis of melanoma brain metastases (MBM) compared to extracranial metastases (ECM).

Abstract

9521 Background: Previous work has shown that MBM have a unique molecular profile compared to ECM. Description of the biology of MBM will facilitate the design of rational therapies for patients (pts) with MBM. Methods: We analyzed a previously published dataset from MD Anderson Cancer Center, which includes RNA-seq on surgically resected FFPE MBM (88 tumors from 74 pts) and surgically resected ECM from the same pts (50 from 34 pts). WES on 18 matched pairs of MBM and ECM was available. The STAR pipeline was used to estimate mRNA abundance. The DESeq2 package was used to perform differential gene expression (DGE) analyses. Pathway analysis was performed using Gene Set Enrichment Analysis (GSEA). Paired DGE and GSEA analyses comparing MBM vs. lymph node metastases (LN mets, n = 16) and MBM vs. skin mets (n = 10) were performed. CIBERSORT estimated relative abundance of immune cell types in MBM and ECM. The GATK Mutect2 pipeline was used to call somatic mutations using paired normal tumor samples. Mutations were annotated using the Ensembl Variant Effect Predictor and visualized using the Maftools package in R. RNA-seq was available on 54 primary cutaneous melanoma (CM) pt samples, including 19 CM which did not recur, 19 CM which recurred as MBM, and 16 CM which recurred as ECM. Gene Ontology or KEGG Pathway analysis was performed using goana function of limma package in R. Results: Comparing MBM vs. LN and MBM vs. skin mets, paired DGE identified 136 and 89 up-regulated genes with a fold change > 2 and false-discovery rate (FDR) q-value < 0.05. Moreover, 308 and 659 down-regulated genes with a fold change < 0.5 were identified in MBM vs. LN and MBM vs. skin mets, respectively (q < 0.05). Paired GSEA found that autophagy signaling pathways may be up-regulated in MBM vs. LN and MBM vs. skin mets. On a single-gene level, comparing both MBM vs. LN and skin mets, the most strongly up-regulated genes in autophagy pathways were GFAP and HBB, whereas fold changes in the majority of other autophagy-related genes were low and did not reach significance. Comparison between CM which recurred in brain vs. CM which did not recur identified up-regulation of autophagy pathways. No difference in autophagy pathway expression was observed comparing between CM with any recurrence vs. without recurrence. CIBERSORT identified an increased proportion of immune suppressive M2 macrophages compared to tumor suppressive M1 macrophages in both MBMs and ECMs. Conclusions: Up-regulation of autophagy pathways was observed in pt-matched MBM vs. LN and skin mets. This finding seemed to be driven by up-regulation of GFAP and HBB, which could reflect changes in the tumor microenvironment (TME). Future studies using single-cell RNA-seq or spatial transcriptomic technology will dissect the TME. A higher M2:M1 ratio may contribute to an immune suppressive tumor microenvironment in MBM and ECM and is targetable. Validation of our findings in an independent Duke dataset is ongoing.