Comparative transcriptomic analysis to identify similarities and therapeutic vulnerabilities in olfactory neuroblastoma (ONB), sinonasal neuroendocrine carcinoma (SNEC) and sinonasal undifferentiated carcinoma (SNUC).

Abstract

6095 Background: ONB, SNEC and SNUC are rare sinonasal epithelial/neuroepithelial tumors, underserved by clinical trials, with few treatments available despite novel therapeutic agents against surface targets approved or in clinical development. Transcriptomic similarities of ONB with small cell lung cancer (SCLC), pheochromocytoma (PH), paraganglioma (PG), glioblastoma (GB) and low-grade glioma (LGG) are reported, but not for SNUC or SNEC. We examined the transcriptome of ONB, SNUC, SNEC, neuroendocrine (NE) and central nervous system tumors from a real-world (RW) patient cohort to identify similarities and uncover therapeutic vulnerabilities. Methods: Tumor specimens (pathology per referring clinician) tested (Caris Life Sciences, Phoenix, AZ) included ONB (n = 26), SNUC (n = 9), SNEC (n = 6), SCLC (n=1751), pancreatic NE tumors (PNET, n=16), PH (n=23), PG (n=50), LGG (n=657), GB (n=4524) and neuroblastoma (NB, n=47). RNA sequencing data were processed to obtain transcripts per million (TPM) values. Clustering was performed with a random subset of 50 samples for tumor types with n>100. ONBs were subtyped to neural and basal (Classe et al . 2018). RW overall survival (rwOS) was calculated from insurance claims (tissue collection to last contact), compared with log-rank test; Cox proportional hazard model was used for hazard ratio (HR). Selected genes encoding surface targets included DLL3 , PMEL , PVRL4 , TACSTD2 , ERBB2 , F3 , CLDN18 , EGFR, ERBB3 , MET , GPC3 , CD276 , VTCN1 and FOSL1 . Median TPM values for genes of interest in ONB, SNUC and SNEC were examined. Results: There were 5 transcriptomic clusters (C1-5) (Table). Across clusters, median rwOS was worse for C1 (11.6 mo) and C3 (16.6 mo) vs. C2, C4, and C5 (all not reached), (p = 0.0) and was numerically better in neural (C4) vs. basal ONB (C0) (HR = 0.388, p=0.199). Highest expression for F3 (34.5), CD276 (15.6), GPC3 (7.2) and CLDN18 (1.1) was in ONB; ERBB2 (16.4), TACSTD2 (13.9), EGFR (9.2), PVRL4 (6.2), PMEL (2.0) and FOLR1 (1.5) in SNUC ; ERBB3 (83.3), MET (32.4), DLL3 (3.5) and VTCN1 (1.7) in SNEC. Conclusions: We show that neural ONBs cluster independently and basal ONBs co-cluster with SCLC and PNET, joined also by SNUC and SNEC. In our dataset, this cluster is associated with worse rwOS. We also show expression of surface target genes in ONB, SNUC and SNEC, indicating the presence of actionable subsets with existing drugs approved in other tumor types. Our findings provide targets for protein expression validation and expansion of therapeutic options for patients with these rare tumors. Cluster Tumor type, samples over N C1 ONB basal, 6/7 SNUC, 9/9 SNEC, 5/6 SCLC, 48/50PNET, 15/16NB, 2/47 C2 ONB neural, 2/18 SNEC, 1/6 PH, 22/23PG, 47/50NB, 4/47 C3 LGG, 50/50GBM, 50/50PG, 1/50 C4 NB, 41/47PG, 1/50PH, 1/23 C5 ONB neural, 16/18 ONB basal, 1/7 PG, 1/16