(P 376) MULTI-OMIC PROFILING OF LEIOMYOSARCOMA IDENTIFIES MARKERS PREDICTIVE OF RESPONSE TO GEMCITABINE AND DOCETAXEL

Objective: Combination of gemcitabine and docetaxel (Gem/Tax) is one of the most frequently considered systemic first-line regimens for leiomyosarcoma (LMS) patients with advanced disease. However, there is wide heterogeneity in clinical responses with only a subset of patients responding to this treatment. There is currently a poor understanding of the biology underlying this differential response and there are no predictive biomarkers for the response to Gem/Tax in LMS. The identification of biomarkers involved in the response to Gem/Tax would allow clinicians to treat only those patients that are most likely to receive long-term durable benefit, while sparing the toxicity of drug exposure for those patients who are unlikely to benefit this therapy.

Methods: We profiled archival tumor specimens from the Leiomyosarcoma Direct Research Foundation (LMSdr) Tumor Bank hosted at Stanford University. The LMSdr Tumor Bank contains fully annotated 288 primary, recurrent and metastatic tumor specimens that were used to construct a tissue microarray. Among these cases, we identified a cohort of primary specimens from 54 patients who have been treated with Gem/Tax (26 responders and 28 non-responders). To identify candidate mechanisms of Gem/Tax resistance and response in LMS, we performed multi-omic profiling of these specimens by whole transcriptome RNA-seq (Illumina), DNA methylation profiling (EPIC microarrays, Illumina). High-throughput proteomic profiling, miRNA sequencing and validation experiments by immunohistochemistry are ongoing.

Results: Between responders and non-responders to Gem/Tax therapy, we identified 41 significantly differentially methylated CpGs, and 163 significantly differentially expressed genes. Differentially methylated CpGs were associated with 25 protein-coding genes. Among the differentially methylated CpGs, hypermethylation in the gene body of the RGL3 in two separate CpGs was positively correlated with mRNA expression of this gene (r = 0.53 and p=0.005 for cg17848054, and r = 0.46 and p=0.02 for cg02774542). Genes significantly upregulated in responders included the COL1A1 oncogene and several genes encoding protein kinases, transcription factors and growth factors. We validated strong expression of the COL1A1 protein in the primary tumors of responders to Gem/Tax by immunohistochemistry (strong expression in 20/25 (80%) of responders compared to 12/27 (44%) of non-responders; Fisher’s exact test p=0.01). In the non-responders, the significantly upregulated genes included PAK3 protein kinase and THRB transcription factor. Gene Set Enrichment Analysis indicated higher expression of the genes involved in the IGF1R, PDGF and thrombopoietin signaling pathways in responders to Gem/Tax. Analysis of the paired primary and metastatic tumors from 3 patients identified 66 differentially expressed genes. Metastatic tumors demonstrated significant overexpression of the gene encoding oncogenic ROS1 protein kinase. By immunohistochemistry, we validated expression of ROS1 protein in 134/261 (51.3%) of all specimens included in the tissue microarray. Metastatic tumors showed higher expression of ROS1 compared to primary tumors (expression of ROS1 in 68% (39/57) of metastatic tumors, compared to 47% (95/204) of primary tumors; Fisher’s exact test p=0.004). Among patients treated with Gem/Tax, 13/17 (76%) of metastatic tumors showed expression of ROS1 compared to 27/53 (51%) of primary tumors (Fisher’s exact test p=0.09).

Conclusion: This is the first study to identify candidate markers and molecular pathways associated with response and resistance to Gem/Tax therapy in LMS. This is also the first study that demonstrates the expression of ROS1 protein kinase in LMS, which can be potentially therapeutically targeted by crizotinib. Our findings warrant further studies (including proteomics, miRNA profiling and immunohistochemistry) on the current cohort and in an independent cohort to validate these findings.