(P 430) QUANTITATIVE EMT GENE SCORES AT DIAGNOSIS INDICATE METASTASIS RISK AND SURVIVAL IN SYNOVIAL SARCOMA

Objective: Synovial sarcoma (SS) is a soft tissue sarcoma with an undetermined cell of origin and a hallmark translocation of the SS18 gene on chromosome 18 with the SSX gene on chromosome X. SS is divided into two histologic variants with similar behavior and prognosis: monophasic (homogenous spindle morphology) and biphasic (spindle and epithelial morphology). Epithelial mesenchymal transition (EMT) and its reverse, mesenchymal epithelial transition (MET) occur naturally as a part of normal tissue development. This reversible transition can be exploited by cancer to increase invasiveness and resilience depending on tumor environment. This study sought to determine the impact of EMT/MET and phenotype heterogeneity on SS behavior.

Methods: Ribonucleic acid sequencing was performed on primary SS tumor specimens from patients treated at our institution. Two cohorts were established based on their ability to metastasize. The primary objective was to determine if expression of EMT genes contributed to risk of metastasis and whether there was a difference in phenotype between tumors that metastasized and those that did not. Given the lack of consensus on what technique is most accurate in determining E/M phenotype, we employed three different methods with unique transcriptomic signatures. An EMT score for each cohort was generated by: (1) core enrichment genes from a gene set enrichment analysis (GSEA) gene list, (2) a previously published default 76-gene signature (76GS) for EMT and (3) a two-sample Kolmogorov–Smirnov (KS) test.

The EMT scores for method 1 and 2 were calculated based on the weighted sum of expression levels. Each gene was weighted by the correlation coefficient between its expression level and that of epithelial marker E-cadherin (CDH-1) in the dataset. Negative scores were interpreted as mesenchymal (M) phenotype whereas the positive scores were epithelial (E) phenotype. The EMT score for method 3 used a two-sample KS test which compared the cumulative distribution functions (CDFs) of E and M signatures. The score was determined through hypothesis testing with a null hypothesis as no difference in CDFs of M and E signatures. The alternative hypotheses were: (1) the CDF of M signature is greater than that of E signature and (2) the CDF of E signature is greater than that of M signature. Samples with a positive EMT score were classified as M while a negative EMT score were classified as E.

Differences in survival between E and M phenotype was determined by dividing samples binomially based on their EMT scores with log-rank testing and cox proportional hazards model testing performed.

Results: Fifty-one patients with SS were diagnosed between 2009 and 2019. Half (51%) were male and mean age was 35. Five patients had metastasis at presentation and an additional 11 patients developed metastasis thereafter. Of the 51 patients, 32 had specimens located and of sufficient quality for RNA sequencing. Twenty-one specimens were treatment-naïve and therefore included in the analysis.

RNA sequencing revealed differentially expressed genes between cohorts in untreated primary tumors with FDR < 0.1 threshold (n=69) visualized in heat map figure 1. The hallmark EMT gene set identified by GSEA was significantly expressed in the metastatic cohort compared to nonmetastatic (normalized enrichment score 1.71; p-val 0.025). Additionally, there was a phenotypic difference between cohorts demonstrated by all three testing methods. In the 76GS method, 52 of the 76 genes reported in the default EMT signature were expressed in our samples. Most metastatic specimens had negative EMT scores (M phenotype) whereas non-metastatic specimens were generally more positive (E phenotype) (Figure 2a). Concordant, statistically significant findings were found using GSEA transcriptomic signature as well as the KS method, wherein the metastatic cohort had a more M phenotype than the non-metastatic cohort. In KS method, all EMT scores were positive, given SS is a mesenchymal tumor (Figure 3).

Patients with an M phenotype were consistently found to have significantly worse outcomes (Figure 4). The log-rank testing demonstrated χ²(1, N=21) = 4.05 (p-value 0.04) for 76GS method, and χ²(1, N=21) = 11.05 (p-value 0.0009) for GSEA core enrichment genes. The 5-year survival (95% CI) in the E group was 89% (71%-100%) versus 55% (32%-94%) in the M group. The hazard ratio for overall survival was 7.25 with a p-value of 0.076. Due to limited sample size, p< 0.1 was deemed significant.

Of note, there were two outliers in the nonmetastatic cohort that are visualized in the heatmap and circled in figure 2. They expressed a more M phenotype than the other nonmetastatic patients. One patient had primary pleuropulmonary synovial sarcoma (S16-1330-1A) with very aggressive disease invading along the pleural lining with a survival of only 6.6 months from diagnosis. The second patient had no evidence of metastasis three years after surgery but was lost to follow up and is well overdue for surveillance imaging (S18-42136).

Conclusion: The importance of EMT/MET has been well described in carcinoma but its impact in sarcoma has not been fully elaborated. Sarcomas are mesenchymal tumors with intra-tumoral heterogeneity. Using three different methods we demonstrated a consistent theme that a more mesenchymal phenotype of SS portends increased chance of metastasis compared to the more epithelial phenotype; this finding translated to a significantly worse survival with patients 7.25 times more likely to die at any point during the follow-up. Because these findings were evident in primary tumor specimens, calculating an EMT expression score at the time of diagnosis may be a tool to predict which case is more likely to metastasize. This could have implications on decisions regarding systemic therapies and how vigilant to be in surveillance.