A gp78/AMFR protein-driven gene signature that predicts breast cancer outcome.

553 Background: gp78, also known as the autocrine motility factor receptor (AMFR) or RNF45, is a polytopic RING-type E3 ubiquitin ligase resident to the endoplasmic reticulum (ER) that plays major role in the cellular response to stress by regulating ER homeostasis and signaling through its participation in the unfolded protein response (UPR) and ER associated degradation. We used machine learning (ML) and statistical modeling (SM) to assess gp78 as a protein biomarker that is an independent predictor of breast cancer (bc) survival exclusively in women of self-reported African descent as opposed to European ancestry. Methods: We examined a cohort of racially diverse 555 BC bc patients who underwent surgery for their primary BC in Greenville, NC using ML and SM approach. We leveraged the availability of RNA-seq gene expression data on a portion of our bc cohort (N=136 of 555) to construct gene expression signatures. Results: Using antibodies developed in the Weissman lab and established methods for quantitative IHC, we have found that gp78 expression is significantly increased in the tumors of bc patients compared to normal breast epithelia. In addition, we found that gp78 is expressed at significantly higher levels in bc of non-Hispanic black women (NHB) compared to non-Hispanic white women (NHW) (p=0.0038), and that bc subtypes known to be more aggressive and associated with higher grades like, Basal (p=1.6e-12), Luminal B (p=2.3e-4) and HER2(8.3e-4), display significantly higher levels of gp78 compare to Luminal A. Moreover, Kaplan-Meier survival curve analyses show that gp78 protein expression is more significantly associated with poor survival in NHB women (HR:1.65, p=0.073) compared to NHW women (HR:2.01, p=0.004). Finally, multivariate analysis reveals that gp78 protein expression, based on quantitative IHC, is an independent predictor of poor bc survival exclusively in women of African (NHB) ancestry (HR:1.99, p=0.017). We leveraged the availability of RNA-seq gene expression data on a portion of our bc cohort to construct gene expression signatures or gene modules. An analysis of pooled publicly available data from 845 patients that underwent neoadjuvant chemotherapy for bc (primarily taxane and anthracycline based), reveals that gp78 gene modules are highly predictive of patient response to therapy. gp78-derived gene modules show both high fold difference and significance in predicting response to therapy (AUC:0.72) which is very similar to other multi-gene panels that are currently in clinical use including Prosigna, MammaPrint, and Oncotype Dx. Conclusions: Our results show that gp78/AMFR is an independent predictor of bc survival and response to therapy, based on race, thus implicating a role for this protein, and potentially the UPR, as underlying biological differences in tumor properties linked to genetic ancestry.