Targeting the fibroblast growth factor pathway in molecular subtypes of castration-resistant prostate cancer.

Abstract

Androgen receptor (AR) pathway inhibition remains the cornerstone for prostate cancer therapies. However, castration-resistant prostate cancer (CRPC) tumors can resist AR signaling inhibitors through AR amplification and AR splice variants in AR-positive CRPC (ARPC), and conversion to AR-null phenotypes, such as double-negative prostate cancer (DNPC) and small cell or neuroendocrine prostate cancer (SCNPC). We have shown previously that DNPC can bypass AR-dependence through fibroblast growth factor receptor (FGFR) signaling. However, the role of the FGFR pathway in other CRPC phenotypes has not been elucidated. RNA-Seq analysis was conducted on patient metastases, LuCaP patient-derived xenograft (PDX) models, and CRPC cell lines. Cell lines (C4-2B, VCaP, and 22Rv1) and ex vivo LuCaP PDX tumor cells were treated with enzalutamide (ENZA) and FGFR inhibitors (FGFRi) alone or in combination and sensitivity was determined using cell viability assays. In vivo efficacy of FGFRi in ARPC, DNPC, and SCNPC were evaluated using PDX models. RNA-Seq analysis of FGFR signaling in metastatic specimens, LuCaP PDX models, and CRPC cell lines revealed significant FGF pathway activation in AR-low PC (ARLPC), DNPC, and SCNPC tumors. In vitro/ex vivo analysis of erdafitinib and CH5183284 demonstrated robust and moderate growth suppression of ARPC, respectively. In vivo studies using four ARPC PDX models showed that combination ENZA and CH5183284 significantly suppressed tumor growth. Additional in vivo studies using four ARPC PDX models revealed that erdafitinib monotherapy was as effective as ENZA in suppressing tumor growth, and there was limited combination benefit. Furthermore, two of three DNPC models and two of four SCNPC models responded to CH5183284 monotherapy, suggesting FGFRi responses were model dependent. RNA-Seq and gene set enrichment analysis of end-of-study ARPC tumors treated with FGFRi displayed decreased expression of E2F and MYC target genes and suppressed G2M checkpoint genes, whereas end-of-study SCNPC tumors had heterogeneous transcriptional responses. Although FGFRi treatments suppressed tumor growth across CRPC phenotypes, our analyses did not identify a single pathway or biomarker that would identify tumor response to FGFRi. This is very likely due to the array of FGFR1-4 expression and tumor phenotypes present in CRPC. Nevertheless, our data nominate the FGFR pathway as a clinically actionable target that promotes tumor growth in diverse phenotypes of treatment-refractory metastatic CRPC.