Deciphering Molecular Subtypes in Advanced Prostate Cancer by Transcriptional Profiling of Circulating Tumor Cells

Abstract

Standard therapy for prostate cancer is directed against the androgen receptor (AR) signaling pathway. As a result of increased therapeutic pressure from novel hormonal agents, patients are more frequently progressing to AR-independent, aggressive variant prostate cancer (AVPC). The acquisition of neuroendocrine features in a transdifferentiation process leads to the emergence of neuroendocrine prostate cancer (NEPC) and presents one resistance mechanism to AR-targeted therapy. As PSA no longer is a reliable biomarker in AVPC and NEPC, there is an urgent need to identify new biomarkers for AR-independent progression and transdifferentiation. The molecular analysis of circulating tumor cells (CTC) and tumor cell fragments shed into body fluids is known as liquid biopsy (LBx). LBx poses significantly reduced risk to patients, can be performed repeatedly and represents tumor heterogeneity more effectively than a tissue biopsy. Thus, this study aimed to establish a liquid biopsy-based biomarker for the detection and monitoring of patients with NEPC. For this purpose, a transcript panel was selected based on literature review and validated in cell lines and published tumor tissue data sets. The combined workflow of CTC enrichment and transcript detection was validated in spike-in controls. Blood samples were collected from patients with AVPC, NEPC and hormone-sensitive prostate cancer at the University Medical Center Hamburg-Eppendorf. CTC counts were measured by CellSearch. After validation of the enrichment methods, the expression of a 22 gene panel was analyzed in AdnaTestenriched CTCs. Gene expression profiles were evaluated using supervised and unsupervised approaches and correlated to clinical data. Comparison of different enrichment methods showed that CTCs were more frequently detected with EPCAM-based CTC enrichment. The detection of neuroendocrine markers was not improved by Parsortix enrichment compared to AdnaTest. CellSearch analysis revealed significantly increased CTC counts in NEPC and AVPC patients accompanied by a heterogeneous CTC morphology. Gene expression profiles revealed a high degree of heterogeneity between patients with significant deregulation of several individual markers. Unsupervised analysis identified four distinct clusters, which were termed ARhigh, CTClow, amphicrine and pure NEPC. Using a random forest model, HSPC and NEPC samples could be distinguished with an AUC of 95.5 % and an out-of-bag error rate of 15.5 % in cross-validation. In longitudinal samples of single patients, the detection of neuroendocrine markers in CTCs recapitulated the clinical course of therapy response and progression. In conclusion, AVPC and NEPC patients had a high CTC burden, which facilitated subsequent molecular analyses. The gene expression profiles of the marker panel in CTCs reflected the histology of the tumors and were sufficient to distinguish intrinsic molecular subtypes of advanced prostate cancer. In the future, the convenient PCR-based analysis pipeline may allow monitoring of advanced prostate cancer patients for earlier adjustment of therapy for NEPC.