Xenograft-derived mRNA/miR and protein interaction networks of systemic dissemination in human prostate cancer

Distant metastasis formation is the major clinical problem in prostate cancer (PCa) and the underlying mechanisms remain poorly understood. Our aim was to identify novel molecules that functionally contribute to human PCa systemic dissemination based on unbiased approaches. We compared mRNA, microRNA (miR) and protein expression levels in established human PCa xenograft tumours with high (PC-3), moderate (VCaP) or weak (DU-145) spontaneous micrometastatic potential. By focussing on those mRNAs, miRs and proteins that were differentially regulated among the xenograft groups and known to interact with each other we constructed dissemination-related mRNA/miR and protein/miR networks. Next, we clinically and functionally validated our findings. Besides known determinants of PCa progression and/or metastasis, our interaction networks include several novel candidates. We observed a clear role of epithelial-to-mesenchymal transition (EMT) pathways for PCa dissemination, which was additionally confirmed by an independent human PCa model (ARCAP-E/-M). Two converging nodes, CD46 (decreasing with metastatic potential) and DDX21 (increasing with metastatic potential), were used to test the clinical relevance of the networks. Intriguingly, both network nodes consistently added prognostic information for patients with PCa whereas CD46 loss predicted poor outcome independent of established parameters. Accordingly, depletion of CD46 in weakly metastatic PCa cells induced EMT-like properties in vitro and spontaneous micrometastasis formation in vivo. The clinical and functional relevance of the dissemination-related interaction networks shown here could be successfully validated by proof-of-principle experiments. Therefore, we suggest a direct pro-metastatic, clinically relevant role for the multiple novel candidates included in this study; these should be further exploited by future studies. •Spontaneous metastasis formation of human prostate cancer can be modelled in vivo.•Transcriptomics and proteomics reveal metastasis-related interaction networks.•The network converging nodes could be clinically validated for proof-of-concept.•CD46 independently predicts poor prognosis and has a functional role for metastasis.•Interaction networks include several novel potential determinants of metastasis.