Genome-wide individual prostate cancer (PC) cell expression profiling identifies heterogeneity in effects of androgen receptor (AR) targeting
Karolina Nowakowska, Paolo Cremaschi, Daniel Wetterskog, Stefano Lise, Gerhardt Attard
The Institute of Cancer Research, Centre for Evolution and Cancer, London, UK
Introduction: Patients with metastatic PC initially respond to androgen deprivation therapy (ADT) but invariably relapse with resistant disease. We hypothesise that sub-populations of cells rapidly adapt to AR targeting by up-regulation of distinct expression profiles to give rise to incurable disease.
Methods: Single-cell RNAseq data from PC cells after exposure to the potent AR antagonist, enzalutamide were compared to untreated cells. Following AR inhibition, cells were sorted either by FACS or processed using the DropSeq- high-throughput microfluidic system capable of capturing and barcoding the transcriptome of individual cells.
Results: RNAseq and ddPCR obtained from FACS sorted and SMART-seq processed cells identified 20% of treated cells maintained expression of AR-regulated PSA and/or TMPRSS2, suggesting sub-populations persist with active AR signalling despite potent AR targeting. Single-cell RNAseq confirmed a high correlation with ddPCR (R-square: 0.715, P value<2.2e-16) and whole-transcriptome analysis identified an association between cell cycle genes and AR expression. The use of microfluidics allowed us to increase the number of tested cells to 1000. A low doublet rate was confirmed by a clear separation of mouse and human transcripts in a 1:1 mix of LNCaP and NIH-3T3 cells. High-quality mappable RNAseq data identified highly variable sets of genes within and across tested groups and allowed clustering of cells into sub-populations. We observed a distinct separation between treated vs untreated cells based on the transcription profile of ~350 most variable genes, with the emergence of Treatment Outlier cells that clustered with the untreated group.
Conclusion: Single cell analyses reveals changes within the transcriptome that may be missed in a bulk sample approach. Our model offers an opportunity to characterise outlier cells/sub-populations in order to dissect treatment resistance and identify new therapeutic targets.