Prostate cancer is a heterogeneous disease and it is proposed that the sub-population of cancer cells that survive androgen deprivation possess a lethal phenotype. Identifying the unique biological characteristics of these cells is essential to determine their role in tumour progression and susceptibility to novel therapeutic agents. Using patient-derived xenografts (PDX) we can identify a population of castrate-tolerant cancer cells that ‘pre-exist’ in localised tumours. We hypothesise that the castrate-tolerant cells represent a select clone that evades treatment without the need to evolve under this selective pressure. To test this, we established PDX of human primary specimens from 3 men with localised prostate cancer that underwent castration and testosterone replacement. Our data analysis includes DNA copy number and protein expression to identify biomarker and molecular profiles of castrate-tolerant prostate cancer cells. Using immunohistochemistry, we showed that localisation and expression of PTEN, FoxA1, FGFR1 and p21 were similar in T-restored grafts compared to intact controls. Whole genome copy number analysis revealed genomic amplifications and deletions were similar between intact and T restored PDX tumours, and that these aberrations were consistent with the original specimen. Whilst these data do not preclude the possibility that the castrate-tolerant cells may attain further alterations after prolonged or additional therapies, they do indicate that new targeted therapies that eliminate them earlier in disease progression may prevent or delay the onset of advanced disease.