In this study, we have shown in two independent sets of PCa samples that the PTEN genomic deletion was associated with early disease recurrence and reduced levels of AR expression. In microarray gene expression data, the PTEN deletion was also associated with a down regulation of AR-driven genes.
The frequency of PTEN deletion in our FISH study (40%) is within the range of previous reports
[8, 10, 12, 14, 15]. Our survival analysis further confirms the association of PTEN genomic deletion and poor outcome of PCa reported earlier
 and its potential use as a prognostic marker. Clinical relevance is also supported by the recent literature detecting PTEN deletion at high frequency in CRPC samples
, in circulating tumor cells
 and its association with PCa death
[11, 28]. Further validation in larger cohorts would be critical to compare its predictive value with the current prognostication tools.
The intriguing finding of our study was the reduced levels of AR expression quantified by H-score in tumors harboring a PTEN deletion. We found a similar association between PTEN deletion and AR transcript levels in a PCa microarray dataset. The differential expression of AR according to the PTEN tumor status has not been well documented so far. A pilot IHC study has found a positive correlation between AR and PTEN expression
. In contrast, Sircar et al. reported a positive correlation between PTEN deletion status and AR expression
 in CRPC samples. These results likely reflect two different stages of the disease: CRPC and untreated PCa. The genomic amplification of AR is known to occur in CRPC but rarely in untreated PCa
, thereby explaining differences in results.
Previous in vitro studies in cell lines derived from advanced PCa suggested that PTEN could act as suppressor of AR activity
[31, 32]. It was also reported that the activation of PI3K/AKT pathway can suppress the AR activity in low passage LNCaP and enhance AR activity in high passage, hence suggesting modulation as cells evolve towards less responsive status
. In models representing less advanced disease, re-expression of PTEN in PTEN null murine cells did not affect AR expression, but upregulated the AR transcriptional activity
. Another group reported that PTEN null murine prostate cells had a reduced AR protein levels compared to wild-type PTEN cells and the AR protein levels were partly restored by the PI3K/mTOR inhibitor BEZ235
. The latter observation would suggest that the activation of PI3K pathway may in part explain the reduced AR levels in PTEN deleted tumors. A shown by Lin et al., it is also possible that PTEN interacts directly with AR and promotes its degradation
. Underlying mechanisms of how PTEN deletion in human tumors is associated with lower AR expression and transcriptional activity need to be further explored.
Given their reduced levels of AR expression, the PTEN deleted tumor cells are expected to be less responsive to androgen ablation treatment. In support of this hypothesis, it was reported that CRPC and early biochemical recurrence were associated with reduced immunoreactivity of PTEN and AR in the PCa samples harvested before treatment initiation
. The addition of an inhibitor of PI3K/mTOR to the standard androgen ablation treatment of advanced PCa may therefore be beneficial to patients with PTEN deleted tumor.
Some previous studies have found that low levels of AR were associated with PCa recurrence
[36, 37] while others reported the opposite
[38, 39]. In our study, AR levels of expression were not significantly associated with PCa recurrence. The antibody used, IHC technique and scoring methods may explain the differences in the findings. Given the limited number of patients of our study, a detailed analysis of AR and PTEN in a large cohort of patients with follow-up is warranted.
During the course of our study, two groups also showed a reduced expression of androgen regulated genes in human PTEN deleted PCa by microarray analysis
[34, 35]. In our analysis, the androgen regulated genes enriched in tumor with no deletion of PTEN include genes expressed in normal prostate luminal epithelium such as KLK3 (PSA), TMPRSS2, and NKX3-1. Of interest, the list includes AZGP1 previously reported as a surrogate marker for subtype-1 tumors, a favourable prognostic subclass of PCa defined by gene expression pattern analysis
. AZGP1 prognostic value was further confirmed by two other studies
[40, 41]. Previous GSEA has also revealed enrichment of androgen-responsive genes in subtype-1 tumors
. Consistant with our findings, the confirmation of intact PTEN status in subtype-1 tumors from the array CGH data may, at least in part, explain their androgen-regulated gene expression feature and good clinical outcome.