Treatment-dependent androgen receptor mutations in prostate cancer exploit multiple mechanisms to evade therapy

Abstract

Mutations in the androgen receptor (AR) that enable activation by antiandrogens occur in hormone-refractory prostate cancer, suggesting that mutant ARs are selected by treatment. To validate this hypothesis, we compared AR variants in metastases obtained by rapid autopsy of patients treated with flutamide or bicalutamide, or by excision of lymph node metastases from hormone-naïve patients. AR mutations occurred at low levels in all specimens, reflecting genetic heterogeneity of prostate cancer. Base changes recurring in multiple samples or multiple times per sample were considered putative selected mutations. Of 26 recurring missense mutations, most in the NH(2)-terminal domain (NTD) occurred in multiple tumors, whereas those in the ligand binding domain (LBD) were case specific. Hormone-naïve tumors had few recurring mutations and none in the LBD. Several AR variants were assessed for mechanisms that might underlie treatment resistance. Selection was evident for the promiscuous receptor AR-V716M, which dominated three metastases from one flutamide-treated patient. For the inactive cytoplasmically restricted splice variant AR23, coexpression with AR enhanced ligand response, supporting a decoy function. A novel NTD mutation, W435L, in a motif involved in intramolecular interaction influenced promoter-selective, cell-dependent transactivation. AR-E255K, mutated in a domain that interacts with an E3 ubiquitin ligase, led to increased protein stability and nuclear localization in the absence of ligand. Thus, treatment with antiandrogens selects for gain-of-function AR mutations with altered stability, promoter preference, or ligand specificity. These processes reveal multiple targets for effective therapies regardless of AR mutation.

Figure 1

Recurring AR mutations from prostate cancer metastases. A. Mutations found in multiple cases. For codons carrying mutations to different amino acids, both changes are shown. B. Mutations in multiple clones per sample. Only ΔQ86 was shared among groups. AR domains and repeats are boxed. Mutations above the map are silent or nonsense; mutations below are missense. Codon color indicates treatment group. Q: Polyglutamine tract, NTD: N-terminal domain, G: Polyglycine tract, DBD: DNA binding domain, H: Hinge region, LBD: Ligand binding domain. C. V716M was the only AR sequence in 3 metastases from patient 28, but did not occur in normal kidney. Electropherograms, left to right: amplified cDNA clone from metastasis 1 with G3261A (numbering from Genbank NM_000044) resulting in V716M; wild type sequence from normal kidney genomic DNA; cDNA and genomic DNA of metastasis 2. Green arrow: mutation; black arrow: wild type base.

Figure 2

Splice variant AR23 has altered subcellular localization and enhances wild type AR (wtAR) activity. A. Punctate cytoplasmic localization of AR23. AR23 transfected into PC-3 cells shows diffuse cytoplasmic localization without ligand (top) like wtAR (not shown), but forms cytoplasmic puncta with 10 nM R1881 (middle) unlike wtAR nuclear localization (bottom). AR detection used AR N20 and FITC-conjugated secondary antibody. B. Transactivation of wtAR, AR23, or 1:1 wtAR:AR23 (4 ng each) with 200 ng PSA-luc and 100 ng promoterless renilla in PC-3 cells. Cells were harvested 24 hours after agonist or antagonist treatment (HOF, hydroxyflutamide; Bic, bicalutamide) and luciferase activity assayed. Average normalized values of three independent trials are presented as percent wtAR transactivation at 1 nM R1881. C. Transrepression of NF-κB activity in CV-1 cells. WtAR or AR23 was transfected with the NF-κB reporter pBVIx-luc and NF-kB was activated with TPA. WtAR reduced activation to 20% of vector alone with 10 nM R1881 + 1 nM TPA; with AR23, NF-κB activity remained 80% of control. Error bars are standard error of the mean (SEM). * Significant difference based on p<0.05 by Student's t-test.

Figure 3

Promoter- and cell-context dependence of AR-W435L. Transactivation of ARW435L was assessed in CV-1 (A), RWPE (B), and PC-3 (C) cells, revealing promoter-specific increases most pronounced in RWPE and absent in PC-3 cells. Cells were transfected with 4 ng wtAR or AR-W435L and 400 ng of the indicated reporters. PSA activated poorly in CV-1 cells so MMTV was tested instead. After 24 hours cells were fed with phenol red free medium +/− 1 nM R1881. Average values normalized to renilla for three trials are represented as percent induced wtAR activity. D. AR-W435L increased N-C interaction in mammalian 2-hybrid assays compared to wtAR. PC-3 cells were transfected with 100 ng VP16-wtAR-NTD or W435LNTD, 100 ng Gal4-AR LBD, 200 ng Gal4-luc and 100 ng renilla. 24 hours after transfection cells were fed +/− R1881 for 24 hours. Activity was compared to induced wtAR NTD + LBD (100%). Values are the average of 5 trials +/− standard error of the mean (SEM). * p<0.05, ** p<0.005, significant differences between wtAR and AR-W435L based on Student's t-test.

Figure 4

AR-E255K has increased stability and ligand-independent nuclear localization. A. AR degradation following cycloheximide treatment. 100 ng wtAR or AR-E255K plasmid was transfected into CV-1 cells, which were treated after 24 hours with 30 μM cycloheximide. Cells were harvested at times indicated and 20 μg of total protein electrophoresed (left). AR bands from scanned immunoblots were quantified using ImageJ, values normalized to the amount of protein at time 0 (100%). The log₁₀ of the percentage was plotted versus time for wtAR and ARE255K without hormone (right). Half-life was calculated as log₁₀ of 50% based on the linear regression. AR-E255K shows a longer half-life (t_1/2= 12.5 hrs) than wtAR (t_1/2=5.2 hrs). Full gels are in Supplementary Data. B. Proteasome inhibition with lactacystin increased unliganded wtAR but not AR-E255K levels. CV-1 cells were transfected as above, treated after 24 hours with 10μM lactacystin, harvested 18 hours later, and immunoblotted. C. Following transfection into PC-3 cells, wtAR was largely cytoplasmic without hormone (top) while most cells with ARE255K showed more nuclear staining (bottom). Color images and composite are in Supplementary Data. AR was detected as for Fig. 2. Percent of cells with cytoplasmic to nuclear AR fluorescence are graphed as: C, exclusively cytoplasmic; C>N, cytoplasmic greater than nuclear; C=N, equal cytoplasmic and nuclear; N>C, nuclear greater; N, exclusively nuclear. n= number of cells counted for all three trials. Mean percentages +/− SEM are shown. D. AR-E255K showed increased transactivation of PSA-luc in RWPE cells, compared to wtAR. Bars are average % wtAR activation of three trials +/− SEM. ** p<0.005 based on Student's t-test.