Copy number architectures define treatment-mediated selection of lethal prostate cancer clones.

Despite initial responses to hormone treatment, metastatic prostate cancer invariably evolves to a lethal state. To characterize the intra-patient evolutionary relationships of metastases that evade treatment, we perform genome-wide copy number profiling and bespoke approaches targeting the androgen receptor (AR) on 167 metastatic regions from 11 organs harvested post-mortem from 10 men who died from prostate cancer. We identify diverse and patient-unique alterations clustering around the AR in metastases from every patient with evidence of independent acquisition of related genomic changes within an individual and, in some patients, the co-existence of AR-neutral clones. Using the genomic boundaries of pan-autosome copy number changes, we confirm a common clone of origin across metastases and diagnostic biopsies, and identified in individual patients, clusters of metastases occupied by dominant clones with diverged autosomal copy number alterations. These autosome-defined clusters are characterized by cluster-specific AR gene architectures, and in two index cases are topologically more congruent than by chance (p-values 3.07 × 10-8 and 6.4 × 10-4). Integration with anatomical sites suggests patterns of spread and points of genomic divergence. Here, we show that copy number boundaries identify treatment-selected clones with putatively distinct lethal trajectories.

Allele-informed copy number evaluation of plasma DNA samples from metastatic prostate cancer patients: the PCF_SELECT consortium assay.

Sequencing of cell-free DNA (cfDNA) in cancer patients' plasma offers a minimally-invasive solution to detect tumor cell genomic alterations to aid real-time clinical decision-making. The reliability of copy number detection decreases at lower cfDNA tumor fractions, limiting utility at earlier stages of the disease. To test a novel strategy for detection of allelic imbalance, we developed a prostate cancer bespoke assay, PCF_SELECT, that includes an innovative sequencing panel covering ∼25 000 high minor allele frequency SNPs and tailored analytical solutions to enable allele-informed evaluation. First, we assessed it on plasma samples from 50 advanced prostate cancer patients. We then confirmed improved detection of genomic alterations in samples with <10% tumor fractions when compared against an independent assay. Finally, we applied PCF_SELECT to serial plasma samples intensively collected from three patients previously characterized as harboring alterations involving DNA repair genes and consequently offered PARP inhibition. We identified more extensive pan-genome allelic imbalance than previously recognized in prostate cancer. We confirmed high sensitivity detection of BRCA2 allelic imbalance with decreasing tumor fractions resultant from treatment and identified complex ATM genomic states that may be incongruent with protein losses. Overall, we present a framework for sensitive detection of allele-specific copy number changes in cfDNA.

Plasma androgen receptor and response to adapted and standard docetaxel regimen in castration-resistant prostate cancer: A multicenter biomarker study.

BACKGROUND: Plasma AR status has been identified as a potential biomarker of response in metastatic castration-resistant prostate cancer (mCRPC) patients receiving docetaxel or AR-targeted therapies. However, the relevance of plasma AR in the overall management of CRPC patients receiving different docetaxel doses is unknown. PATIENTS AND METHODS: This was a multi-institution study of associations between baseline plasma AR copy number status, assessed by droplet digital PCR, and outcome in 325 mCRPC patients receiving docetaxel at standard or adapted regimen at the discretion of the treating physician. Upon analysis, patients were assigned randomly to either a training (n = 217) or validation (n = 108) cohort. RESULTS: In the training cohort, AR-gained patients treated with adapted docetaxel regimen had a significantly worse median progression-free survival (PFS) (3.8 vs 6.3 months, hazard ratio [HR] 2.58, 95% confidence interval [CI] 1.34-4.95, p < 0.0001), median overall survival (10.8 vs 20.6 months, HR 1.98, 95% CI 1.09-3.62, p = 0.0064) and PSA response (PSA > -50%: odds ratio 4.88 95%CI 1.55-14.32, p = 0.013) as compared to plasma AR normal patients. These findings were all confirmed in the validation cohort. However, in patients treated with standard docetaxel regimen, these differences were not seen. The interaction between AR CN status and dose reduction of docetaxel was considered as independent factor for PFS in both the training and validation cohort (HR 2.84, 95% CI 1.41-5.73, p = 0.003, and HR 4.79, 95% CI 1.79-12.82, p = 0.002). CONCLUSION: Despite the retrospective non-randomised design of this study, our hypothesis-generating findings could suggest plasma AR as a potential biomarker for optimal docetaxel timing and dose in mCRPC patients. Prospective trials are warranted.

Plasma tumour DNA as an early indicator of treatment response in metastatic castration-resistant prostate cancer.

BACKGROUND: Plasma tumour DNA (ptDNA) levels on treatment are associated with response in a variety of cancers. However, the role of ptDNA in prostate cancer monitoring remains largely unexplored. Here we characterised on-treatment ptDNA dynamics and evaluated its potential for early assessment of therapy efficacy for metastatic castration-resistant prostate cancer (mCRPC). METHODS: Between 2011 and 2016, 114 sequential plasma samples from 43 mCRPC abiraterone-treated patients were collected. Targeted next-generation sequencing was performed to determine ptDNA fraction. ptDNA progressive disease was defined as a rise in the fraction compared to the pre-treatment. RESULTS: A ptDNA rise in the first on-treatment sample (interquartile range (IQR) 2.6-3.7 months) was significantly associated with increased risk of early radiographic or any prostate-specific antigen (PSA) rise (odds ratio (OR) = 15.8, 95% confidence interval (CI) 3.5-60.2, p = 0.0002 and OR = 6.0, 95% CI 1.6-20.0, p = 0.01, respectively). We also identified exemplar cases that had a rise in PSA or pseudoprogression secondary to bone flare but no rise in ptDNA. In an exploratory analysis, initial ptDNA change was found to associate with the duration of response to prior androgen deprivation therapy (p < 0.0001) but not to prior taxanes (p = 0.32). CONCLUSIONS: We found that ptDNA assessment for therapy monitoring in mCRPC is feasible and provides data relevant to the clinical setting. Prospective evaluation of these findings is now merited.

Genome-wide plasma DNA methylation features of metastatic prostate cancer.

Tumor DNA circulates in the plasma of cancer patients admixed with DNA from noncancerous cells. The genomic landscape of plasma DNA has been characterized in metastatic castration-resistant prostate cancer (mCRPC) but the plasma methylome has not been extensively explored. Here, we performed next-generation sequencing (NGS) on plasma DNA with and without bisulfite treatment from mCRPC patients receiving either abiraterone or enzalutamide in the pre- or post-chemotherapy setting. Principal component analysis on the mCRPC plasma methylome indicated that the main contributor to methylation variance (principal component one, or PC1) was strongly correlated with genomically determined tumor fraction (r = -0.96; P < 10-8) and characterized by hypermethylation of targets of the polycomb repressor complex 2 components. Further deconvolution of the PC1 top-correlated segments revealed that these segments are comprised of methylation patterns specific to either prostate cancer or prostate normal epithelium. To extract information specific to an individual's cancer, we then focused on an orthogonal methylation signature, which revealed enrichment for androgen receptor binding sequences and hypomethylation of these segments associated with AR copy number gain. Individuals harboring this methylation pattern had a more aggressive clinical course. Plasma methylome analysis can accurately quantitate tumor fraction and identify distinct biologically relevant mCRPC phenotypes.

Identification of single nucleotide variants using position-specific error estimation in deep sequencing data.

BACKGROUND: Targeted deep sequencing is a highly effective technology to identify known and novel single nucleotide variants (SNVs) with many applications in translational medicine, disease monitoring and cancer profiling. However, identification of SNVs using deep sequencing data is a challenging computational problem as different sequencing artifacts limit the analytical sensitivity of SNV detection, especially at low variant allele frequencies (VAFs). METHODS: To address the problem of relatively high noise levels in amplicon-based deep sequencing data (e.g. with the Ion AmpliSeq technology) in the context of SNV calling, we have developed a new bioinformatics tool called AmpliSolve. AmpliSolve uses a set of normal samples to model position-specific, strand-specific and nucleotide-specific background artifacts (noise), and deploys a Poisson model-based statistical framework for SNV detection. RESULTS: Our tests on both synthetic and real data indicate that AmpliSolve achieves a good trade-off between precision and sensitivity, even at VAF below 5% and as low as 1%. We further validate AmpliSolve by applying it to the detection of SNVs in 96 circulating tumor DNA samples at three clinically relevant genomic positions and compare the results to digital droplet PCR experiments. CONCLUSIONS: AmpliSolve is a new tool for in-silico estimation of background noise and for detection of low frequency SNVs in targeted deep sequencing data. Although AmpliSolve has been specifically designed for and tested on amplicon-based libraries sequenced with the Ion Torrent platform it can, in principle, be applied to other sequencing platforms as well. AmpliSolve is freely available at https://github.com/dkleftogi/AmpliSolve .

Plasma AR status and cabazitaxel in heavily treated metastatic castration-resistant prostate cancer.

BACKGROUND: Plasma androgen receptor (AR) copy number status has been identified as a potential biomarker of response in patients with metastatic castration-resistant prostate cancer (mCRPC) receiving docetaxel or the AR-targeted therapies abiraterone or enzalutamide. However, the relevance of plasma AR status in the context of cabazitaxel therapy is unknown. PATIENTS AND METHODS: Between September 2011 and January 2018, pretherapy plasma samples were collected from 155 patients treated with second- or third-line cabazitaxel at standard or reduced dose in different biomarker protocols. Droplet digital polymerase chain reaction was used to identify plasma AR gain and normal samples. The primary objective was to evaluate associations of plasma AR status with treatment outcome. In an exploratory analysis, a comparison between plasma AR and treatment type was investigated by incorporating updated data from our prior study of 85 post-docetaxel patients receiving abiraterone or enzalutamide. RESULTS: We observed a shorter median overall survival (OS) and progression-free survival (PFS) in AR-gained compared to AR-normal patients (OS 10.5 versus 14.1 months, hazard ratio (HR) = 1.44, 95% confidence interval [CI] 0.98-2.13, P = 0.064 and PFS 4.0 versus 5.0 months, HR = 1.47, 95% CI 1.05-2.07, P = 0.024). In patients with mCRPC receiving second-line therapies, a significant treatment interaction was observed between plasma AR and cabazitaxel versus AR-directed therapies for OS (P = 0.041) but not PFS (P = 0.244). In an exploratory analysis, AR-gained patients treated with initial reduced dose of cabazitaxel had a significantly shorter median OS (7.3 versus 11.5 months, HR = 1.95, 95% CI 1.13-3.38, P = 0.016) and PFS (2.7 versus 5.0 months, HR = 2.27, 95% CI 1.39-3.71, P = 0.001). CONCLUSION: Plasma AR status has a potential clinical utility in patients being considered for cabazitaxel. Validation of these findings in prospective trials is warranted.

Plasma Androgen Receptor and Docetaxel for Metastatic Castration-resistant Prostate Cancer.

Plasma androgen receptor (AR) gain identifies metastatic castration-resistant prostate cancer (mCRPC) patients with worse outcome on abiraterone/enzalutamide, but its relevance in the context of taxane chemotherapy is unknown. We aimed to evaluate whether docetaxel is active regardless of plasma AR and to perform an exploratory analysis to compare docetaxel with abiraterone/enzalutamide. This multi-institutional study was a pooled analysis of AR status, determined by droplet digital polymerase chain reaction, on pretreatment plasma samples. We evaluated associations between plasma AR and overall/progression-free survival (OS/PFS) and prostate-specific antigen (PSA) response rate in 163 docetaxel-treated patients. OS was significantly shorter in case of AR gain (hazard ratio [HR]=1.61, 95% confidence interval [CI]=1.08-2.39, p=0.018), but not PFS (HR=1.04, 95% CI 0.74-1.46, p=0.8) or PSA response (odds ratio=1.14, 95% CI=0.65-1.99, p=0.7). We investigated the interaction between plasma AR and treatment type after incorporating updated data from our prior study of 73 chemotherapy-naïve, abiraterone/enzalutamide-treated patients, with data from 115 first-line docetaxel patients. In an exploratory analysis of mCRPC patients receiving first-line therapies, a significant interaction was observed between plasma AR and docetaxel versus abiraterone/enzalutamide for OS (HR=0.16, 95% CI=0.06-0.46, p<0.001) and PFS (HR=0.31, 95% CI=0.12-0.80, p=0.02). Specifically, we reported a significant difference for OS favoring abiraterone/enzalutamide for AR-normal patients (HR=1.93, 95% CI=1.19-3.12, p=0.008) and a suggestion favoring docetaxel for AR-gained patients (HR=0.53, 95% CI=0.24-1.16, p=0.11). These data suggest that AR-normal patients should receive abiraterone/enzalutamide and AR-gained could benefit from docetaxel. This treatment selection merits prospective evaluation in a randomized trial. PATIENT SUMMARY: We investigated whether plasma androgen receptor (AR) predicted outcome in metastatic castration-resistant prostate cancer (mCRPC) patients treated with docetaxel, and we performed an exploratory analysis in patients treated with docetaxel or AR-directed drugs as first-line mCRPC therapy. We showed that plasma AR normal favored hormonal treatment, whilst plasma AR-gained patients may have had a longer response to docetaxel, suggesting that plasma AR status could be a useful treatment selection biomarker.

Plasma Androgen Receptor Copy Number Status at Emergence of Metastatic Castration-Resistant Prostate Cancer: A Pooled Multicohort Analysis.

PURPOSE: Increases in androgen receptor (AR) copy number (CN) can be detected in plasma DNA when patients develop metastatic castration-resistant prostate cancer. We aim to evaluate the association between AR CN as a continuous variable and clinical outcome. PATIENTS AND METHODS: PCR2023 was an international, multi-institution, open-label, phase II study of abiraterone acetate plus prednisolone (AAP) or abiraterone acetate plus dexamethasone that included plasma AR assessment as a predefined exploratory secondary end point. Plasma AR CN data (ClinicalTrials.gov identifier: NCT01867710) from this study (n = 133) were pooled with data from the following three other cohorts: cohort A, which was treated with either AAP or enzalutamide (n = 73); the PREMIERE trial (ClinicalTrials.gov identifier: NCT02288936) of biomarkers for enzalutamide (n = 94); and a phase II trial from British Columbia (ClinicalTrials.gov identifier: NCT02125357) that randomly assigned men to either AAP or enzalutamide (n = 201). The primary outcome measures for the biomarker analysis were overall survival and progression-free survival. RESULTS: Using multivariable fractional polynomials analysis using Cox regression models, a nonlinear relationship between plasma AR CN and outcome was identified for overall survival, where initially for small incremental gains in CN there was a large added hazard ratio that plateaued at higher CN. The CN cut point associated with the highest local hazard ratio was 1.92. A similar nonlinear association was observed with progression-free survival. In an exploratory analysis of PCR2023, the time from start of long-term androgen-deprivation therapy to start of AAP or abiraterone acetate plus dexamethasone was significantly shorter in patients with plasma AR CN of 1.92 or greater than patients with plasma AR CN of less than 1.92 (43 v 130 weeks, respectively; P = .005). This was confirmed in cohort A (P = .003), the PREMIERE cohort (P = .03), and the British Colombia cohort (P = .003). CONCLUSION: Patients with metastatic castration-resistant prostate cancer can be dichotomized by a plasma AR CN cut point of 1.92. Plasma AR CN value of 1.92 or greater identifies aggressive disease that is poorly responsive to AR targeting and is associated with a prior short response to primary androgen-deprivation therapy.

Circulating tumor DNA in advanced prostate cancer: transitioning from discovery to a clinically implemented test.

The genomic landscape of metastatic castration-resistant prostate cancer (mCRPC) differs from that of the primary tumor and is dynamic during tumor progression. The real-time and repeated characterization of this process via conventional solid tumor biopsies is challenging. Alternatively, circulating cell-free DNA (cfDNA) containing circulating tumor DNA (ctDNA) can be obtained from patient plasma using minimally disruptive blood draws and is amenable to sequential analysis. ctDNA has high overlap with the genomic sequences of biopsies from metastases and has the advantage of being representative of multiple metastases. The availability of techniques with high sensitivity and specificity, such as next-generation sequencing (NGS) and digital PCR, has greatly contributed to the development of the cfDNA field and enabled the detection of genomic alterations at low ctDNA fractions. In mCRPC, a number of clinically relevant genomic alterations have been tracked in ctDNA, including androgen receptor (AR) aberrations, which have been shown to be associated with an adverse outcome to novel antiandrogen therapies, and alterations in homologous recombination repair (HRR) genes, which have been associated with a response to PARP inhibitors. Several clinical applications have been proposed for cfDNA analysis, including its use as a prognostic tool, as a predictive biomarker, to monitor tumor response and to identify novel mechanisms of resistance. To date, the cfDNA analysis has provided interesting results, but there is an urgent need for these findings to be confirmed in prospective clinical trials.

Phase II pilot study of the prednisone to dexamethasone switch in metastatic castration-resistant prostate cancer (mCRPC) patients with limited progression on abiraterone plus prednisone (SWITCH study).

BACKGROUND: Despite most metastatic castration-resistant prostate cancer (mCRPC) patients benefit from abiraterone acetate plus prednisone 5 mg bid (AA + P), resistance eventually occurs. Long-term use of prednisone has been suggested as one of the mechanisms driving resistance, which may be reversed by switching to another steroid. METHODS: SWITCH was a single-arm, open-label, single-stage phase II study. The primary objective was to evaluate the antitumour activity of abiraterone acetate plus dexamethasone 0.5 mg daily (AA + D) in mCRPC patients progressing to AA + P. Clinically stable mCRPC patients who had prostate-specific antigen (PSA) and/or limited radiographic progression after at least 12 weeks on AA + P, were eligible. The primary endpoint was measured as the proportion of patients achieving a PSA decline of ≥ 30% (PSA30) from baseline after 6 weeks on AA + D. Secondary endpoints included: PSA50 response rate at 12 weeks, time to biochemical and radiological progression, overall survival, safety profile evaluation, benefit from subsequent treatment lines and the identification of biomarkers of response (AR copy number, TMPRSS2-ERG status and PTEN expression). RESULTS: Twenty-six patients were enrolled. PSA30 and PSA50 were 46.2% and 34.6%, respectively. Median time to biochemical and radiological progression were 5.3 and 11.8 months, respectively. Two radiological responses were observed. Median overall survival was 20.9 months. Patients with AR gain detected in plasma circulating tumour DNA did not respond to switch, whereas patients with AR normal status benefited the most. No significant toxicities were observed and PSA50 response rate to subsequent taxane was 50%. CONCLUSIONS: In selected clinical stable mCRPC patients with limited disease progression on AA + P, a steroid switch from prednisone to dexamethasone can lead to PSA and radiological responses.

Assessment of the Validity of Nuclear-Localized Androgen Receptor Splice Variant 7 in Circulating Tumor Cells as a Predictive Biomarker for Castration-Resistant Prostate Cancer.

Importance: A blood test to determine whether to treat patients with metastatic castration-resistant prostate cancer (mCRPC) with an androgen receptor signaling (ARS) inhibitor or taxane is an unmet medical need. Objective: To determine whether a validated assay for the nuclear-localized androgen receptor splice variant 7 (AR-V7) protein in circulating tumor cells can determine differential overall survival among patients with mCRPC treated with taxanes vs ARS inhibitors. Design, Setting, and Participants: This blinded correlative study conducted from December 31, 2012, to September 1, 2016, included 142 patients with histologically confirmed mCRPC and who were treated at Memorial Sloan Kettering Cancer Center, The Royal Marsden, or the London Health Sciences Centre. Blood samples were obtained prior to administration of ARS inhibitors or taxanes as a second-line or greater systemic therapy for progressing mCRPC. Main Outcomes and Measures: Overall survival after treatment with an ARS inhibitor or taxane in relation to pretherapy AR-V7 status. Results: Among the 142 patients in the study (mean [SD] age, 69.5 [9.6] years), 70 were designated as high risk by conventional prognostic factors. In this high-risk group, patients positive for AR-V7 who were treated with taxanes had superior overall survival relative to those treated with ARS inhibitors (median overall survival, 14.3 vs 7.3 months; hazard ratio, 0.62; 95% CI, 0.28-1.39; P = .25). Patients negative for AR-V7 who were treated with ARS inhibitors had superior overall survival relative to those treated with taxanes (median overall survival, 19.8 vs 12.8 months; hazard ratio, 1.67; 95% CI, 1.00-2.81; P = .05). Conclusions and Relevance: This study suggests that nuclear-localized AR-V7 protein in circulating tumor cells can identify patients who may live longer with taxane chemotherapy vs ARS inhibitor treatment.

Plasma DNA and Metastatic Castration-Resistant Prostate Cancer: The Odyssey to a Clinical Biomarker Test.

Comprehensive plasma DNA analysis identifies clinically actionable genomic aberrations. Cancers harboring disruption of TP53 or BRCA2 or ATM detected in plasma have significantly worse outcomes on novel AR targeting. Cancer Discov; 8(4); 392-4. ©2018 AACR.See related article by Annala et al., p. 444.

Prostate-specific Antigen Decline After 4 Weeks of Treatment with Abiraterone Acetate and Overall Survival in Patients with Metastatic Castration-resistant Prostate Cancer.

BACKGROUND: The availability of multiple new treatments for metastatic castration-resistant prostate cancer (mCRPC) mandates earlier treatment switches in the absence of a response. A decline in prostate-specific antigen (PSA) is widely used to monitor treatment response, but is not validated as an intermediate endpoint for overall survival (OS). OBJECTIVE: To evaluate the association between early PSA decline and OS following abiraterone acetate (AA) treatment. DESIGN, SETTING, AND PARTICIPANTS: We identified mCRPC patients treated with AA before or after docetaxel at the Royal Marsden NHS Foundation Trust between 2006 and 2014. Early PSA decline was defined as a 30% decrease in PSA at 4 wk relative to baseline, and early PSA rise as a 25% increase. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: Association with OS was analyzed using multivariate Cox regression and log-rank analyses. Spearman's rho correlation coefficient (r) was calculated to evaluate the association between PSA changes at 4 wk and 12 wk. RESULTS AND LIMITATIONS: There were 274 patients eligible for this analysis. A 30% PSA decline at 4 wk was associated with longer OS (25.8 vs 15.1 mo; hazard ratio [HR] 0.47, p<0.001), and a 25% PSA rise at 4 wk with shorter OS (15.1 vs 23.8 mo; HR 1.7, p=0.001) in both univariate and multivariable models. The percentage PSA decline at 4 wk was significantly correlated with the percentage PSA change at 12 wk (r=0.82; p<0.001). Patients achieving a 30% PSA decline at 4 wk were 11.7 times more likely to achieve a 50% PSA decrease at 12 wk (sensitivity 90.9%, specificity 79.4%). Limitations include the retrospective design of this analysis. CONCLUSIONS: Patients not achieving 30% PSA decline after 4 wk of AA have a lower likelihood of achieving PSA response at 12 wk and significantly inferior OS. Prospective multicentre validation studies are needed to confirm these findings. PATIENT SUMMARY: Prostate-specific antigen (PSA) is commonly used to evaluate response to treatment in metastatic castration-resistant prostate cancer. Expert recommendations discourage reliance on PSA changes earlier than 12 wk after treatment initiation. Our data suggest that early PSA changes are associated with survival in patients receiving abiraterone acetate.

Managing Metastatic Castration-Resistant Prostate Cancer in the Pre-chemotherapy Setting: A Changing Approach in the Era of New Targeted Agents.

In recent years, the therapeutic options for treating men with metastatic castration-resistant prostate cancer have increased substantially. The hormonal treatments abiraterone acetate and enzalutamide, the chemotherapeutics docetaxel and cabazitaxel, the radiopharmaceutical alpharadin and the immunotherapeutic Sipuleucel-T have entered the field. Additionally, corticosteroids, which are used extensively, have documented activity but no documented survival benefit. Physicians treating patients with metastatic prostate cancer immediately after castration resistance develops currently have at least four different options to choose from for the first treatment. These therapeutic choices and their several possible ways of sequential use have not yet been compared to each other head-to-head and may never be. Therefore, there is an unmet need to inform their use with prospective clinical data. Additionally, the new indications of docetaxel for hormone naïve prostate cancer is changing the landscape of prostate cancer treatment and questions the traditional classifications 'pre-chemotherapy' and 'post-chemotherapy'. In this work we attempt to address these challenges in the treatment of metastatic castration-resistant prostate cancer with the focus mainly on the non-cytotoxic agents. We try to integrate available clinical and preclinical information to suggest optimal ways of treatment.

Plasma AR and abiraterone-resistant prostate cancer.

Androgen receptor (AR) gene aberrations are rare in prostate cancer before primary hormone treatment but emerge with castration resistance. To determine AR gene status using a minimally invasive assay that could have broad clinical utility, we developed a targeted next-generation sequencing approach amenable to plasma DNA, covering all AR coding bases and genomic regions that are highly informative in prostate cancer. We sequenced 274 plasma samples from 97 castration-resistant prostate cancer patients treated with abiraterone at two institutions. We controlled for normal DNA in patients' circulation and detected a sufficiently high tumor DNA fraction to quantify AR copy number state in 217 samples (80 patients). Detection of AR copy number gain and point mutations in plasma were inversely correlated, supported further by the enrichment of nonsynonymous versus synonymous mutations in AR copy number normal as opposed to AR gain samples. Whereas AR copy number was unchanged from before treatment to progression and no mutant AR alleles showed signal for acquired gain, we observed emergence of T878A or L702H AR amino acid changes in 13% of tumors at progression on abiraterone. Patients with AR gain or T878A or L702H before abiraterone (45%) were 4.9 and 7.8 times less likely to have a ≥50 or ≥90% decline in prostate-specific antigen (PSA), respectively, and had a significantly worse overall [hazard ratio (HR), 7.33; 95% confidence interval (CI), 3.51 to 15.34; P = 1.3 × 10(-9)) and progression-free (HR, 3.73; 95% CI, 2.17 to 6.41; P = 5.6 × 10(-7)) survival. Evaluation of plasma AR by next-generation sequencing could identify cancers with primary resistance to abiraterone.