Role of Circulating Tumor Cells (CTC), Androgen Receptor Full Length (AR-FL) and Androgen Receptor Splice Variant 7 (AR-V7) in a Prospective Cohort of Castration-Resistant Metastatic Prostate Cancer Patients

Background: Circulating tumor cells (CTC), androgen receptor full-length (AR-FL), and androgen receptor splice variant 7 (AR-V7) are prognostic in patients (pts) with metastatic castration-resistant prostate cancer (mCRPC). AR-V7 seems to predict resistance to androgen-receptor signaling inhibitors (ARSi). Methods: We assessed the association of CTC, AR-FL, and AR-V7 with prostate-specific antigen (PSA) response and overall survival (OS). We used a modified AdnaTest CTC-based AR-FL and AR-V7 mRNA assay. Chi-square test, Fisher Exact test, Kaplan–Meier method, log-rank test, Cox proportional hazards models were used as appropriate. Results: We enrolled 39 mCRPC pts, of those 24 started a first-line treatment for mCRPC (1L subgroup) and 15 had received at least two lines for mCRPC (>2L subgroup). CTC, AR-FL, and AR-V7 were enriched in >2L compared to 1L subgroup. Detection of these biomarkers was associated with a lower percentage of biochemical responses. Only 1/7 AR-V7+ pts had a PSA response and received cabazitaxel. Median OS was 4.7 months (95% CI 0.6–8.9) in AR-V7+ pts and not reached in AR-V7− pts. AR-V7 was the only variable with prognostic significance in the Cox model. Conclusion: AR-V7, CTC, and AR-FL are associated with advanced mCRPC and AR-V7+ predicts for shorter OS.


Introduction
The treatment landscape of metastatic castration-resistant prostate cancer (mCRPC) has radically changed over recent years. Androgen-receptor signaling inhibitors (ARSi) and taxane-based chemotherapy have been established as valid treatment options for patients with mCRPC, by demonstrating a significant survival advantage in phase 3 trials [1]. However, the androgen-receptor plasticity and the adaptive mechanisms of prostate cancer cells eventually produce resistance to therapies and treatment failure [2]. The androgen-receptor isoform encoded by splice variant 7 (AR-V7) lacks the ligand-binding domain, leading to constitutive activation of the androgen receptor, and its detection in circulating tumor cells (CTC) has been associated with resistance to ARSi [3].

Patient Selection
Ethics approval and consent to participate: informed written consent to blood collection and use for experimental purposes was obtained by all patients, following study-protocol approval by the local Ethical Committee (P.R.505REG2015).
Patients with mCRPC referred to our Institution, who were starting a first-line treatment for mCRPC (1L subgroup) or who had been already treated with at least two lines for mCRPC (>2L subgroup), including at least one ARSi and one chemotherapy regimen, were invited to participate in this study. After informed consent, patients underwent a blood sample collection and were prospectively followed with PSA assessments every 4-6 weeks, until death or the limit date of 31 December 2018.

CTC Capture and AR-FL/AR-V7 Analysis
To evaluate AR-FL and AR-V7 on CTC, 5 ml of blood was collected before starting a new line of therapy into dedicated test tube AdnaCollect (Qiagen, Hilden, Germany) and processed within 2 hours as described in the study published by Antonarakis and colleagues [3]. Briefly, CTC were isolated by immuno-magnetic beads with an EpCAM-based method (AdnaTest Prostate Cancer Select), mRNA was obtained and retro-transcribed using the Adna Test Prostate Cancer Detect and SensiScript RT kits (Qiagen, Hilden, Germany), respectively. cDNA was used for AR-FL and AR-V7 detection by a dedicated kit developed by Bird LTD company laboratories (Rezzato, Italy) [22]. CTC positivity was carried out by multiplex PCR reaction using primer specific for prostate-specific antigen (PSA), prostate-specific membrane antigen (PSMA), epidermal growth factor receptor (EGFR), and actin as internal PCR control. PCR products were analyzed on Agilent chip by Bioanalyzer electrophoresis (Agilent Technologies, Santa Clara, CA, USA). Given the intrinsic limits related to the methods for CTC isolation and RNA detection, patients with detectable (below 10 copies/mL), but not quantifiable, AR-V7 and AR-FL on CTC were considered as positive.

Statistical Analysis
The statistical analysis was conducted using the chi-square test for categorical variables and Fisher Exact test when appropriate. Pearson's correlation was used to test correlations among variables. PSA50 was the endpoint chosen for the analysis of biochemical response and was defined as a decline of at least 50% in PSA values from treatment start. OS was defined as the time elapsed from blood collection date and the date of death for any cause. OS curves were constructed according to the Kaplan-Meier method and compared using the log-rank test [23]. The Cox proportional model hazard ratios (HR) estimates and their 95% confidence intervals (CI) were also calculated. Only variables with a p value < 0.05 at univariable analysis were included in the multivariable models. The performance of the models was measured using the concordance C-index [24]. All p values were two-tailed. The IBM software Statistical Package for Social Sciences (SPSS) version 25.0 for Windows (SPSS Inc., Chicago, IL, USA) and the Software for Statistics and Data Science (STATA) version 11 were used for data analysis. Given the small sample size and the exploratory purpose of this study, adjustments for multiplicity were not performed.

Patients Cohort
Thirty-nine patients with mCRPC who met the inclusion criteria were enrolled in this study, of them 24 were included in the 1L subgroup and 15 in the >2L subgroup. After sample collection, 1L patients received abiraterone acetate, enzalutamide, or docetaxel, whereas treatments of >2L subgroup included abiraterone acetate, cabazitaxel, docetaxel, cyclophosphamide, enzalutamide, mitoxantrone, and vinorelbine. Drug dosages and schedules were those commonly used in clinical practice, according to physician choice and international guidelines on prostate cancer [25]. The main characteristics of cohort patients are summarized in Table 1 and Supplementary Table S1.
Neither of the two AR-V7+ patients treated with ARSi reached the PSA50, whereas one AR-V7+ patient showed PSA50 response to third-line cabazitaxel. Of the four highly-pretreated patients who received palliative treatments (cyclophosphamide, mitoxantrone, vinorelbine), none showed a PSA response (all were CTC positive and two were AR-V7 positive) ( Figure 2B).

Association of CTC, AR-FL, and AR-V7 with OS
We performed univariable analysis for significant prognostic variables (p < 0.05), and we found that median PSA value, treatment line at study entry, number of metastatic sites, CTC status, AR-FL status, and AR-V7 status were all associated with OS (Supplementary Table S2  Patients included in the 1L subgroup showed a higher number of PSA50 responses than patients who started a >2L treatment. PSA50 was reached in 19/24 (79.2%) patients included in the 1L cohort, compared to 4/15 (26.7%) patients included in the >2L subgroup (Figure 2A-C).
Neither of the two AR-V7+ patients treated with ARSi reached the PSA50, whereas one AR-V7+ patient showed PSA50 response to third-line cabazitaxel. Of the four highly-pretreated patients who received palliative treatments (cyclophosphamide, mitoxantrone, vinorelbine), none showed a PSA response (all were CTC positive and two were AR-V7 positive) ( Figure 2B).

Association of CTC, AR-FL, and AR-V7 with OS
We performed univariable analysis for significant prognostic variables (p < 0.05), and we found that median PSA value, treatment line at study entry, number of metastatic sites, CTC status, AR-FL status, and AR-V7 status were all associated with OS (Supplementary Table S2 Patients included in the 1L subgroup showed a higher number of PSA50 responses than patients who started a >2L treatment. PSA50 was reached in 19/24 (79.2%) patients included in the 1L cohort, compared to 4/15 (26.7%) patients included in the >2L subgroup (Figure 2A-C).
Neither of the two AR-V7+ patients treated with ARSi reached the PSA50, whereas one AR-V7+ patient showed PSA50 response to third-line cabazitaxel. Of the four highly-pretreated patients who received palliative treatments (cyclophosphamide, mitoxantrone, vinorelbine), none showed a PSA response (all were CTC positive and two were AR-V7 positive) ( Figure 2B).

Discussion
In the present study, we used a modified-AdnaTest CTC-based AR-FL and AR-V7 mRNA assay with custom primers [22], similar to that used by Antonarakis and colleagues [3], to assess the prognostic and predictive significance of CTC, AR-FL, and AR-V7 in a monocentric prospective cohort of patients with mCRPC at different phases of disease.
In our exploratory cohort, we observed that CTC, AR-FL, and AR-V7 were all enriched in pretreated mCRPC patients compared to those starting a first-line regimen (Table 1). This figure is well known in respect to AR-V7 [5,13], but less information is currently available regarding AR-FL in CTC [20].
Overall, a lower percentage of biochemical responses was observed in the AR-V7+ patients than in AR-V7− (Figures 1 and 2). Neither of the two AR-V7+ patients treated with ARSi showed a biochemical response. Notably, a higher percentage of biochemical responses were also observed in CTC− compared to CTC+ patients and in AR-FL-compared to AR-FL+ patients. Despite our results potentially being biased due to the small sample size and heterogeneity of the cohort, these observations are consistent with the literature data [3][4][5][6][7][8][9][10][11][12][13][14][15]. Among patients with >10 copies/mL of AR-FL, 2/2 treated with taxane reached the PSA50 compared to 1/3 treated with ARSi. Although few patients are considered, this observation might be consistent with a possible positive effect of taxanes in patients with AR gain [19].

Discussion
In the present study, we used a modified-AdnaTest CTC-based AR-FL and AR-V7 mRNA assay with custom primers [22], similar to that used by Antonarakis and colleagues [3], to assess the prognostic and predictive significance of CTC, AR-FL, and AR-V7 in a monocentric prospective cohort of patients with mCRPC at different phases of disease.
In our exploratory cohort, we observed that CTC, AR-FL, and AR-V7 were all enriched in pretreated mCRPC patients compared to those starting a first-line regimen (Table 1). This figure is well known in respect to AR-V7 [5,13], but less information is currently available regarding AR-FL in CTC [20].
Overall, a lower percentage of biochemical responses was observed in the AR-V7+ patients than in AR-V7− (Figures 1 and 2). Neither of the two AR-V7+ patients treated with ARSi showed a biochemical response. Notably, a higher percentage of biochemical responses were also observed in CTC− compared to CTC+ patients and in AR-FL-compared to AR-FL+ patients. Despite our results potentially being biased due to the small sample size and heterogeneity of the cohort, these observations are consistent with the literature data [3][4][5][6][7][8][9][10][11][12][13][14][15]. Among patients with >10 copies/mL of AR-FL, 2/2 treated with taxane reached the PSA50 compared to 1/3 treated with ARSi. Although few patients are considered, this observation might be consistent with a possible positive effect of taxanes in patients with AR gain [19].
To our knowledge, this is the first report that included a group of highly pretreated patients receiving palliative therapies. This population well represents a condition of intensive selective pressure from treatments (ARSi and chemotherapy). Of the four patients treated with cyclophosphamide, mitoxantrone, and vinorelbine, none showed a PSA response (all were CTC positive and two were AR-V7 positive). This observation suggests that such treatments might not able to overcome the resistance induced by prior therapies and confirms the strong enrichment of AR-V7 and CTC in end-stage populations.
AR-V7, AR-FL, and CTC were all identified as prognostic variables for OS in univariable analysis (Supplementary Table S2), but AR-V7 was the only factor that retained the statistical significance in the Cox multivariable model (Table 2).
In our exploratory prognostic nomograms, we did not find that the combination of AR-V7 with CTC or with AR-FL could better predict for patients' prognosis compared to AR-V7 alone (Figures 3 and 4). This result needs confirmation in a larger cohort and might be the result of a small sample size.
We also acknowledge that the significance of CTC, AR-FL, and AR-V7 in our prospective series might be at least partly amplified by the method for CTC detection and heterogeneity of the cohort. Notably, AR-V7 correlated with both AR-FL and CTC detection. There is a higher probability of detecting CTC, and in turn AR-FL/AR-V7, in a population with more advanced and incurable disease stage, because of the higher disease burden and higher number of CTC. This figure is confirmed by the highest PSA values that were observed in CTC+ AR-V7+ patients, followed by CTC+ AR-V7− and by CTC− AR-V7− patients (Table 1). Similarly, a negative AR-V7 status does not necessarily imply that AR-V7 is not playing a role in the resistance processes. In fact, the detection of AR-V7 is subordinated to the challenging detection of CTC, which might be rare in patients with low disease burden or low tumor shrinkage. This notion is confirmed by a recent study performed by Sharp and colleagues [13]. The authors analyzed 181 mCRPC patients and did not find a significant difference in OS between CTC+/AR-V7+ and CTC+/AR-V7−patients when adjusting for baseline characteristic including CTC count. They also demonstrated that AR-V7 protein expression in mCRPC biopsies is often not consistent with the result on CTC. False positives and false negatives can derive from the limits of CTC isolation and AR-V7 analysis or can be due to intrapatient tumor sampling variability. Finally, although AR-V7 seems to predict for response to ARSi, this androgen-receptor variant does not explain all of the mechanisms of resistance, since some AR-V7+ patients still respond to ARSi therapy, and some AR-V7+ patients do not respond to chemotherapy [26]. This observation is confirmed in our small cohort of patients, in which three AR-V7 negative patients did not respond to first-line ARSi, and many AR-V7+ patients did not show response to chemotherapy.

Conclusions
Our study confirms the prognostic role of AR-V7, whose detection might be relevant during the treatment choices for patients with mCRPC. AR-V7 seems to be the most promising biomarker to predict for response to ARSi, however a prospective validation study in chemotherapy-treated versus ARSi-treated patients is still needed to implement this test in clinical practice. In our small cohort, the combination of AR-V7 with CTC or with AR-FL did not outperform the prognostic model including only AR-V7. Further studies are strongly warranted to assess the role of these biomarkers in patients with mCRPC.
Author Contributions: F.B. conceived and designed the study. F.B., C.C., L.C. and E.Z. collected patients' data and blood samples. P.B. and M.C. analyzed the blood samples for CTC, AR-FL and AR-V7. A.R. and L.Z. performed the statistical analyses. F.B. and C.C. were major contributors in writing the manuscript. All authors read and approved the final manuscript.