EURURO Vol. 72 No. 5

830 860

RNA was extracted from blood samples according to the manufacturer’s
instructions using the PAXgene blood RNA kit (Qiagen). A NanoDrop
1000 spectrophotometer (Thermo Scientific, Waltham, MA, USA) was
used for quantification and purity assessment of RNA samples. cDNAwas
synthesized from 1 m g of RNA using qScript XLT SuperMix (Quanta
Biosciences, Beverly, MA, USA) according to the manufacturer’s
instructions.
2.3. ddPCR analysis
AR-V7 and AR-FL mRNA levels were simultaneously quantified in a dual
color assay using custom primer and hydrolysis probe sets on a QX200
ddPCR systemwith automatic droplet generation (Bio-Rad Laboratories,
Hercules, CA, USA). Analyses were performed and reported according to
the digital MIQE guidelines . Further information on the ddPCR
assays is provided in the Supplementary material. All operators involved
in the measurements were blinded to the assignment of samples to
healthy control subjects or patients and their outcome.
2.4. Healthy donors
We included 28 healthy male subjects (age < 40 yr) to determine
background levels of AR-V7 and AR-FL transcripts in peripheral whole
blood. Samples from healthy subjects were obtained and stored under
the same conditions as for patient samples to minimize any bias.
2.5. Statistical analysis
The Supplementary material contains details on the statistical analysis.
3. Results
3.1. Patient characteristics
We enrolled 85 mCRPC patients who were scheduled to
undergo a new line of therapy with either abiraterone
(n = 56) or enzalutamide (n = 29). provides detailed
information on baseline characteristics and clinical out-
comes. Bone and visceral metastases were present in 96%
and 28% of patients, respectively. Prior systemic treatment
regimens for mCRPC included chemotherapy with doce-
taxel in 79% of patients and abiraterone in 28% of patients.
None had previously received enzalutamide.
3.2. AR-V7 detection in peripheral whole blood
First we assessed the analytical validity of our ddPCR assay
for AR-V7 and AR-FL isoform detection. AR-V7 mRNA from
one VCaP cell against a background of 1 million leukocytes
could be repeatedly detected (Supplementary Fig. 1). Next
we quantified AR-V7 and AR-FL transcript levels in
peripheral whole blood samples from 85 mCRPC patients
and 28 healthy men as control subjects ). Notably,
18 of 28 healthy control subjects had detectable (non-zero)
AR-V7 levels. To normalize AR-V7 expression, we calculated
the fraction of AR-V7 transcripts over total AR (AR-V7 plus
AR-FL) transcripts, and used this ratio in all subsequent
analyses. The fraction of AR-V7 transcripts inwhole blood of
mCRPC patients ranged from0% to 4.0% (mean 0.3%; ).
Using the maximum AR-V7 fraction observed among
healthy men (0.6%) as a cutoff, we dichotomized patients
into “AR-V7 high” and “AR-V7 low” groups ). Overall,
15/85 patients (18%) had high AR-V7 levels. According to
prior therapy with zero, one, two, and three lines of
systemic treatment, the number of patients with high AR-
V7 levels was 0/9 (0%), 9/39 (23%), 3/27 (11%), and
3/10 (30%), respectively.
3.3. AR-V7 status in whole blood predicts PSA response under
abiraterone or enzalutamide
The overall proportion of patients with a PSA response
defined as a PSA decline of 50% was 41% (31 of 74 menwith
available PSA follow-up). The PSA response rate was 0% in
patients with high AR-V7 levels (0 of 12 men), and 50% in
patients with low AR-V7 levels (31 of 62 men, ). Thus,
AR-V7 status was significantly associated with PSA response
in univariable analysis (p < 0.001). In a multivariable
logistic regression analysis we modeled the influence on
PSA response of AR-V7 status together with clinical
variables . Only the association between AR-V7
status and PSA response remained significant, and high
AR-V7 levels in whole blood were confirmed as an
independent predictor of no PSA response to treatment
with abiraterone or enzalutamide (p = 0.03). Furthermore,
AR-V7 was confirmed as an independent predictor of no PSA
response in a multivariable model with AR-V7 expression as
a continuous variable (p = 0.04; Supplementary Table 1).
Table 1 – Patient characteristics and clinical outcomes
Parameter Value
Patients (n) 85
Median age, yr (IQR) [ n = 85] 71 (66–74)
Median PSA, ng/ml (IQR) [n = 84] 211 (29–768)
Eastern Cooperative Oncology Group performance score, n (%) [n = 83]
0 43 (52)
1 29 (35)
2 11 (13)
Prior systemic treatments for mCRPC, n (%) [n = 85]
Docetaxel 67 (79)
Abiraterone 24 (28)
Cabazitaxel 14 (17)
Enzalutamide 0 (0)
Radium-223 6 (7)
Other 12 (14)
Prior lines of systemic treatment regimens for mCRPC (n) [n = 85]
0 9
1 39
2 27
3 10
Site of metastasis, n (%) [n = 83]
Bone 80 (96)
Visceral 23 (28)
Deceased, n (%) [ n = 84] 51 (60)
Median follow-up, mo (IQR) [n = 84] 7.6 (4.7–12.7)
With event (death) 7.3 (3.3–12.7)
Without event (death) 7.7 (5.4–12.6)
Median PSA-PFS, mo (95% CI) [n = 74] 3.6 (3.2–4.1)
Median clinical PFS, mo (95% CI) [ n = 82] 4.6 (3.1–6.2)
Median overall survival, mo (95% CI) [n = 84] 10.1 (5.8–14.5)
IQR = interquartile range; CI = con fidence interval; mCRPC = metastatic
castration-resistant prostate cancer; PSA = prostate-specific antigen;
PFS = progression-free survival.
E U R O P E A N U R O L O GY 7 2 ( 2 0 17 ) 8 2 8 – 8 3 4 830