Clinical Significance of AR-V567es in Prostate Cancer—Response

We thank Drs. Plymate and Uo for their assessment of the article by Bernemann and colleagues (1) and for sharing their concerns about the validity of the AR-v567es assay used in our study (2).

The intent of Bernemann and colleagues was to compare six published assays for the detection of AR-v567es mRNA, together with their own in-house assay, using the same input RNA. While the intent is important, they only compared the primers/probes used by other groups, and did not adopt the complete assay platforms utilized in the original studies. For example, while Liu and colleagues (3) used a Taq Man approach with 40 amplification cycles and whole blood mRNA as input, Bernemann and colleagues used 45 cycles for the same assay and EpCAM-selected circulating tumor cells (CTC) as input.

With respect to our assay, different conditions, including RNA input, assay platform, and CTC enrichment method, were utilized for the comparison. Bernemann and colleagues used a Taq Man assay with RNA reverse transcribed to cDNA as input followed by 45 cycles of amplification, while we used a highly sensitive digital droplet PCR (ddPCR) platform, with template RNA as input, which has been reported to enhance the specificity and sensitivity of RNA target quantitation even in comparison with conventional ddPCR methods (4). In addition, our CTC enrichment platform involved PSMA (Prostate Specific Membrane Antigen)-based immunocapture that likely enriches for a different CTC subpopulation compared with the EpCAM-based approach. Use of different CTC enrichment methodologies adds another important variable that could account for some of the discrepancies.

The main concern raised by Bernemann and colleagues was that our assay may not specifically detect AR-v567es, due to potential nonspecific AR-FL amplification in the absence of AR-v567es (1). While this claim is theoretically plausible, it is not supported by our data showing no signal for AR-v567es transcript in cells expressing endogenous AR-FL alone (Supplementary Fig. S4 in Tagawa and colleagues; ref. 2). Accordingly, our study showed no AR-v567es signal detection in 8 of 46 patient samples expressing AR-FL (Supplementary Table S2 in Tagawa and colleagues; ref. 2).

In addition, a head-to-head comparison of the two assays is challenging, as there is no information on the analytic specificity and sensitivity of the Bernemann assay, such as inclusion of positive and negative controls to benchmark the assay or detection limit threshold. For example, our assay reliably and reproducibly detected the intended transcript in single cells (Supplementary Fig. S1 in Tagawa and colleagues; ref. 2), whereas their assay did not detect AR-v567es transcript in the established 136 xenograft model at passage 8 (Fig. 10; Supplementary Fig. S2 in Bernemann and colleagues; ref. 1) originally shown to express robust AR-v567es mRNA and protein in passages 5–10 (5). Differences in assay sensitivity is another important factor that should be taken into account in assay comparison.

Finally, to properly address discrepancies in the reported AR-v567es prevalence, we should take into account the important clinical distinctions such as prior treatment with abiraterone and/or enzalutamide (or any other potent AR pathway inhibitor) versus treatment-naïve mCRPC, versus taxane chemotherapy-pretreated metastatic castration-resistant prostate cancer. Each of these prior treatments might select for CTCs (or tissue) clones with potentially distinct AR-variant expression.

Taken together, this letter's affirmation that Bernemann and colleagues performed a comprehensive comparison of previously published assays, while passing “…the rigorous standardized checklist” seems to represent an overstatement. In addition, the comparison between untargeted, whole-transcriptome RNA-Seq data and targeted RNA quantitative assays is troublesome as it is already established that targeted qRT-PCR–based assay are significantly more sensitive for specific detection of known transcripts.

In closing, we agree with the authors of the letter that the exact role and clinical significance of AR-v567es in mCRPC remains uncertain, and that a more comprehensive proteogenomic approach is required before we conclude on its clinical utility as a prognostic or predictive biomarker. We would also remind the authors that such biomarker interrogations can only occur after rigorous analytic validation of the assay methodology, platform, and experimental conditions used; and that clinical qualification requires testing of the assay in a uniform patient population within a specific context of use. We plan to conduct such studies before concluding on the clinical relevance and utility of our AR-v567es assay.

See the original Letter to the Editor, p. 6009