A clinical-grade liquid biomarker detects neuroendocrine differentiation in prostate cancer

Abstract

BackgroundNeuroendocrine prostate cancer (NEPC) is an aggressive subtype, the presence of which changes the prognosis and management of metastatic prostate cancer.MethodsWe performed analytical validation of a Circulating Tumor Cell (CTC) multiplex RNA qPCR assay to identify the limit of quantification (LOQ) in cell lines, synthetic cDNA, and patient samples. We next profiled 116 longitudinal samples from a prospectively collected institutional cohort of 17 patients with metastatic prostate cancer (7 NEPC, 10 adenocarcinoma) as well as 265 samples from 139 patients enrolled in 3 adenocarcinoma phase II trials of androgen receptor signaling inhibitors (ARSIs). We assessed a NEPC liquid biomarker via the presence of neuroendocrine markers and the absence of androgen receptor (AR) target genes.ResultsUsing the analytical validation LOQ, liquid biomarker NEPC detection in the longitudinal cohort had a per-sample sensitivity of 51.35% and a specificity of 91.14%. However, when we incorporated the serial information from multiple liquid biopsies per patient, a unique aspect of this study, the per-patient predictions were 100% accurate, with a receiver-operating-curve (ROC) AUC of 1. In the adenocarcinoma ARSI trials, the presence of neuroendocrine markers, even while AR target gene expression was retained, was a strong negative prognostic factor.ConclusionOur analytically validated CTC biomarker can detect NEPC with high diagnostic accuracy when leveraging serial samples that are only feasible using liquid biopsies. Patients with expression of NE genes while retaining AR-target gene expression may indicate the transition to neuroendocrine differentiation, with clinical characteristics consistent with this phenotype.FundingNIH (DP2 OD030734, 1UH2CA260389, R01CA247479, and P30 CA014520), Department of Defense (PC190039 and PC200334), and Prostate Cancer Foundation (Movember Foundation - PCF Challenge Award).

Keywords: Oncology; Prostate cancer.

Figure 1. Multiplex qPCR in 22RV1 cells can detect RNA down to below a single cell.

22RV1 cell number is plotted against qPCR Cq (n = 47 for each gene). Four genes, SYP, CHGA, AR-V7, and AR-V9, were selected for their variable expression in CTCs from patients with metastatic prostate cancer. qPCR is able to reliably detect RNA as low as an equivalent of 0.4 cells; above this, the relationship between Cq and input amount appears to be linear.

Figure 2. Limits of quantification for CTC multiplex qPCR.

(A) Logistic regression demonstrating the limits of quantification using serial dilutions of 22RV1 cells (n = 145) including SYP, CHGA, AR-V7, and AR-V9. Five replicates were performed per gene and concentration. (B) Logistic regression demonstrating the limits of quantification using serial dilutions of synthetic DNA oligonucleotides (n = 324), including the genes SYP, CHGA, AR-V7, AR-V9, TMPRSS2, KLK3, KLK2, and FOLH1. Nine replicates were performed per gene and concentration. (C) Logistic regression demonstrating the limits of quantification using CTCs captured from patients with prostate cancer (n = 148). The same 4 genes were tested as in A. 4 replicates were performed per gene and sample. In addition, the ubiquitously expressed housekeeping gene RPII was added (shown in gray) for visualization purposes only because the range of relative quantity (RQ) values for the 4 genes was narrow, as expression levels were low even when detectable (shown in black). Addition of RPII allowed the visualization of the plateau region.

Figure 3. NEPC liquid biomarker performance.

(A) Waffle plot shows the accuracy of NEPC classification on a per-sample basis across our institutional longitudinal samples. (B) Bar plot shows the performance metrics of NEPC classification on an individual per-sample basis versus leveraging serial sampling using liquid biomarkers to calculate the percent positive samples per patient (cutoff of 33%) from across our institutional longitudinal samples.

Figure 4. NEPC liquid biomarker serial sample performance.

(A) ROC curve of NEPC classification accuracy leveraging serial sampling using liquid biomarkers to calculate the percent positive samples per patient across our institutional longitudinal samples. (B) Waffle plot shows the accuracy of NEPC classification on a per-patient basis using the percent positive across our institutional longitudinal samples.

Figure 5. Clinical outcomes from ARSI trials and emergence of neuroendocrine markers.

(A) Kaplan-Meier curves show that patients with expression of neuroendocrine markers, even with preserved AR target gene expression, have worse overall survival (OS) in 2 phase II adenocarcinoma ARSI (enzalutamide and abiraterone) clinical trials. P is calculated via log-rank test. (B) Kaplan-Meier curves show that patients with expression of neuroendocrine markers, even with preserved AR target gene expression, have worse time to treatment failure (TTF) in a phase II adenocarcinoma ARSI (seviteronel) clinical trial. P values were calculated via log-rank test.