Combined Cell-free DNA and RNA Profiling of the Androgen Receptor: Clinical Utility of a Novel Multianalyte Liquid Biopsy Assay for Metastatic Prostate Cancer

Abstract

Background:

The androgen receptor (AR) remains a critical driver in metastatic castration-resistant prostate cancer (mCRPC). Profiling AR aberrations in both circulating DNA and RNA may identify key predictive and/or prognostic biomarkers in the context of contemporary systemic therapy.

Objective:

To profile AR aberrations in circulating nucleic acids and correlate with clinical outcomes.

Design, setting, and participants:

We prospectively enrolled 67 mCRPC patients commencing AR pathway inhibitors (ARPIs; n = 41) or taxane chemotherapy (n = 26). Using a first-in-class next-generation sequencing-based assay, we performed integrated cell-free DNA (cfDNA) and cell-free RNA (cfRNA) profiling from a single 10 ml blood tube.

Outcome measurements and statistical analysis:

Kaplan-Meier survival estimates and multivariable Cox regression analyses were used to assess associations between clinical outcomes and the following AR aberrations: copy number variation, splice variants (ARV7 and AR-V9) and somatic mutations.

Results and limitations:

Cell-free DNA and cfRNA were successfully sequenced in 67 (100%) and 59 (88%) patients, respectively. Thirty-six (54%) patients had one or more AR aberrations. AR gain and cumulative number of AR aberrations were independently associated with clinical/radiographic progression-free survival (PFS; hazard ratio [HR] 3.2, p = 0.01 and_, HR 3.0 for, 0 vs ≥2, p = 0.04) and overall survival (HR 2.8, p = 0.04 and HR 2.9 for 0 vs ≥2 p=0.03) Notably,concurrent AR gain and AR splice variant expression (AR gain/AR-V+) was associated with shorter prostate-specific antigen PFS on both ARPIs (HR 6.7, p = 0.009) and chemotherapy (HR 3.9, p = 0.04). Importantly, key findings were validated in an independent cohort of mCRPC patients (n = 40), including shorter OS in AR gain/AR-V+ disease (HR 3.3, p = 0.02). Limitations include sample size and follow-up period.

Conclusions:

We demonstrate the utility of a novel, multianalyte liquid biopsy assay capable of simultaneously detecting AR alterations in cfDNA and cfRNA. Concurrent profiling of cfDNA and cfRNA may provide vital insights into disease biology and resistance mechanisms in mCRPC.

Patient summary:

In this study of men with advanced prostate cancer, DNA and RNA abnormalities in the androgen receptor detected in blood were associated with poor outcomes on available drug treatments. This information could be used to better guide treatment of advanced prostate cancer.

1. Introduction

Advances in therapeutic strategies have significantly improved both quantity and quality of life for metastatic castration-resistant prostate cancer (mCRPC) [1]. However, there remains a pressing need to identify predictive and prognostic biomarkers for contemporary systemic therapies. Identification of such biomarkers may hold great utility, informing optimal treatment selection, design of innovative clinical trials, and facilitating discussions with patients and caregivers around expected outcomes.

In the era of precision medicine, liquid biopsies have emerged as a minimally invasive alternative for interrogating the prostate tumour genome, either through the capture of intact circulating tumour cells (CTCs) or through the analysis of cell-free tumour nucleic acids [2]. Circulating assays have demonstrated strong congruence with tumour biopsies [3], whilst simultaneously encapsulating the complexity of intrapatient heterogeneity evident within the mCRPC genomic landscape.

As sustained androgen receptor (AR) signalling remains a key driver in mCRPC [4], considerable efforts have been made to profile AR aberrations using circulating nucleic acids. Resistance to novel AR pathway inhibitors (ARPIs; eg, abiraterone and enzalutamide) has been observed in patients harbouring AR copy number gain [5–9], somatic AR mutations [6–8], and constitutively active AR splice variants (AR-Vs) [10,11]. Similarly, AR copy number gain has been associated with poor outcomes in patients receiving chemotherapy [12]. Importantly, nearly all assays have principally focused on genomic aberrations, which likely explain only a proportion of the phenotypic responses observed clinically. Transcriptomic analysis has been proposed as the bridge between genotype and phenotype where liquid biopsies are concerned [13], with obvious advantages to combining DNA and RNA sequencing data.

De Laere et al [14,15] recently combined sequencing of the AR in plasma cell-free DNA (cfDNA) with RNA sequencing of AR-Vs expressed in CTCs. However, current isolation methods are unable to consistently detect CTCs in a high proportion of patients [10,16,17], highlighting the ongoing need to develop and optimise CTC-independent platforms to detect AR-Vs as well as other transcriptomic resistance markers. Importantly, aberrations found specifically in cell-free RNA (cfRNA) have not been correlated previously with clinical outcomes in mCRPC patients receiving contemporary systemic treatments.

Here, we use a novel, high-sensitivity, multianalyte, next-generation sequencing assay to simultaneously profile the AR in pretreatment plasma cfDNA and cfRNA obtained from two independent cohorts of mCRPC patients. We found that concurrent DNA and RNA AR aberrations portend poor treatment outcomes, likely reflecting patients with aggressive intrinsic disease biology and resistance to existing therapies. Our data suggest that combined cfDNA and cfRNA sequencing may have high clinical value in mCRPC.

2. Patients and methods

2.1. Study cohorts and sample processing

Patients with mCRPC were prospectively enrolled between September 2016 and August 2018 across two Australian institutions (Monash Health and Chris O’Brien Lifehouse). All patients provided written informed consent, with ethics approval obtained from each institution’s human research ethics committee. Peripheral blood (10 ml) was collected in a single EDTA-containing or dedicated cfDNA-stabilising tube (Streck, La Vista, Nebraska, USA) immediately prior to commencing systemic therapy (ARPIs or taxane chemotherapy). Two-step centrifugation was performed (1900× g for 10 min followed by 16 000 ×g for 10 min) to separate and clarify plasma and buffy coat (containing peripheral blood mononuclear cells [PBMCs]). Plasma and PBMCs were stored at –80^∘C until required.

A second independent group of mCRPC patients served as a validation cohort. Patients were prospectively enrolled between September 2009 and March 2014 to a circulating biomarker programme at a single tertiary-level US cancer centre (Mayo Clinic). Details of ethics approval, sample collection, and processing have been published previously [18].

2.2. Nucleic acid extraction, next-generation sequencing, and bioinformatics analysis

A detailed description of nucleic acid extraction, sequencing, and bioinformatics analysis is provided in the Supplementary material. Briefly, PBMC-derived germline DNA (gDNA) and plasma cfDNA/cfRNA were extracted using a combination of established proprietary kits and in-house column-based methods. After quality assessment and quantification, 40 ng of fragmented gDNA, 5–30 ng of cfDNA, and up to 30 ng of fragmented cfRNA were used for library preparation, panel-based hybridisation (180-gene PredicinePLUS panel; Supplementary Table 1), and enrichment prior to paired-end sequencing on the Illumina HiSeq XTen. Predicine’s proprietary GeneRADAR technology and DeepSea machine learning bioinformatics algorithm were then used to identify point mutations, insertions/deletions, splice site alterations, and copy number alterations. Circulating tumour DNA (ctDNA) fraction was determined using a methodology described previously [19].

2.3. Outcome measures and statistical methods

Follow-up time was calculated from the date of sample acquisition to the date of last patient contact. AR aberrations were defined as AR copy number variation (ctDNA), AR somatic mutations (ctDNA), and AR-Vs (cfRNA), which were restricted to AR-V7 and AR-V9 due to their strong association with pathogenicity [10,20,21]. Kaplan-Meier survival estimates (log-rank test) and multivariable Cox regression models (covariates: ctDNA fraction dichotomised into below or above 2% [22] (see Supplementary material); prior taxane chemotherapy; prior ARPIs; performance status; presence of visceral metastases; and pain at enrolment) were then used to assess the association between AR aberrations and clinical outcomes, including (1) overall survival (OS; time from treatment commencement until death from any cause), (2) prostate-specific antigen (PSA) response (PSA decline from baseline of ≥50%, confirmed ≥3 wk later), (3) PSA progression-free survival (PSA-PFS, as per Prostate Cancer Working Group 3 criteria [23]), and (4) clinical/radiographic progression-free survival (clin/rPFS). Evaluation of PSA response required≥12 wk of follow-up, and the 12-wk PSA response rate is reported. All three survival outcomes were censored at the date of last patient contact if the event had not occurred. Statistical significance was defined as p < 0.05.

3. Results

3.1. Assay validation

Multiple steps were taken for analytical and orthogonal validation of the assay (Supplementary material). Serially diluted reference DNA was used to determine an assay limit of detection of 0.25%, or 0.1% for hotspot mutations. For AR-V7 detection, sensitivity was found to be 100% for samples with a minimum of 10 AR-V7 molecules, whilst no AR-Vs were detected in healthy controls. Regarding orthogonal validation, AR gains were validated using low-pass whole-genome sequencing. Similarly, AR-V7 detection in cfRNA was previously validated using digital droplet polymerase chain reaction.

3.2. Study cohort

For the Australian cohort, pretreatment blood samples were obtained from 67 patients immediately prior to commencing ARPIs (n = 41) or taxane chemotherapy (n = 26). In total, eight (12%) failed cfRNA quality assessment and were consequently censored for AR-V analysis; all cfDNA samples passed quality assessment (Supplementary material). The clinical characteristics of the cohort and prior therapy exposure are presented in Table 1 and Supplementary Table 2, respectively. Median follow-up time for nondeceased patients was 14.7 mo.

Table 1:

Baseline patient characteristics of the Australian cohort

	Study cohort n = 67
TREATMENT ALLOCATION, N (%)
Enzalutamide	33 (49)
Abiraterone	8 (12)
Docetaxel	20 (30)
Cabazitaxel	6 (9)
AGE
Q1	63
Q2	73
Q3	78
Q4	91
RACE, N (%)
White	63 (94)
Asian	2 (3)
Other	2 (3)
EXTENT OF DISEASE, N (%)
Bone	64 (96)
Nodal	37 (55)
Visceral	11 (16)
GLEASON GRADE GROUP, N (%)
Grade Group 1–3 (Gleason ≤ 7)	16 (24)
Grade Group 4–5 (Gleason ≥ 8)	31 (46)
No biopsy/unknown	20 (30)
PRIMARY TREATMENT, N (%)
Surgery	19 (28)
Radiation +/− ADT	11 (16)
Primary ADT	3 (5)
Metastatic disease at diagnosis	34 (51)
PRIOR TREATMENT EXPOSURE, N (%)
Prior chemotherapy only	15 (22)
Prior ARPI only	6 (9)
Prior chemotherapy and ARPI	12 (18)
LINE OF SYSTEMIC TREATMENT AT SAMPLE COLLECTION, N (%)
First line	34 (51)
Second line	21 (31)
Third line	10 (15)
Fourth line	2 (3)
PAIN AT BASELINE, N (%)
No	29 (43)
Yes	38 (57)
BASELINE PSA (ng/ml)
Median	51

ADT = androgen deprivation therapy; ARPI = androgen receptor pathway inhibitor; PSA = prostate-specific antigen.

3.3. Distribution of AR aberrations

AR aberrations of any type were present in 36/67 (54%) patients at baseline; the distribution of AR aberrations is shown in Figure 1. Quality metrics for these samples are provided in Supplementary Tables 3–6. AR copy number gain was found in 26/67 (39%) patients; of note, ctDNA fraction was not significantly higher in patients with AR copy number gain (Supplementary Fig. 1). AR somatic mutations were seen in 16/67 (24%) patients. Supplementary Table 7 summarises the landscape of AR mutations, the most common of which was H875Y (n = 10). The median allelic frequency of AR mutations was 2.1% (range 0.13–26%). Eleven (19%) and six (10%) patients harboured transcript expression of AR-V7 and AR-V9, respectively. As no patients exhibited all three types of AR aberrations, associations between cumulative AR aberrations and time-to-event outcomes were analysed using a three-level model (zero/ one/two or more aberrations).

Fig. 1 –

Landscape of AR aberrations identified in cell-free DNA and RNA across the Australian cohort. Asterisks denote ARPI therapy, whilst daggers denote taxane chemotherapy. Prior therapy at baseline is also shown. Orange tiles represent presence of aberration; missing cfRNA data are denoted by grey tiles.

AR = androgen receptor; ARPI = AR pathway inhibitor; AR-V = AR splice variant; cfRNA = cell-free RNA.

3.4. AR aberrations and biochemical outcomes

PSA responses were seen in 42/67 (63%) patients, with median PSA-PFS of 7.7 mo. Supplementary Table 8 summarises PSA response rates according to the type of AR aberration. The presence of AR copy number gain was associated with significantly lower PSA response rates (neutral/gain: 28/41 [68%] vs 9/26 [35%], p = 0.007), although this difference was observed only in ARPI-treated patients (Supplementary Tables 9 and 10). Supplementary Fig. 2 shows the waterfall plot of best PSA response according to AR copy number status. Conversely, the presence of an AR mutation correlated with higher PSA response rates (wild-type/mutated: 26/51 [51%] vs 11/16 [69%], p = 0.2); Supplementary Fig. 3 shows the waterfall plot of the best PSA response according to AR mutation status. Neither AR-V7 nor AR-V9 transcript expression predicted for a confirmed ≥50% PSA response.

Figure 2 demonstrates Kaplan-Meier estimates of PSA-PFS according to AR copy number variation, presence of any AR aberration, and total number of AR aberrations, all of which were associated with significantly shorter PSA-PFS in univariable (Table 2) and multivariable (Table 3) analyses. Furthermore, patients with simultaneous AR gain and AR-V expression (n = 9) experienced particularly poor outcomes in univariable analysis (Fig. 3 and Table 2; median PSA-PFS 8.1 vs 3.4 mo, p < 0.001), remaining significant in multivariable analysis (hazard ratio [HR] 4.2, 95% confidence interval [CI] 1.1–15, p = 0.03; Table 3). A summary of the univariable associations between AR aberrations and PSA-PFS analysed by treatment subgroup is shown in Supplementary Tables 11 and 12. Importantly, for patients with concurrent AR gain and AR-V expression, PSA-PFS was significantly shorter following both ARPIs (HR 6.7, 95% CI 1.6–29, p = 0.009) and chemotherapy (HR 3.9, 95% CI 1.1–15, p = 0.04), with median PSA-PFS of 1.8 and 4.3 mo, respectively.

Fig. 2 –

Kaplan-Meier analysis of PSA-PFS (top), clinical or radiographic PFS (middle), and overall survival (bottom), according to AR copy number status (left), the presence of at least one of AR gain, AR splice variant, or AR somatic mutation (centre), and the total number of AR aberrations (0, 1,≥2) present (right). Two and eight patients were excluded from analysis for any AR aberration and total AR aberrations, respectively, due to insufficient information on AR-V expression.

AR = androgen receptor; AR-V = AR splice variant; Clin/rPFS = clinical/radiographic progression-free survival; OS = overall survival; PFS = progression-free survival; PSA = prostate-specific antigen.

Table 2:

Univariable Cox proportional hazards analysis of clinical endpoints based on AR aberrations in two independent cohorts

			PSA progression-free survival			Clin/r progression-free survival			Overall survival
	Variable	n	HR	95% CI	p	HR	95% CI	p	HR	95% CI	p
Australian cohort	AR gain	26	2.9	1.5–5.7	0.002	3.1	1.5–6.5	0.003	3.0	1.4–6.5	0.006
	AR mutation	16	0.80	0.44–1.9	0.8	0.69	0.29–1.6	0.4	1.0	0.41–2.3	0.9
	AR-V+ (V7 or V9)	13	2.6	1.2–5.6	0.02	2.8	1.3–6.4	0.01	2.1	0.91–5.1	0.08
	AR-V7+	11	2.1	0.97–4.9	0.07	2.4	1.1–5.6	0.04	2.2	0.84–5.7	0.1
	AR-V9+	6	3.1	1.2–8.6	0.03	3.0	1.1–8.3	0.03	2.2	0.74–6.5	0.2
	Any AR aberration	36	2.5	1.3–5.2	0.01	2.6	1.2–5.6	0.02	2.6	1.1–6.1	0.03
	AR-V+ plus AR gain	9	4.6	1.8–12	0.001	4.0	1.6–10	0.003	2.3	0.92–5.9	0.07
	AR total aberrations 0	29	1.0	-	-	1.0	-	-	1.0	-	-
	1	12	2.5	1.0–6.4	0.048	2.7	1.0–7.0	0.043	2.0	0.62–6.4	0.3
	≥2	18	2.8	1.2–6.4	0.02	3.1	1.3–7.8	0.01	3.5	1.4–9.2	0.01
Mayo cohort	AR gain	18	-	-	-	-	-	-	7.8	3.0–21	<0.001
	AR mutation	8	-	-	-	-	-	-	1.6	0.74–3.6	0.2
	AR-V+ (V7 or V9)	5	-	-	-	-	-	-	2.1	0.74–5.7	0.2
	AR-V7+	4	-	-	-	-	-	-	2.5	0.83–7.7	0.1
	AR-V9+	2	-	-	-	-	-	-	0.83	0.19–3.6	0.8
	Any AR aberration	21	-	-	-	-	-	-	3.3	1.3–8.9	0.02
	AR-V+ plus AR gain	5	-	-	-	-	-	-	3.3	1.2–8.9	0.02
	AR total aberrations 0	7	-	-	-	-	-	-	1.0	-	-
	1	9	-	-	-	-	-	-	8.2	1.7–40	0.009
	≥2	9	-	-	-	-	-	-	8.8	1.8–44	0.007

All p-values <0.05 are highlighted in bold.

AR = androgen receptor; AR-V = androgen receptor variant; Clin/r = clinical/radiographic; PSA = prostate-specific antigen.

Table 3:

Multivariable Cox proportional hazards analysis of clinical endpoints based on AR aberrations

		PSA progression-free survival			Clin/r progression-free survival			Overall survival
Variable	n	HR	95% CI	p	HR	95% CI	p	HR	95% CI	p
AR gain	26	3.3	1.5–7.6	0.004	3.2	1.3–8.0	0.01	2.8	1.1–7.2	0.04
AR mutation	16	-	-	-	-	-	-	-	-	-
AR-V+ (V7 or V9)	13	1.8	0.71–4.5	0.2	2.1	0.88–5.2	0.1	-	-	-
AR-V7+	11	-	-	-	2.2	0.86–5.8	0.1	-	-	-
AR-V9+	6	1.3	0.42–4.1	0.7	1.1	0.34–3.8	0.8	-	-	-
Any AR aberration	36	2.3	1.1–4.8	0.03	2.1	0.87–4.8	0.07	1.9	0.72–4.7	0.2
AR-V+ plus AR gain	9	4.2	1.1–15	0.03	2.8	0.85–9.3	0.09	-	-	-
AR total aberrations 0	29	1.0	-	-	1.0	-	-	1.0	-	-
1	12	1.6	0.58–4.3	0.36	1.5	0.51–4.5	0.5	1.6	0.58–4.3	0.36
≥2	18	2.9	1.1–7.3	0.03	3.0	1.1–8.3	0.04	2.9	1.1–7.3	0.03

NOTE: only AR aberrations with p < 0.05 in univariable analysis were included in multivariable analysis (MVA). All p-values < 0.05 in MVA are highlighted in bold.

Clinical variables included in MVA: ctDNA fraction ≥2%, prior chemotherapy, prior ARPI therapy, presence of visceral metastasis, presence of pain at baseline and ECOG PS >2.

AR = androgen receptor; ARPI = androgen receptor pathway inhibitor; AR-V = androgen receptor variant; Clin/r = clinical/radiographic; ctDNA = circulating tumour DNA; ECOG PS = Eastern Cooperative Oncology Group performance status; PSA = prostate-specific antigen.

Fig. 3 –

Kaplan-Meier analysis of (A) PSA-PFS, (B) clinical or radiographic PFS, and (C) overall survival, according to concurrent expression of both an AR-V and an AR copy number gain. Six patients were excluded from analysis due to insufficient information on AR-V expression.

AR = androgen receptor; AR-V = AR splice variant; Clin/rPFS = clinical/radiographic progression-free survival; OS = overall survival; PFS = progression-free survival; PSA = prostate-specific antigen.

3.5. AR aberrations and survival outcomes

The median clin/rPFS and OS for the overall cohort were 10.4 and 17.1 mo, respectively. Patients with any AR aberration, AR copy number gain, and cumulative AR aberrations experienced significantly shorter clin/rPFS and OS (Fig. 2), with the latter two variables remaining significant in multivariable analysis (Table 3). Similar to biochemical outcomes, presence of both AR gain and AR-V expression at baseline predicted poorer clin/rPFS (median 12 vs 3.8 mo, p = 0.001) and shorter OS (median 17.1 vs 10.4 mo, p = 0.07; Fig. 3 and Table 2). Univariable Cox regression analyses of clin/rPFS and OS for ARPI- and chemotherapy-treated subgroups are shown in Supplementary Tables 11 and 12. Notably, for ARPI-treated patients with concurrent AR gain and AR-V expression, clin/rPFS (HR 4.6, 95% CI 1.2–17, p = 0.03) and OS (HR 5.1, 95% CI 1.5–17, p = 0.008) were significantly shorter, with median clin/rPFS and OS of 2.1 and 11 mo, respectively. Interactions between treatment subgroup and AR aberrations were estimated using a nested Cox proportional hazard regression model for OS (Supplementary Table 13). AR gain, cumulative AR aberrations, and concurrent AR gain plus AR-V expression demonstrated a significant interaction with ARPI therapy.

3.6. Independent mCRPC cohort

Pretreatment samples were collected from 40 patients at Mayo Clinic. Baseline clinical characteristics are presented in Supplementary Table 14. The landscape of AR somatic mutations is shown in Supplementary Table 15. Quality metrics for these samples are provided in Supplementary Tables 3–6. Frequencies of AR gain, AR somatic mutations, and AR-V expression were 45%, 20%, and 20%, respectively (Supplementary Fig. 4), which were similar to the profile observed in the Australian cohort (Fig. 1).

Median OS in the Mayo Clinic cohort was 13.9 mo, with a median follow-up time for nondeceased patients of 79.6 mo. Importantly, similar to the Australian cohort, all patients with AR copy number gain, any AR aberration, and cumulative AR aberration experienced significantly shorter OS (Table 2). Notably, the concurrent presence of AR gain and AR-V expression was also associated with significantly worse OS (HR 3.3, 95% CI 1.2–8.9, p = 0.02; Supplementary Fig. 5).

4. Discussion

In this prospective multicentre study, we demonstrated the application of a novel, first-in-class liquid biopsy assay capable of simultaneously detecting both cfDNA and cfRNA aberrations in a single 10 ml blood tube from 67 mCRPC patients commencing contemporaneous systemic therapy. Using this assay to comprehensively profile both genomic and transcriptomic aberrations in the AR, we confirmed the ongoing critical impact that AR aberrations have in shaping the disease course in mCRPC. Importantly, our observations on the prognostic relevance of AR aberrations were confirmed in a second, independent cohort of mCRPC patients.

We observed that AR gain was an independent negative prognostic biomarker for OS and PFS. These findings are consistent with previous reports supporting an association between AR amplification and resistance to ARPI therapy [6–8]. However, we observed no association between AR point mutations and time-to-event outcomes, even with ligand promiscuity–conferring point mutations L702H, T878A, H875Y, and W742C [24]. We note data on whether these mutations confer resistance to novel ARPI therapies are conflicting [6,7,22,25]. Interestingly, we also observed an association between AR point mutations and increased PSA response rates. The underlying biology of this observation is still unclear, but has been noted previously [26], again illustrating the unclear relationship between AR single nucleotide variants and outcomes on ARPIs.

Other multianalyte liquid biopsy assays in mCRPC have compared and/or combined genomic and transcriptomic analyses of ctDNA together with CTCs [14,15,27–29]. However, the variability in CTC counts in mCRPC and RNA profiling of individual CTCs together yield inherent challenges, both from a technical perspective (paucity of viable CTCs) and from an analytical standpoint (noisier sequencing data [30], restricting low-frequency variant detection). Additionally, achievement of adequate analytical sensitivity remains a major obstacle, hindering liquid biopsy assay implementation into routine clinical care. By employing molecular barcoding to uniquely label individual nucleic acid fragments and perform error correction, complemented by deep sequencing of both plasma cfDNA and cfRNA, we were able to successfully identify ultra-low-frequency variants down to 0.1% with minimal input material.

Using this assay, we sought to expand our understanding of the relationship between AR genomic alterations and aberrant RNA splicing independent of CTC status. We found that possession of any type of AR aberration and the cumulative number of AR aberrations were independently associated with shorter PFS. In addition, the cumulative number of AR aberrations was independently associated with OS. These findings were not merely a consequence of an increased tumour burden, as the associations remained significant after adjustment for ctDNA fraction. Our data illustrate the potential impact of multianalyte assessment, with cfDNA and cfRNA providing complementary molecular insights into the global influence of the AR in mCRPC.

The impact of AR aberrations was unsurprisingly generally more apparent in the ARPI-treated cohort than in the chemotherapy-treated cohort. However, it is noteworthy that a subset of both ARPI- and chemotherapy-treated patients with concurrent AR gain and AR-V expression had particularly poor outcomes (median PSA-PFS 1.8 and 4.3 mo, respectively; HR 6.7 and 3.9, respectively). These results suggest that concurrent DNA and RNA aberrations in the AR may portend to exceedingly aggressive disease biology and inferior outcomes irrespective of treatment.

Importantly, our findings were largely confirmed in the independent cohort, with shorter OS linked to the presence of AR gain, any AR aberration, and cumulative AR aberrations. Critically, we also observed shorter OS in patients with both AR gain and AR-V expression, a trend that was also observed in the Australian cohort (p = 0.07), again highlighting the importance of concurrent DNA and RNA analysis.

We acknowledge several limitations of our study. Small sample size of both cohorts, together with heterogeneity of treatment administered, may impact generalisability of findings. Furthermore, follow-up periods were relatively short in the Australian cohort. In addition, due to low ctDNA fractions and poor RNA quality in some specimens, some false negative results may have been recorded.

5. Conclusions

We demonstrate the application of a first-in-class multianalyte liquid biopsy assay capable of simultaneously detecting AR genomic alterations in cfDNA and cfRNA with high sensitivity from a single blood tube. In two independent cohorts, we identified a novel poor prognosis subgroup harbouring concurrent AR gain and AR-V expression. These data support further evaluation of minimally invasive, multianalyte circulating nucleic assays in advanced prostate cancer patients, including integration into prospective clinical trials and cross-specimen comparison with CTCs and tumour tissue.