Androgen receptor splice variants are known to facilitate resistance of prostate cancer cells toward antihormonal therapies. However, detection of the most prominent variant, AR-V7, on its own, is not sufficiently accurate for prediction of response. Thus, simultaneous detection of other variants might improve prediction. AR-V567es has been shown to be expressed in late stages of prostate cancer. Yet, there have been discrepant results regarding incidence of AR-V567es. We therefore aimed to perform a comprehensive comparison of different detection approaches for AR-V567es mRNA.
We compared a custom-made, probe-based PCR assay with 6 published AR-V567es detection PCR assays in distinct samples, that is, cancer cell lines, LuCaP xenografts, primary and metastatic tumor samples, and circulating tumor cells (CTC).
Using distinct approaches, we concordantly detected expression of AR-V567es in only three of 45 samples (LuCaP xenografts 86.2 and 136s2 as well as one CTC sample). We observed varying results in all other samples. Specificity analysis displayed nonspecific binding of 5 previously published PCR assays to AR full-length mRNA in the absence of AR-V567es.
Validation of biomarker detection approaches is one of the most critical steps before transfer into clinical application. By performing comparative analysis of different detection approaches, we revealed eminent variability among previously described systems. Furthermore, we demonstrate an overestimation of AR-V567es in prostate cancer, presumably due to nonspecific detection of AR-FL mRNA. Therefore, any correlation between AR-V567es expression and clinical responses is highly doubtful and does not reflect the biological nature of the disease.
Detection of AR splice variants might one day become a clinical tool in stratifying patients with prostate cancer for either antihormonal or chemotherapeutic therapies. Besides other AR variants, which are being intensively discussed, AR-V567es detection might improve the predictive potential of these biomarkers in clinical settings. Nonetheless, different studies demonstrated varying percentages of patients with AR-V567es–positive prostate cancer. By performing comprehensive comparison of previously published detection approaches, we demonstrate considerable variation of detection of AR-V567es in tumor samples, which might be based on false-positive detection of AR-V567es, using several published systems. We conclude that AR-V567es expression has been significantly overestimated in its in vivo incidence, rendering conclusions based on previous detection systems highly doubtful. Our study further demonstrates the necessity of a comprehensive validation of detection approaches for future biomarkers before drawing clinical conclusions.
The number of therapies for treatment of metastatic castration-resistant prostate cancer (mCRPC) has considerably evolved during the last decade, primarily by using novel androgen receptor (AR)-axis–targeting agents as well as taxane-based chemotherapy. Nonetheless, virtually all patients suffering from mCRPC develop resistance to these agents.
One mechanism leading to resistance is the expression of AR splice variants (1, 2). These variants display a shortened protein structure missing the ligand-binding domain (LBD), making these proteins invulnerable to androgen deprivation therapy (ADT). The most prominent variant, AR-V7, has been analyzed extensively in patients with mCRPC. Its expression seems to correlate to resistance (3). However, a high number of patients without AR-V7 also fails to respond. Furthermore, a non-negligible number of patients show response to abiraterone or enzalutamide despite AR-V7 expression (4). Thus, the use of AR-V7 as a predictive biomarker in clinical routine is currently being discussed intensively (5–7).
AR-V567es (AR-V12) is another prominent AR variant. Skipping of exons 5, 6, and 7 leads to a shorter AR protein structure, which harbors a dysfunctional LBD (Fig. 1A; refs. 8, 9). This aberrant LBD is also invulnerable to ADT. In vitro studies revealed the constitutive activity of AR-V567es, that is, AR-V567es is able to translocate to the nucleus without hormonal activation, thereby inducing expression of AR target genes (10, 11).
Several studies describe AR-V567es expression in patients with prostate cancer, mostly related to late-stage prostate cancer (8, 9, 12–14). However, there are conflicting results regarding incidence of AR-V567es, displaying detection levels from 3% up to more than 70% (13–15).
Clinically, AR-V567es has gained interest because of its unique protein structure. In particular, the presence of an intact hinge region—in contrast with most other AR splice variants—might be related to taxane sensitivity. It has been shown that the hinge region is required for AR-microtubule association and downstream nuclear translocation (16). Taxane treatment inhibits translocation by microtubule stabilization, thereby disturbing the transcriptional activity of AR (17–19). Thus, tumors driven by AR-V567es expression might be particularly affected by taxane-based treatment (16).
Given that the presence of a single variant, that is, AR-V7, is not sufficient in stratifying patients with mCRPC for resistance to anti-hormonal therapies, a combination of AR variants might serve as a more reliable tool to further increase the predictive property of AR variant expression. In addition, tumors expressing an AR variant with an intact hinge region might be predictive for response to chemotherapeutic agents rather than hormonal therapies.
However, similar to AR-V7, different detection systems for analysis of AR-V567es mRNA expression have been described without an accurate comparison or a validation of performance. This might lead to fluctuating detection levels (15, 20). Therefore, we compared a set of AR-V567es mRNA detection PCR assays in different tumor samples (e.g., cancer cell lines, xenograft tissue as well as clinical samples).
Materials and Methods
Study design and ethics approval
We performed comparison of 7 AR-V567es mRNA detection systems by using three different approaches: comparison using (i) cancer cell lines in vitro, (ii) xenograft tissue, and (iii) clinical tumor samples [primary prostate cancer tissue, lung and lymph node metastatic tissues and circulating tumor cell (CTC) samples from patients with mCRPC]. Studies on patient material were conducted in accordance with The Declaration of Helsinki. The local institutional review board approved the study and all patients provided written informed consent (vote 2007-467-f-S). Primary prostate cancer tissue samples were obtained by the Department of Urology, Molecular Urooncology, University of Heidelberg School of Medicine, Heidelberg, Germany (votes 206/2005 and 207/2005 of the Ethics committee of the University of Heidelberg School of Medicine). Samples were provided by the tissue bank of the National Center for Tumor Diseases (NCT) Heidelberg in accordance with the regulations of the tissue bank. Xenograft and metastatic RNA and genomic DNA samples were obtained by the Prostate Cancer Biorepository Network (PCBN).
Cell lines, mRNA isolation, and cDNA synthesis
Human prostate cancer cell lines 22Rv1, LNCaP, and PC-3, human renal cancer cell lines (CAKI-1, CAKI-2) and human bladder cancer cell lines (RT-4, RT-112) were purchased from the Leibniz-Institute DSMZ GmbH (Braunschweig, Germany) in 2018. The human prostate cancer cell line VCaP was purchased from ECACC (European Collection of Authenticated Cell Cultures) via SigmaAldrich (Pasching, Germany) in 2018. All cell lines were cultured under matching protocols at 37°C and 5% CO2. We purchased all media from SigmaAldrich (Pasching, Germany). Trypsin-EDTA, PBS, and FCS were purchased from ThermoFisher Scientific. Upon arrival, cell lines were immediately expanded. RNA isolation was performed within the first 10 passages. For isolation of total RNA, we used the RNeasy Mini kit (Qiagen, Hilden, Germany) following the manufacturer's guide. 500 ng of total RNA (from primary prostate cancer, LuCaP xenograft or metastatic tissue) were reverse transcribed using the Quantitect Reverse Transcription Kit along with gDNA wipeout buffer (Qiagen, Hilden, Germany).
We collected peripheral venous blood (5–10 mL) in EDTA blood collection tubes (Sarstedt, Nuernbrecht, Germany). Samples were processed within 4 hours after blood drawing.
For CTC enrichment, we used the Dynabeads Epithelial Enrich Kit followed by mRNA isolation using the Dynabeads mRNA DIRECT Purification Kit (both ThermoFisher Scientific).
Reverse transcription of mRNA was achieved using the SuperScript IV VILO Master Mix (ThermoFisher Scientific) in a 20 μL reaction using the following settings: 25°C for 10 minutes, 55°C for 20 minutes, and 85°C for 5 minutes.
For determination of CTC, we performed detection of KLK3-PSA mRNA using a KLK3-PSA TaqMan PCR assay (Hs03063374_m1; ThermoFisher Scientific). A patient sample was determined as CTC-positive when displaying a qPCR signal for KLK3-PSA. AR-V7 status was analyzed by using a previously described custom-made TaqMan PCR assay (2). All qPCR runs were performed along with TaqMan PCR assays for housekeeping genes RPL37A (Hs01102345_m1) and HPRT1 (Hs99999909_m1; ThermoFisher Scientific).
Detection of AR-V567es expression using qPCR analysis
We performed qPCR analysis of AR-V567es expression using either SYBR Green chemistry for PCR assays #1 (9), #2 (12), and #3 (21) or TaqMan chemistry for PCR assays #4 (22), #5 (23), #6 (24), and #7. Primer and probe sequences are listed in Supplementary Table S1. For analysis of AR full length (AR-FL) cDNA, we used pSG5-AR-FL plasmid-DNA as template DNA. All experiments were performed using appropriate positive- and negative cDNA-, as well as water controls. PCR reactions were performed as follows:
SYBR Green qPCR.
SYBR Green qPCR reactions were performed using appropriate PCR assays along with the PowerUp SYBR Green Master Mix (ThermoFisher Scientific) and the following PCR protocol: 95°C x 15s and 60°C x 60s for 40 cycles followed by melting curve analysis.
TaqMan qPCR reactions were performed using appropriate PCR assays along with the TaqMan Fast Advanced Master Mix (ThermoFisher Scientific) using the following protocol: 95°C x 15s and 60°C x 60s for 45 cycles.
Both SYBR Green and TaqMan qPCRs were run on a QuantStudio 3 qPCR cycler (ThermoFisher Scientific). PCR products of 6 samples (LuCaP 86.2, LuCaP 136s2, LuCaP 136p8, patient samples VL4KJ2, 9V9DIB and pSG5-AR-FL plasmid) along with non-template control (NTC) were run on a Bioanalyzer HS Chip (Agilent).
Analysis of AR-FL and AR splice variant expression
We performed analysis of AR-FL and AR splice variants (AR-V3, AR-V7, and AR-V9) expression using AR-FL TaqMan PCR assay (Hs00171172_m1; ThermoFisher Scientific) as well as custom-made TaqMan PCR assays for AR variant detection. PCR assay sequences are listed in Supplementary Table S1. TaqMan qPCR was performed as described above.
Genomic DNA analysis
Genomic DNA (gDNA) of a benign prostate hyperplasia (BPH) tissue sample was used as non-cancer control. gDNA was isolated using the ChargeSwitch gDNA Mini Tissue Kit (ThermoFisher Scientific). PCR was performed using the HotStarTaq Master Mix Kit (Qiagen) along with the following protocol: 95°C x 45s and 60°C x 45s and 72°C for 90s for 30 cycles. PCR products were run on a 2% agarose gel. Imaging was performed on a Fusion FX (Vilber Lourmat, Marne-la-Vallée Cedex 3, France).
Design and validation of a custom-made AR-V567es detection PCR assay
First, we designed a custom-made TaqMan PCR assay for detection of AR-V567es mRNA using primers flanking the exon junction of exon 4 and exon 8 in combination with a hydrolysis probe spanning the junction of exon 4 to 8 (Fig. 1A and B). PCR assay specificity was verified by using synthetic custom-made DNA containing the region of exon 4 adjacent to exon 8, thereby displaying the unique cDNA sequence of AR-V567es (not shown). We additionally detected AR-V567es in LuCaP xenografts 86.2 and 136s2, in which this variant has been detected initially (Fig. 1D; ref. 9). These results demonstrate the validity of our custom-made probe-based PCR assay.
Comparative analysis of AR-V567es detection approaches
Next, we performed comparison of 7 distinct AR-V567es mRNA detection systems. Six of these PCR assays have been published previously (#1 (9), #2 (12), #3 (21), #4 (22), #5 (23), #6 (24). In addition, we used the custom-made probe-based PCR assay (#7; Fig. 1B and C). For comprehensive analysis we used various tumor samples (i.e., cancer cell lines, xenografts, primary, and metastatic tumor samples, CTC samples).
Cancer cell lines.
Expression of AR-V567es has been detected in prostate cancer cell lines, for example, 22Rv1 and VCaP (1, 9, 12). Hence, we performed analysis in different cancer cell lines, that is, 4 prostate cancer cell lines, 2 renal cancer cell lines and 2 urothelial cancer cell lines (Fig. 2A). We did not detect AR-V567es expression in urothelial cancer cell lines. Two PCR assays (#2 and #4) displayed signals in renal cancer cell lines. In prostate cancer cell lines, 5 PCR assays showed signals, although at different frequencies. No signals were detected using PCR assays #5 and #7. These results demonstrate inconsistency of detection between distinct AR-V567es detection systems.
Initially, expression of AR-V567es has been detected in LuCaP xenograft samples (9). Thus, we compared expression in six different LuCaP xenograft samples. All PCR assays robustly detected expression of AR-V567es in xenografts 86.2 and 136s2. The remaining samples displayed variable signals (Fig. 2B). Three PCR assays (#2, #4 and #6) detected signals in all xenografts. PCR assay #3 detected a signal in one additional xenograft. Again, these results demonstrate discrepancy of detection among different approaches.
Clinical tissue samples.
Besides expression in xenografts, AR-V567es has been found in prostate cancer patient tumor samples, most frequently at late stages (8, 9). Therefore, we performed analysis of AR-V567es expression in distinct prostate cancer tissue samples [primary prostate cancer tissue (n = 5) and metastatic prostate cancer tissue derived from either lymph node or lung metastases (n = 10)].
In primary prostate cancer tissue, 4 PCR assays did not detect signals, whereas 3 PCR assays (#2, #4, and #6) detected signals in 80%, 100%, or 60% of the samples, respectively (Fig. 2C). The latter three PCR assays also detected signals in 100% of metastatic samples (Fig. 2D). The PCR assays #1 and #3 detected signals in 4 out of 10 and 1 out of 10 metastatic samples, respectively. No signals were detected using PCR assays #5 and #7, which again demonstrates variability of different systems with respect to AR-V567es expression.
Circulating tumor cells.
Next, we conducted analysis of CTC samples of patients with mCRPC (n = 205). We used dichotomous KLK3-PSA expression for determination of CTC as being described previously (2). We detected CTC in 139 patients (68%). AR-V7 expression was detected in 40% of patients (82 out of 205).
Given the limited sample volume upon enrichment, we performed comparative analysis of 7 PCR assays in n = 20 patient samples, all of which were determined CTC positive (Fig. 2E). We concordantly detected a positive signal in one patient sample (VL4KJ2) in all PCR assays. PCR assays #4 and #2 detected signals in 100% and 90% of all samples, respectively. PCR assay #6 detected signals in 35% of all samples. PCR assays #1 and #3 detected signals in 10% of all samples. These results demonstrate an extreme variability from 5% to 100% positive samples.
Next, we analyzed the remaining number of n = 185 patient samples using PCR assay #7 (custom-made probe-based PCR assay). We did not detect a signal in any of the remaining patient samples. Essentially, using PCR assay #7 we detected one AR-V567es positive sample in 205 CTC samples (0,5%).
Overall, the results regarding expression of AR-V567es in numerous distinct tumor samples display eminent inconsistency in signal detection among different systems (Table 1).
|Assay .||Number of positive signals .||Rate of positive detection (%) .||Number of positive signals (including 205 CTC samples; n = 220) .||Rate of positive detection (including 205 CTC samples; n = 220; %) .|
|Assay .||Number of positive signals .||Rate of positive detection (%) .||Number of positive signals (including 205 CTC samples; n = 220) .||Rate of positive detection (including 205 CTC samples; n = 220; %) .|
Nonspecific detection in the absence of AR-V567es
Two of the seven PCR assays (#5 and #7) displayed concordant results in all tumor samples. All remaining 5 PCR assays displayed varying signals. Of these 5 PCR assays, 4 assays (#1, #2, #3, and #6) share a common feature: One oligonucleotide (either the forward or reverse primer) spans the exon junction of exon 4 to exon 8. Thus, we performed sequence analysis of the edges of the exon junction of AR-V567es (exon 4 to exon 8) and the AR wildtype junctions of exon 4 to exon 5 and exon 7 to exon 8. We detected some similarity in nucleotide sequence between the AR-FL exon junctions and the AR-V567es exon junction (Fig. 3A).
This similarity prompted us to assess whether nonspecific interaction of detection PCR assays with AR-FL in the absence of AR-V567es might occur. Therefore, we conducted qPCR reactions using pSG5-AR-FL plasmid-DNA (25) as template DNA in decreasing concentrations. This plasmid contains the AR-FL cDNA without presence of sequences similar to AR-V567es cDNA. We detected signals using PCR assays #1, #2, #3, #4, and #6 at varying concentrations, whereas no signals were detected using PCR assays #5 and #7 (Fig. 3B).
Next, we aimed to visually analyze the size of PCR products obtained in different samples. Therefore, PCR products of 6 samples (positive controls LuCaP 86.2 and LuCaP 136s2, LuCaP 136p8 (varying signals), patient samples VL4KJ2 (positive in all PCR assays) and 929DIB (varying signals) and pSG5-AR-FL plasmid DNA) were run on a highly sensitive DNA Bioanalyzer chip (Supplementary Fig. S2). Two PCR assays (#2 and #6) showed PCR products of AR-V567es in all samples. PCR assay #3 showed AR-V567es in both positive controls, both patient samples as well as the pSG5-AR-FL sample. PCR assay #1 showed AR-V567es in both positive controls, one patient (VL4KJ2) as well as pSG5-AR-FL sample. PCR assay #4 showed a PCR product of 694bp (AR-V567es cDNA) in both positive controls. In all other samples however, we detected a PCR product of 1128bp displaying AR-FL cDNA rather than AR-V567es cDNA. PCR assays #5 and #7 displayed AR-V567es in both positive controls as well as one patient sample (VL4KJ2). These results demonstrate correct detection of AR-V567es of all 7 PCR assays in positive control samples. Nonetheless, 5 of 7 PCR assays (#1, #2, #3, #4 and #6) displayed false-positive signals in samples without AR-V567es.
AR-V567es expression is based on genomic structural rearrangements rather than alternative splicing events
A study from Nyquist and colleagues (26) revealed genomic structural rearrangements (GSR) as the underlying reason for expression of AR-V567es in xenografts LuCAP 86.2 as well as 136s2. Whereas in LuCaP 86.2 the authors detected a deletion of the genomic region containing exons 5, 6, and 7, in LuCaP 136s2 an inversion of this genomic region leads to a structural rearrangement of the AR gene. A study by Henzler and colleagues (27) detected AR-V567es mRNA expression as a result of a serial GSR of the AR genomic structure in one out of 15 CRPC patient samples.
Therefore, we evaluated the genomic DNA (gDNA) of LuCaP xenograft and metastatic tissues. When using deletion specific primers, we detected a PCR product in LuCaP 86.2, demonstrating a deletion of the genomic region of exons 5, 6, and 7 (Supplementary Fig. S3A). By using primers specific for an inversion of the exons 5, 6, and 7, we detected a signal in LuCaP 136s2. No such GSR was detected in any of the other tumor samples (Supplementary Fig. S3B).
These results are in line with results using PCR assays #5 and #7, displaying AR-V567es exclusively in xenografts 86.2 as well as 136s2. Unfortunately, we had no opportunity to perform gDNA sequencing of the one patient sample displaying AR-V567es in all PCR assays. Nonetheless, these results validate that expression of AR-V567es is based on GSR rather than alternative splicing events.
AR-V567es expression occurs subclonally
Xenograft LuCaP 86.2 displays expression of both AR-FL and AR-V567es mRNA throughout passaging (26). By running deletion specific as well as wildtype-specific PCR, we detected both wildtype gDNA as well as gDNA containing a GSR (Supplementary Fig. S3A). These results demonstrate the presence of a stable subclone of AR-V567es–positive cells.
In case of xenograft LuCAP 136, expression of AR-V567es has been exclusively detected in an early passage (s2), whereas later passages did not display AR-V567es (26). When running a TaqMan-based qPCR we detected extremely low levels of AR-FL mRNA in the early passage compared with a later passage [approximately 1,000-fold change lower (ΔCt 10,69; Fig. 4A; Supplementary Fig. S4A; Supplementary Table S2]. This demonstrates the presence of a small population of AR-FL containing cells in the early passage. The later passage, however, did not show signs of GSR, indicating the loss of the AR-V567es subclone during passaging of the xenograft.
These results reveal co-existence of both, cells without GSR expressing AR-FL and cells which have undergone GSR to express AR-V567es in the same tumor sample. This co-expression might occur in both a stable form (xenograft 86.2) as well as an unstable form (xenograft 136).
Coexpression of AR-V567es and AR splice variants
A GSR of the genomic region of exons 5, 6, and 7 would still contain the exons 1 to 4 in its correct order. Given that most AR variants are based on alternative splicing of exon 3 to cryptic exons located between exon 3 and 4 (1, 10), a GSR-positive cell might still be able to express AR splice variants in concurrence to GSR-related AR-V567es expression. Thus, we analyzed expression of AR splice variants in LuCaP xenografts. We detected expression of AR-V7, AR-V3 and AR-V9 at similar levels in all xenografts, even in 136s2, which almost exclusively contains GSR-positive cells (Fig. 4B; Supplementary S4B–S4D; Supplementary Table S2). This suggests that cells which have undergone a GSR might still express AR splice variants in concurrence to AR-V567es.
Here, we report a comparative analysis of approaches for detection of AR variant AR-V567es in a variety of distinct tumor samples. Concordant results for AR-V567es using 7 distinct detection systems were found in two xenograft tissues as well as one out of 35 patient samples, whereas the majority of samples displayed discrepant results. These discrepancies are relevant from a technical, biological as well as a clinical perspective.
From a technical perspective, the comprehensive validation of a biomarker detection approach is one of the most critical steps before transferring it into clinical applications. Nonetheless, different approaches might lead to identical results (28).
AR-V567es was initially found in LuCaP xenografts which were derived from metastatic tissues (9). Further studies revealed its expression mostly related to late-stage mCRPC, whereas expression in primary tumor was low or even absent (8, 12, 14, 23, 29). Recently, AR-V567es expression has been detected in liquid biopsy samples of patients with prostate cancer (13, 22, 24). However, there have been conflicting results regarding the incidence of AR-V567es, that is, a range of patients with AR-V567es–positive mCRPC from 3% (14) up to over 70% (13, 15, 24). Besides, different approaches for detection of AR-V567es mRNA have been published without evaluation of a higher specificity or validity of one approach over another (15, 20).
These discrepancies prompted us to first, construct a highly sensitive AR-V567es mRNA detection system and second, perform a comprehensive comparison of distinct detection systems all of which have been described to detect AR-V567es in clinical patient material.
Analysis of distinct clinical tumor samples (primary prostate cancer tissue, metastatic tissue, and CTC) using 7 detection systems displayed extensive variety of AR-V567es positivity, ranging from 1 to 35 out of 35 samples. Whereas two PCR assays (#5 and #7) detected AR-V567es expression in only one clinical sample (VL4KJ2), the remaining PCR assays detected signals heterogeneously. Conclusively, different detection systems failed to consistently demonstrate AR-V567es expression, except for three samples.
Initial detection of AR-V567es was based on PCR reactions run by two oligonucleotides, one of which spans the exon junction of exon 4 to exon 8. Three of the analyzed PCR assays (#1, #2, #3) rely on this feature. However, sequence similarity exists between the 5′ edges of exon 4 and exon 7 as well as the 3′ edges of exon 5 and exon 8. Thus, nonspecific binding of exon-spanning oligonucleotides might occur in the absence of AR-V567es. Indeed, we observed signals of these PCR assays when using AR-FL DNA as PCR template. Therefore, a positive signal using these PCR assays does not stratify whether true AR-V567es was present or whether nonspecific binding to AR-FL in the absence of AR-V567es has led to a positive signal.
As the use of two oligonucleotides might be at risk of nonspecific binding, thereby leading to false-positive results, a probe-based PCR assay might increase sensitivity of a detection system. The PCR assays #4 and #6 rely on a hydrolysis probe along with two oligonucleotides.
PCR assay #4 demonstrates the highest frequency of positive signals in clinical samples (100%). This PCR assay consists of primers located in exon 1 and exon 8 along with a hydrolysis probe spanning the exon junction of exon 2 and exon 3. Thus, the use of this PCR assay will lead to amplification of two PCR products: AR-V567es and AR-FL. To avoid variety in amplification efficiency, qPCR assays should not exceed a certain PCR size of approximately 150 bp (30). However, both PCR products significantly exceed this size limitation (694 bp for AR-V567es vs. 1128 bp for AR-FL). Therefore, this PCR assay on its own is not valid in discriminating AR-FL from AR-V567es expression and would require further downstream analysis, that is, PCR product size analysis.
PCR assay #6 comprises of two oligonucleotides along with a hydrolysis probe and was used in the very sensitive setting of digital droplet PCR for detection of AR-V567es (24). Sanger sequencing validated both the correct size and correct sequence of the PCR product. This study revealed expression of AR-V567es in 78% of patients with mCRPC. Using this PCR assay, we detected the second most common number of signals in clinical samples (20 out of 35 total samples; 57%). Typically, the use of a hydrolysis probe located in-between the flanking primers enhances the sensitivity of a PCR assay. However, this particular PCR assay consists of a reverse primer that spans the AR-V567es exon junction of exon 4 to exon 8. Given the sequence similarity shared by AR-FL and AR-V567es, this PCR assay might be at risk for nonspecific binding to AR-FL in the absence of AR-V567es. This nonspecific binding would lead to a 5′-specific binding of the reverse primer along with an 3′-overhang. This overhang, however, would be refilled during PCR. Thus, after several cycles a correct sequence would be obtained, although the original sample did not include the accurate complementary sequence. Therefore, both gel electrophoresis and Sanger sequencing analysis will not surely lead to the precise stratification of an AR-V567es–positive sample or whether nonspecific binding to AR-FL occurred during PCR. Definitely, using this PCR assay we detected signals when running PCR with AR-FL DNA. Therefore, this PCR assay displays nonspecific binding to AR-FL in the absence of AR-V567es.
Overall, we conclude that most published PCR assays for detection of AR-V567es mRNA have to be assumed flawed in a way that—in the absence of AR-V567es—these PCR assays non-specifically bind to AR-FL. For most of these PCR assays however, a valid discrimination between binding to AR-FL rather than AR-V567es is not possible. Decisively, we have to point out that earlier studies correlating AR-V567es expression with treatment results or mechanistic effects have to be considered defective or even wrong based on misclassification of samples.
Nevertheless, two PCR assays (#5 and #7) displayed signals in established AR-V567es–positive tumor samples, that is, LuCaP xenografts 86.2 and 136s2 as well as in one patient sample (VL4KJ2) without any signs of nonspecific detection in the absence of AR-V567es. We consider these PCR assays being the most valid systems for detection of AR-V567es mRNA in clinical samples.
From a biological perspective the dissection of mechanisms of how tumor cells acquire aggressive characteristics is of huge interest to understand the course of the disease. In vitro studies suggest an increase of aggressiveness of tumor cells, when AR-V567es is overexpressed (8, 9, 11, 31). A major mechanisms of AR variants is to either homodimerize or heterodimerize with AR-FL or AR variants to induce AR target gene expression (9, 11). The ability of AR-V567es to heterodimerize to AR-FL has been shown in vitro and as such, is in line with reports describing a heterodimerization of AR variants to AR-FL, thereby leading to induction of proliferative characteristics in tumor cells (11, 12, 31–35).
However, recent publications demonstrate that AR-V567es expression is based on GSR rather than alternative splicing events (26, 27). Given, that only one copy of the X chromosome is present in male cells, co-expression of AR-V567es and AR-FL in the same cell can occur only as a result of extensive GSR: an amplification of the complete AR gene of >180 kb, followed by a GSR leading to alteration of the genomic structure of one of the AR copies (26, 27). So far, AR amplification followed by complex GSR has not been described to occur in vivo. Therefore, any kind of analysis describing dimerization of AR-V567es and AR-FL as a reason for increased aggressiveness has to be considered skeptically.
Nevertheless, the present study as well as previous studies show expression of both AR-FL as well as AR-V567es expression in the same tumor tissue. This simultaneous co-expression, however, occurs clonally, that is, tumor cells express either wildtype AR or have undergone a GSR to express AR-V567es. In LuCaP xenograft 86.2 different passages display stable co-expression of both AR-FL and AR-V567es. Contrary, the LuCaP xenograft 136 displayed expression of AR-V567es in the early passage s2, whereas later passages did not show AR-V567es expression. We now validate the presence of a small population of AR wildtype cells in the early passage of xenograft 136s2. In non-castrate culture conditions these AR wildtype cells gained a survival benefit over GSR positive cells. Thus, the loss of AR-V567es is not based on alternative splicing but rather an overgrow of AR-FL cells over GSR-positive cells.
Nonetheless, heterodimerization of AR-V567es and AR splice variants might occur in GSR positive cells even in the absence of wildtype AR. Most clinically relevant AR splice variants are based on splicing events of exon 3 to cryptic exons located in intron 3 or intron 4, a sequence which is still present even in GSR positive cells (1, 10). We now show for the first time simultaneous co-expression of AR-V567es and other splice variants even in a clinical sample containing almost exclusively AR-V567es–positive cells (LuCaP xenograft 136s2).
The true biological nature of AR-V567es as a potential inducer of tumor aggressiveness by heterodimerization with AR-FL seems to be overestimated by in vitro experiments not mirroring the native biology of the disease. This is in line with reports describing the dissimilarity of in vitro experiments not reflecting the in vivo situation in patients with mCRPC (36).
From a clinical perspective, two major points have to be parsed: (i) the validity of a biomarker for transfer into clinical applications, and (ii) the handling of a clinical biomarker for successful patient stratification for certain treatment regimen.
In case of validity, we have to point out that the presented comprehensive analysis undoubtedly verified the failure of AR-V567es being a reliable clinical biomarker. Concordant expression of AR-V567es was detected only in two xenograft samples and one out of 220 clinical samples, including primary tumors, metastases and CTC. This leads to a percentage of approximately 0.5% positive samples and further implies that AR-V567es—although possible in rare cases—has been extremely super-valued with respect to frequency.
Recent studies showed AR-V567es expression in late-stage prostate cancer, up to an extent of more than 70% (22, 24). Given, however, that these studies might be related to nonspecific detection of AR-FL in absence of true biological expression of AR-V567es, these detection rates have to be considered skeptically. Conclusively, the low rate of detection clearly demonstrates ineffectiveness of AR-V567es as a valid clinical biomarker.
In case of clinical application, AR-V567es has been connected to response to non-hormonal treatment agents, that is, taxane-based agents. This assumption is based on the unique protein structure of AR-V567es, which—in demarcation to other AR splice variants—still contains an intact hinge region which presumably might be a target for taxane-based treatment (17–19). However, we have to point out that clinical response to taxanes might be inhibited by simultaneous co-expression of AR variants lacking an intact hinge region, as these AR variant proteins are not affected by taxane-based microtubule targeting. Therefore, a clinical benefit might occur only if a tumor exclusively expresses AR-V567es. Yet, we demonstrate, that even in a tumor sample, in which the majority of cells expresses AR-V567es, AR splice variants are also present at levels similar to AR-V567es–negative samples. This co-expression of AR-V567es and other AR splice variants, however, would imply intrinsic resistance against taxane-based chemotherapeutics.
Ultimately, we have to sum up, that AR-V567es expression does not reach criteria for being a valid clinical biomarker for stratification of patients being treated with either antihormonal or chemotherapeutic agents. In addition, we have to emphasize, that most studies analyzing the biological and clinical relevance of AR-V567es, are presumably based on false-positive detection, questioning the impact of AR-V567es in prostate cancer progression and further underscores the necessity for valid analysis of biomarker detection approaches.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: C. Bernemann, V. Humberg, J. Steinestel, A.J Schrader, M. Boegemann
Development of methodology: C. Bernemann, V. Humberg
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Bernemann, V. Humberg, X. Chen, S. Duensing, M. Boegemann
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Bernemann, V. Humberg, M. Boegemann
Writing, review, and/or revision of the manuscript: C. Bernemann, J. Steinestel, A.J Schrader, M. Boegemann
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Bernemann, J. Steinestel, X. Chen, S. Duensing, A.J Schrader
Study supervision: C. Bernemann, A.J Schrader
Other (sampling of blood): B. Thielen
This study was supported by the Anneliese-Pohl Stiftung. The authors thank the Prostate Cancer Biorepository Network (PCBN) for providing material (RNA and gDNA of LuCaP xenografts as well as metastatic tissue samples). The pSG5-AR-FL plasmid was a gift from M. Cronauer, University of Luebeck, Germany.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.