Targeted proteomics for the detection of androgen receptor variants in preclinical models of castration-resistant prostate cancer

Abstract

Background

The emergence of Androgen Receptor variants (AR-Vs) during hormonal treatment is believed to be one of the key mechanisms of resistance. AR-Vs lack the ligand-binding domain and function as constitutively active transcription factors, rendering AR-targeted endocrine-based therapeutics ineffective in tumors that express AR-Vs. The goal of this study was to develop a targeted multiplex assay to identify known and unexplored AR-V proteins and help guide the stratification of PCa patients based on their AR-V protein profiles.

Methods

A targeted multiple reaction monitoring (MRM) assay using liquid chromatography-tandem mass spectrometry (LC–MS/MS) was developed to quantify total AR proteins using the AR N-terminal domain (AR-Nterm), a well-characterized AR splice variant AR-V7 (AR-V7), a putative, previously identified but uncharacterized protein isoform (AR-V12) and isoforms that code for an exon 3 duplicated junction seen in AR-V2 and AR- full length with two copies of exon 3. Specimens included three AR-positive PCa cell lines (22Rv1, VCaP16, LNCaP), two AR-negative PCa cell lines (PC3, DU145), and a patient-derived xenograft (PDX) model of CRPC. Endogenous target peptides were quantified according to tier-2 targeted mass spectrometry “fit-for-purpose” guidelines to ensure analytical rigor.

Results

Analytical performance was characterized across a linearity range of 4.15 fmol to 4000 fmol followed by evaluations of precision, accuracy, selectivity, stability, ion suppression and repeatability studies. The overall coefficient of variation (CV) was < 20% across seven independent validation experiments. Endogenous AR-V peptides were accurately quantified using calibration curves, achieving acceptable linear regression coefficients (R² > 0.95). Four targets, including AR-Nterm and 3 AR-variants, were identified and quantified in PCa cell lines and in a PDX pre-clinical model.

Conclusion

This study establishes a robust, validated MRM targeted proteomics assay capable of reproducibly detecting and quantifying both known and previously uncharacterized AR-Vs at the protein level  . The assay demonstrated < 20% analytical variation at the lower limit of quantification of 4.25 femtomole (fmol) for AR-Nterm and 8.25 fmol for AR-V7, AR Exon3-Dup/AR-V2 and AR-V12 and, showed robust analytical performance in preclinical models and is ready for further evaluation in clinical specimens. This provides a clinically applicable platform for the comprehensive profiling of AR-V proteins in human subjects, enabling improved stratification of patients with CRPC and supporting precision therapeutic strategies.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12014-026-09623-2.

Introduction

Prostate cancer (PCa) is the most prevalent cancer in men in the United States with approximately 313,780 new cases in 2025, according to the American Cancer Society [1]. Early detection via regular screening of serum prostate-specific antigen (PSA) levels has facilitated PCa diagnosis at organ-confined stages [2]. The androgen receptor (AR) is the main driver of PCa progression and is the primary target for therapeutic treatment. For example, if a patient has locally advanced or metastatic PCa, contemporary first-line treatment regimens involve androgen deprivation therapy (ADT) with or without the addition of an androgen receptor pathway inhibitor (ARPI) such as abiraterone acetate or enzalutamide [3, 4]. While these treatments are initially effective, most tumors typically progress to a more aggressive disease state known as castration-resistant PCa (CRPC) [3, 5]. Although a small minority of CRPC tumors lose or exhibit very low AR protein expression or activity, the majority restore AR signaling through various mechanisms that include the expression of constitutively active AR variants (AR-Vs) [4, 5]. In particular, the expression of AR-V7 has been associated with abiraterone acetate and enzalutamide resistance in CRPC [5–8].

Over 20 distinct AR-V mRNA species have been identified in bulk CRPC tumors to date [5, 9–11]. AR-Vs lack the C-terminal ligand-binding domain of the AR protein but retain the N-terminal transcriptional activation domain and the DNA binding domain, which enables them to function as constitutively active transcription factors independently of an agonist ligand [9, 12, 13]. AR splicing aberrations caused by alternative polyadenylation [14], structural rearrangements of the AR gene [11, 15], or stabilization of AR-V mRNAs [16], are mechanisms of AR-V emergence in CRPC [14, 17].

Recently, a multicenter prospective study, PROPHECY [18], evaluated AR-V7 mRNA and antibody-based nuclear AR-V7 protein localization in circulating tumor cells (CTCs) of CRPC patients prior to treatment with enzalutamide and/or abiraterone. In this study, among the 28 patients who were AR-V7 positive by either assay (at least one of the two), only 32% (9/28) showed positive detection on both assays. Of the remaining patients, 17 were positive by the mRNA assay only, and 2 were positive by the protein assay only. Furthermore, AR-V7 positivity only partially explained therapy resistance in this study, suggesting that additional AR-Vs may play a role [18].

Other variants such as AR-V12 (also known as ARv567es) is detectable at the mRNA and protein level in CRPC tumors that harbor genomic deletions of AR exons 5–7 [19–22]. Furthermore, 22Rv1 cells harbor a genomic tandem duplication of AR exon 3, which underlies expression of an AR-full length mRNA variant with two tandem spliced copies of exon 3 [15]. This tandem duplication also underlies the expression of AR-V2, which contains two tandem spliced copies of exon 3, but has been detected at the mRNA level only [23]. AR gene rearrangements have been shown to be common in genomes of CRPC cells, and enriched in patients treated with AR inhibitors, indicating that detection of AR-V proteins arising from AR gene rearrangements could have clinical utility [11, 24, 25].

Previous studies have shown that gene expression at the mRNA level is often a poor predictor of protein activity [26] and yet protein expression is more informative as a prognostic biomarker over genomic, epigenomic, or transcriptomic markers alone [27]. Based on these premises, we aimed to identify protein expression of validated and putative AR-Vs in CRPC through targeted liquid chromatography tandem mass spectrometry (LC–MS/MS)-based proteomics. We used a Tier-2 targeted [28] LC–MS/MS approach to evaluate the protein expression of these AR-Vs. This work allows two-fold validation: to confirm detection and quantification of these AR-Vs at the protein level and to enable characterization of their protein expression profiles for future patient stratification based on AR expression patterns.

Results

Candidate peptide selection

Total AR protein using the AR N-terminal domain (AR-Nterm region) and AR-V-derived peptide targets were selected based on whether they have been confirmed at the protein level in PCa cell lines and/or PDX models (i.e. AR full-length, AR-V7, and AR-V12), or whether they have been only described at the mRNA level without any direct protein data (i.e. AR-V2) [9] [19, 20, 29, 30]. AR-Full length including with an Exon-3Dup, AR-V2, and AR-V7 are AR-Vs typically produced by splicing of cryptic exons located in AR intron 3. Conversely, AR-V12 is produced by splicing of AR exons 1, 2, 3, 4, and 8. Due to significant exon overlap with all variants, total AR was measured by designing a peptide in AR exon 1 (N-term). Recombinant full-length androgen receptor (AR) protein was subjected to tryptic digestion. Candidate peptides were evaluated in Skyline and cross-referenced with frequently observed peptides in the Global Proteome Machine (GPM) database (https://thegpm.org) to select the final target peptide. For AR-Exon3-DUP identification, there is no unique region to identify this variant except for the exon3-exon3 junction that overlaps with one other variant composed of exons 1,2,3,3,4,5,6,7,8. This variant was designated as Exon-3Dup that identifies gene-level duplication of exon 3. We designed AR-V-specific peptides based on novel amino acid sequences resulting from AR-V-specific splicing events located in exon junction regions that are unique to each AR-V (Fig. 1A).

Fig. 1.

Schematic of AR and AR-Vs illustrating the exons and the different exons encoding each AR-V. Each color bar above the AR-Vs represents the region of either location of amino acid sequence that was used to design the specific probe (as seen in the legend). Cryptic Exon (CE). Femtomole per microliter (fmol/µL). Single and multiple reaction monitoring (SRM/MRM). Grand Average of Values of hydropathy (GRAVY) score

Peptides were selected based on the following criteria: (1) No readily reactive amino acid residues (Histidine, Tryptophan, Asparagine/Glutamic acid followed by Glycine) or missed trypsin cleavage-containing peptides; (2) Unique sequences were (Fig. 1A-B) determined by BLAST analysis searches; (3) Hydrophobicity: peptide hydropathy score between 10 and 45; (4) Peptide length: 10–13 amino acids; (5) Synthesis purification to level A (ThermoFisher Scientific synthesis criteria A through C; A being the best); (6) Selective Reaction Monitoring/Multiple Reaction Monitoring (SRM/MRM) compatibility level 3 (using ThermoFisher Scientific criteria: Level 1–3, with level 3 being the best). In addition to these criteria, ion suppression studies showed that there was not repression of peptide signal at the specific retention times of the eluting peptides (Supplemental Fig. 1). Initially, we started with eight peptides (two peptides per protein) during method development. However, due to high hydrophilicity and bimodal peak distribution in some peptides, only one peptide per protein passed the acceptability criteria of CV < 20%. We therefore moved forward with one peptide per protein. Thus, a total of four peptides, one per protein, were monitored for the rest of the MRM assay validation studies (Fig. 1B).

Development and validation of the MRM assay

To develop the analytical method and establish the analytical performance of the assay, we used COS-1, a monkey kidney-fibroblast cell line that does not express AR [31] as a matrix blank. We used a liquid–liquid based extraction method applicable to cell lines and patient-derived xenograft (PDX) tumors [32, 33]. The analytical measurement range were determined using calibration curves ranging from 4.5 to 4000 fmol (0.45 to 400 fmol/μL) (Fig. 2A-D). The limit of detection (LOD), lower limit of quantification (LLOQ), upper limit of quantification (ULOQ) and linearity (represented by R² values from linear regression models > 0.95 with a range between 90 to 115% accuracy) for the four peptides ranged from 8.25 to 1000 fmol (0.825 to 100 fmol/μL). Calibration curves for the four peptides (AR-Nterm, AR-Exon3-Dup/ AR-V2, AR-V7, and AR-V12) demonstrated linearity of > 0.95 across all 12 standard points measured (Figs. 2E and 2F).

Fig. 2.

Calibration curve data for, A. AR-term; B. AR-V7; C. AR-Exon3-Dup; D. AR-V12; E. Calibration scheme of working calibrators from each standard (SD) stock solution and percent coefficient of variation (%CV) (SD0-10 n = 5, SD11 n = 3). F. Summary of the limit of quantification (LLOQ) and upper limit of quantification (ULOQ) and %CV per peptide. (The dotted boxed at the lowest range represents limit of detection (LOD). Log10 value was used for the y-axis. Standard data points were excluded if calibration curve was > 20%. Log10 value was used for the x-axis and the y-axis

Assay repeatability was evaluated using three quality control (QC) levels (low/medium/high) for each peptide (Fig. 3A). As expected, there was a higher percent variability at the low QC level relative to the medium QC level across all 4 peptides, which was mostly driven by AR-V7 and AR-Nterm peptides. At the high QC level, variability was mostly driven by AR-Nterm and AR-Exon3-Dup/ AR-V2 peptides. Among the three QC levels measured, two peptides did not pass overall acceptability criteria at the low QC level (AR-V7 and AR-V12) and AR-Nterm did not pass medium QC level (Fig. 3B). For each peptide, acceptable performance was demonstrated at a minimum accross two QC levels, meeting predefined acceptance criteria. Although three QC levels were evaluated, successful performance at two levels was sufficient to qualify the assay for further use. Using this approach, the AR-V peptides exhibited reproducibility across all tested concentration levels over five days, with total imprecision ≤ 20% CV (Fig. 3B).

Fig. 3.

Validation of repeatability demonstrating reproducibility of target peptide measurements. A Total percent variability for 4 peptides at low (0X LOQ), medium (50X LOQ), and high (100X LOQ) levels. B. Inter (between days) and Intra (within samples) percent coefficient of variation (%CV) was measured at each quality control (QC) level (low/medium/high) in 5 different days and 3 technical replicates per day. C. Percent change carryover analysis of QC Low after running a QC High sample. Femtomole per microliter (fmol/uL). Data has a minimum of n = 3–5 graphed data points are based on mean with SD)

Carryover was evaluated to determine whether an analyte from a high-concentration (210 fmol) injection contaminated subsequent analyses. A sequence was run consisting of a low-concentration (3.2 fmol) sample, followed by a high-concentration sample, and then re-injection of the same low-concentration sample. Carryover was defined as residual analyte signal in the second low-concentration sample attributable to the preceding high-concentration injection. Acceptance criteria were set at positive difference of < 20% of the lower limit of quantitation (LLOQ) response. Based on this design, no positive difference carryover was observed in the low sample after high concentration sample, with all measured responses in the post-high low sample well below the established threshold (Fig. 3C). The negative difference is acceptable after post-high injections observed in AR-Nterm and AR-Exon3-Dup/ AR-V2. The reduced signal observed in post-high low-concentration injections reflects transient ion suppression and chromatographic equilibration effects following high-concentration sample injections rather than analyte carryover. Therefore, no evidence of residual carryover analyte was observed. Overall, the total assay imprecision for AR-Nterm, AR-Exon3-Dup/ AR-V2, AR-V7, and AR-V12 at low, medium, and high QC levels remained below 20% CV across all peptides.

To confirm that the target peptides were not influenced by isobaric interferences from the endogenous matrix, the specificity and selectivity of the MRM assay were evaluated. Three peptide concentration levels (low, half-medium, and medium) were assessed in six cancer cell line biological replicates spiked with buffer-only (0X; no spike), 25X, and 50X of the lower limit of quantification (LLOQ), consistent with CPTAC Tier 2 selectivity guidelines. Acceptance criteria included: (1) individual slopes derived from the three concentration levels within 10% of the mean; (2) observed 25 × LLOQ values within 10% of predicted values; and (3) transition ion peak area ratios within 30% of the mean (Supplemental Data 4).

Selectivity was confirmed in a blank (no matrix) and in different PCa cell matrices where all four peptides demonstrated no interference, with average individual slope differences < 2.0% relative to the mean (Supplemental Data 4). In addition, measured concentrations at 25X and 50X of the LLOQ differed by < 1.0% from predicted values, confirming acceptable selectivity across the tested concentration range (Fig. 4A–D; Supplemental Data 4). For the light peptide, ion transition ratios were assessed to determine assay specificity. Peak area intensities for each monitored transition were extracted and averaged across duplicate runs. The mean transition ratios were then calculated to confirm consistency of ion fragmentation patterns and to verify analyte identity [34]. All ion transition ratios [34] for the four peptides were within 30% of the mean (Supplemental Data 3).

Fig. 4.

Validation of selectivity data. A. AR-Nterm, B. AR-Exon3-Dup, C. AR-V7, D. AR-V12. Each cell line represents a PCa cell line biological replicate in each quantification level. Data has a minimum of n = 3 per cell line and six different PCa cell lines (graphed data points are based on the mean with SD). Femtomole per microliter (fmol/µL). Lower Limit of quantification (LLOQ): Low (0X LLOQ), medium (25X LLOQ), and high (50X LLOQ) levels

Peptide stability in the MRM assay was evaluated by calculating the percent difference from day 0 (initial pooled standard mix) to 6 and 24 h post-mixing at various temperatures, a 3-h freeze/thaw post-mixing, and 4 weeks post-mixing. All four peptides that were subjected to varying temperature conditions and multiple freeze–thaw cycles demonstrated acceptable stability, with percent differences between the ranges of 0.1% to < 16% with respect to the concentration of 25 fmol at any given freeze thaw cycle, which is well within the established acceptance criterion of ± 20% (Supplemental Table 1).

In summary, assay development and validation included calibration curve-based determination of LOD and LOQ for the four target peptides. Assay repeatability (within- and between-run imprecision), selectivity, and stability were evaluated in accordance with CLSI EP05/EP15 principles and CPTAC Tier 2 targeted proteomics validation guidelines [28] (Table 1). Of the eight peptides initially evaluated, four met all predefined acceptance criteria and were advanced for subsequent analyses.

Table 1.

Development and validation of the AR-Vs SRM assay in cell lines and solid tumors

Quantification of AR-Vs in PCa cell lines

Quantification of four endogenous peptides, proxies for AR total (AR-Nterm) and three AR-Vs , was performed in multiple PCa cell lines. Each cell line was run in biological triplicate, with technical duplicate measurements. Three AR-positive PCa cell lines (22Rv1, LNCaP, VCaP16) were evaluated for AR-V protein differences. In addition, two AR-negative PCa cell lines (DU145 and PC3) and COS-1 (non-cancer cell line used as a matrix with negative AR expression) were included as negative controls. All peptides were quantified above LOQ. AR-V expression profiles varied across cell lines, with the AR-Nterm highly expressed in 22RV1 compared to the other cell lines (Fig. 5A). AR-V7 was the most abundant AR-V in VCaP16, while 22Rv1 cells expressed multiple AR-Vs of interest including AR-V7, AR-Exon3-Dup/ AR-V2, and AR-V12 (Fig. 5B-D). AR-Exon3-Dup/ AR-V2 was significantly detected in 22Rv1s indicating that exon-3 duplication is positive in this cell line as expected. Profiling of LNCaP demonstrated expression of AR-Nterm and very low levels of AR-V7 (Fig. 5A-B). As expected, no AR- or AR-V-derived peptides were detected in the AR negative cell lines. Following the tier-2 guidelines [27], we performed a 5 × 5 assay to test reproducibility and imprecision over 5 days in 5 different 22Rv1 processed samples. The overall CV was < 30%, which was mainly driven by a few replicates, indicating that the variability was due to sample preparation rather than the detection method (Supplemental Fig. 2). We further validated the expression of AR-full length and AR-V7 via western blot (Supplemental Fig. 3) and this was consistent with our AR proteomic panel for these two peptides.

Fig. 5.

Quantitative expression profiles of four peptides in prostate cancer cell lines. A. AR-Nterm, B. AR-V7, C. AR-Exon3-Dup, D. AR-V12, (n = 3 per sample group). Graphed data used Mean with SD. Statistical significance was obtained by ordinary one-way ANOVA followed by Dunnett’s multiple comparison test. *, **, *** indicate p < 0.05, p < 0.01, p < 0.001 respectively. Dotted line represents the LLOQ, below this is LOD

In parallel. we evaluated the protein expression of these AR-Vs in a clinically relevant castration-resistant PDX model LuCaP77 (non-castrated) and LuCaP77CR (castrated) known to have AR gene rearrangements[10]. All AR-Vs were expressed at baseline in intact model (LuCaP 77) and were significantly increased in the castration resistant model (LuCaP 77CR) (Fig. 6). One exception was AR-Exon3-Dup/ AR-V2 that was detected equally in both tumors grown in intact and castrated mice (Fig. 6) showing for the first time translation of this exon-3Dup regions in a PDX model.

Fig. 6.

Quantitative expression profiles of four peptides in PDX 77 and 77 CR pre-clinical model. A. PDX 77 and 77 castration resistance (CR) measurements of all 4 validated peptides. (n = 3 per sample group). B. Bar graph representing Fold change data from Fig. 6A Log 10 transformed. Graphed data used Mean with SD. Statistical significance was obtained by non-parametric ANOVA (6A) and student unpaired t-test (6B) *, **, *** indicate p < 0.05, p < 0.01, p < 0.001 respectively. Dotted line represents the LLOQ

Discussion

In this study, we developed and validated a targeted MRM assay to detect and quantify certain AR-Vs at the protein level in PCa cell lines and PDX models. While detection of many AR-Vs has been described at the transcript level, protein-level confirmation has been limited by lack of specific antibodies for most AR-V species. Our assay addresses this gap by enabling sensitive, multiplexed detection of total AR (AR-N-term), AR-Exon3-Dup/AR-V2, AR-V7, and AR-V12, with validation performed under CPTAC Tier-2 guidelines. Our assay demonstrated excellent analytical performance, including sub-femtomole sensitivity, linearity (R² > 0.95), reproducibility (< 20% CV across five days in for most peptides), and stability (< 16% CV deviation under varied temperature and freeze–thaw conditions). Selectivity experiments confirmed minimal matrix interference, supporting assay robustness across diverse sample types.

Application to PCa cell lines revealed distinct AR-V-specific protein expression patterns. As expected, 22Rv1 cells expressed multiple AR-Vs (AR-Exon3-Dup/ AR-V2, AR-V7, AR-V12), while VCaP16 predominantly expressed AR-V7, and LNCaP displayed overall very low AR-V levels. AR-negative cell lines (DU145, PC3, COS-1) showed no detectable AR-Vs, confirming assay specificity. Importantly, AR-Exon3-Dup/ AR-V2 protein expression was detected in 22Rv1 cells and PDX tumors, providing the first peptide-level confirmation of this AR-V, previously characterized only at the mRNA level. AR-V12 detection in 22Rv1 and PDXs further supports prior reports of AR-V12 protein presence in advanced PCa [19–21].

This study provides the first endogenous protein-level measurements of AR-Vs at femtomole-level (10^–15 mol) sensitivity, revealing substantial heterogeneity in AR expression dynamics across cell lines. Protein translation peaks during G1 and G2 phases [35], yet the cell lines examined here were not synchronized, introducing additional variability even within the same model. The differences in AR-Nterm abundance among biological replicates, derived from distinct parental cell lines, highlight both methodological and biological heterogeneity. Moreover, because the number of AR molecules per cell is unknown, low expression may reflect genuine heterogeneity, where only a subset of cells may express specific AR-Vs at a given time. Nonetheless, this validated LC–MS/MS assay reliably quantified AR-Nterm and AR-V proteins at concentrations approaching the attomole (10^–18 mol) range.

One limitation of this assay is the peptide design constraints. Due to extensive sequence overlap among AR-Vs, only one unique peptide could be selected per AR-V, limiting redundancy and confidence typically afforded by multi-peptide approaches. Alternatively, the use of a different protease may generate additional measurable peptides, thereby increasing confidence in analyte identification and quantification. Furthermore, although extensive selectivity experiments were performed and rigorous measures were taken to minimize potential interferences, it is not possible to completely exclude all sources of interference. The comparisons performed in this study was based on a limited number of biological replicates to evaluate the technical variability and should be interpreted with caution. Future studies with larger cohorts will be required to provide a more precise estimates of the biological variability.

Therefore, this remains an important consideration when interpreting the results. Future detection of novel AR-Vs will likely require de novo assay development with AR-V-specific peptides that could be incorporated into the current method.

In summary, this work establishes a validated proteomic platform for AR-V detection and quantification, bridging the gap between transcript-level detection and protein-level evidence. The assay robustness, specificity, and sensitivity lay the foundation for future translational studies in PDXs and clinical cohorts. Expanding this platform to incorporate additional AR-Vs and correlating protein-level data with patient outcomes will be critical next step to define the role of AR-Vs in therapeutic resistance and for advancing precision treatment strategies in advanced PCa patients.

Method

Target peptide

Synthesis stable isotope-labeled (heavy) peptides on the C-terminal lysine (K) K (13C6, 5N2) of tryptic peptides, cysteine modification if cysteines were present, valine (V) (13C5, 5N1) for AR-V12 peptide, and the corresponding unlabeled (light) peptides were chemically synthesized by HeavyPeptide™ AQUA Ultimate ThermoFisher Scientific. A total of 8 crude peptides (including 4 light and 4 heavy peptide pairs) were synthesized, and the chemical purity was estimated to be > 97%. The peptide concentrations were balanced and equalized. The isotopic purity of each heavy amino acid was > 99%. MS1 analysis was performed to confirm the monoisotopic mass and the most abundant precursor charge states for each of the 4 light and 4 heavy peptides.

Cell culture

Human PCa cell lines 22Rv1, LNCaP, DU145 and PC3 were obtained from ATCC. As an AR-negative control for experiments, the Cercopithecus aethiops kidney fibroblast-derived COS-1 cell line was obtained from ATCC. 22Rv1, LNCaP, DU145, and PC3 and COS-1 cells were grown in RPMI 1640 Medium with no phenol red (Gibco, cat: 11,835,055) supplemented with 10% fetal bovine serum (FBS) (Sigma-Aldrich, cat: F0926), 1% GlutaMAX™ (ThermoFisher, cat: 35,050,061), and 1% penicillin–streptomycin (Gibco, cat: 15,140,122). VCaP-16, an enzalutamide resistant human PCa cell line, was obtained from Joshua Russo and Steve Balk at Beth Israel Deaconess Medical Center of Boston [36]. VCaP-16 cells were grown in Dulbecco’s Modified Eagle’s Medium (DMEM) with 4.5 g/L glucose, L-glutamine, sodium pyruvate, and phenol red (Corning, cat: 10–013-CV). This medium was supplemented with 10% FBS, 1% penicillin–streptomycin, and 1% GlutaMAX™. All cells were grown in a humidified incubator at 37°C and 5% CO₂, they also all had changes of media every three days. Before experimentation, cells were tested for mycoplasma contamination via PCR amplification. Cell lines were also validated by Promega PowerPlex16HS Assay at the University of Arizona Genetics core.

Cell line and PDX protein extraction sample preparation

Protein extraction from cell lines and PDX was adapted [37] with minor modifications. LuCaP PDX tumor 77 (LuCaP 77) and 77 castration resistance (LuCaP 77CR) from the LuCaP series from bone metastasis was used [38]. The PDX tumors were obtained as a small frozen tumor section that originated from subcutaneously implanted patient-derived advanced PCa from metastatic sites as described previously [38].

Approximately 25 mg of each tumor or 5.0 × 10^6 cells per cell line was processed as previously described[37]. The lysis buffer contained 7 mol/L urea, 2 mol/L thiourea, 0.4 mol/L Tris pH 8.0, 20% acetonitrile (ACN), 10 mmol/L TCEP, 25 mmol/L chloroacetamide, Thermo Scientific’s Halt protease inhibitor cocktail 1X concentration (originally 100X). After adding 500 µL of the lysis buffer to the tumor pieces or cell lines, samples were placed on ice and homogenized, then vortexed and centrifuged at 12,000 g for 10 min. The samples were then sonicated for 5 s using a probe sonicator set at 30% amplitude and kept on ice during the entire sonication process. After sonication, the samples were incubated for 0.5 h at 37°C, then at room temperature for 15 min to reduce and alkylate cysteines and centrifuged at 12,000 g for 10 min at 18 °C.

The total protein concentration from each sample was determined by a Bradford assay (Thermo Fisher Scientific, Waltham, MA) prior to digestion. We then used 1 mg of protein, added 1:250 of lysyl endopeptidase (WAKO, 125–05061) and incubated the samples at room temperature for 5 to 6 h at pH 7.4. Then adjusted pH to 7.5 to 8.0 using 1 mol/L Tris, pH 11. Then, we added Worthington TPCK-treated trypsin (1 mg/mL) dissolved in 1 mmol/L HCl supplemented with 20 mmol/L of CaCl₂ to prevent autolysis and added 1:50 and spiked in twice and incubated at 37°C for overnight digest for (approximately 18 h). Samples were diluted fivefold by adding 10 mmol/L Tris, pH 8.0 to dilute urea. After incubation, samples were acidified with TFA to pH 3 or less. Reverse-phase extraction method was done using hydrophilic–lipophilic balance (HLB) (Oasis). The peptide final yield was measured using the Colorimetric peptide assay (Thermo Fisher Scientific, cat no. 23275). Sample digestion efficiency prior to MS analysis was inspected by evaluating samples before and after enzymatic digestion using SDS-PAGE gels and again after MS analysis.

LC–MS/MS

Analysis was performed by LC–MS/MS using an MPX system from SCIEX (Framingham, MA, USA) with dual Shimadzu XR LC system Shimadzu Corporation (Kyoto, Japan) coupled with a SCIEX QTRAP 6500 + triple quadrupole detector SCIEX. For each run, 20 μL of sample digest was injected onto a 50 × 3 mm Kinetex 2.6 µm C18 100 Å column (Phenomenex, Torrance, CA, USA). Chromatography was carried out at 500 μL/min with a total run time of 7.5 min. The mobile phases comprised solvent A (0.1% formic acid in HPLC grade water) and solvent B (0.1% formic acid in HPLC grade acetonitrile). The chromatographic run began at 10% B and was held for 0.5 min, followed by a 4.5 min linear gradient to 30% B. It then transitioned to 95% B in 0.1 min, held for 0.9 min, returned to 5% B in 0.1 min, and was held at 5% for 1.4 min. The first 2.2 min of the LC run was sent to waste, and data were collected for the next 2.5 min, before again diverting the last 2.8 min of the LC run to waste. The MS/ MS source parameters were CAD: Medium, CUR: 40, GS1: 80, GS2: 85, IS: 5,500, TEM: 650, and EP: 10. Peptide collision energy was empirically measured and done manually on the instrument. Ion transitions were monitored using scheduled MRM with a 30 s MRM detection window based on retention time and a 0.22 s target scan time.

Limit of detection and limit of quantification

The concentration of the AR-Vs in cell lines and PDX tumors were determined by comparing the MRM area ratio of the AR-Vs peak area to the peak area of the internal heavy pair standard for the corresponding transition and referencing to an external calibration curve determined from these same area ratios. Calibration curves were determined using ordinary linear regression analysis with 1/x weighting in MultiQuant software (Version 3.0.2) SCIEX using the MQ4 algorithm. The limit of quantitation (LOQ) was determined empirically by analyzing the lowest working calibrator over 5 replicates and resulting in a percent coefficient of variation (CV) and the Upper limit of quantification was measured empirically until the highest level of the analytical measured ranged was < 20% CV.

Western blot analysis

Prostate cancer cell lines were lysed in 2% SDS with protease and phosphatase inhibitors and quantified by BCA (ThermoFisher) according to manufacturer recommendations. Equal amounts of protein (30 ug) were run on a 4–20% polyacrylamide gel (BioRad) and transferred onto PVDF 0.45µm membrane (Millipore), before being blocked in 5% BSA and incubated with primary antibody. Primary antibodies are N-terminal AR (Millipore SAB5500006), AR-V7 (Precision Antibody, AG10008), and Alpha-tubulin (CST 2144). Secondary antibodies were Licor IR-conjugated, and membranes were scanned using BioRad Chemi-Doc Imager.

Statistical analysis and raw data files

All data was analyzed using Graph pad prism 10 and raw data files 1–9 are in the supplemental data files. The raw targeted mass spectrometry data supporting the findings of this study have been deposited in the PASSEL repository (PeptideAtlas) under dataset identifier PASS05983 (dataset tag: ARV2026).

Abbreviations

22Rv1

22Rv1 is a human prostate carcinoma epithelial cell line derived from a xenograft that was serially propagated in mice.

ADT

Androgen deprivation therapy

AR

Androgen receptor

AR-exon3-Dup

Androgen receptor exon 3 duplication

AR-Nterm

AR N-terminal domain

AR-V12

AR splice variantVariant AR-V12

AR-V2

AR splice variant AR-V2

AR-V7

AR splice variant AR-V7

AR-Vs

Androgen receptor variants

ARPI

Androgen receptor pathway inhibitor

BLAST

Blast alignment sequencing tool

COS-1

An African green monkey kidney fibroblast-like cell line, commonly used in molecular biology for high-efficiency transient transfection and recombinant protein expression and it is androgen receptor negative expressing cell line

CPTAC

Clinical proteomics tumor analysis consortium

CR

Castrated

CRPC

Castration-resistant prostate cancer

CTC

Circulating tumor cells

CV

Coefficient of variation

DU145

DU 145 is a cell line with epithelial morphology that was isolated from the brain of a 69-year-old, White, male with prostate cancer

GPM

Global proteome machine

LC-MS/MS

Liquid chromatography-tandem mass spectrometry

LLOQ

Lower limit of quantification

LNCaP

LNCaP clone that was isolated in 1977 by J.S. Horoszewicz, et al., from a needle aspiration biopsy of the left supraclavicular lymph node of a 50-year-old, White, male (blood type B +) with confirmed diagnosis of metastatic prostate carcinoma

LOD

Limit of detection

MRM

Multiple reaction monitoring

NCR

Non-castrated

PC-3

Is a cell line initiated from a bone metastasis of a grade IV prostatic adenocarcinoma from a 62-year-old, White, male

PCa

Prostate cancer

PDX

Patient derived xenograft

PROPHECY

Prospective-blinded clinical trial that validated the use of circulating tumor cell CTC

PSA

Prostate-specific antigen

QC

Quality control

SRM

Selective reaction monitoring

ULOQ

Upper limit of quantification

VCaP16

Cell lines that were generated by long-term culturing of VCaP cells in androgen depleted media with Abiraterone, an AR antagonist

Author contributions

ZES designed and executed most of the data generated and analysis, wrote the main manuscript text and prepared Figs. 1–6, GL and MLL helped generate and collect data, SS helped set up instrumentation; JWR, and SB provided VCAP cell lines. EV, SP and SMD provided PDX samples. All authors reviewed the manuscript.

Funding

Zoi E Sychev, DOD grants and Targets of Cancer Training Program Fellowship, W81XWH-20–1-0402, T32 CA009138 (ZES).
Kiel Tiez, DOD PCRP Early Investigator Research Award, W81XWH-22–1-0098.
Steven Balk, P01 CA163227.
Eva Corey, Pacific Northwest Prostate Cancer SPORE (P50CA97186), the PO1 NIH grant (PO1CA163227), and the Institute for Prostate Cancer Research (IPCR).
Stephen Plymate, Lopker Foundation-SRP. Emmanuel S Antonarakis, NCI grant P30 CA077598 and US DOD grant W81XWH-22–2-0025.
Scott M Dehm, NIH Grants R01CA174777 and R01CA270539, DOD Grant HT94252510228, Prostate Cancer Foundation Challenge Award.
Jesse C Seegmiller, National Institutes of Health.
Justin M Drake, Masonic Cancer Center Tumor Working Group Award, Veterans Affairs Merit Award I01BX003324 (SRP).

Data availability

All processed data supporting the findings of this study are included in the article and its supplementary information files.

Declarations

Competing interests

The authors declare no competing interests.