Immunohistochemistry-based assessment of androgen receptor status and the AR-null phenotype in metastatic castrate resistant prostate cancer

Abstract

Background

Molecular and immunohistochemistry-based profiling of prostatic adenocarcinoma has revealed frequent Androgen Receptor (AR) gene and protein alterations in metastatic disease. This includes an AR-null non-neuroendocrine phenotype of metastatic castrate resistant prostate cancer which may be less sensitive to androgen receptor signaling inhibitors. This AR-null non-neuroendocrine phenotype is thought to be associated with TP53 and RB1 alterations. Herein, we have correlated molecular profiling of metastatic castrate resistant prostate cancer with AR/P53/RB immunohistochemistry and relevant clinical correlates.

Design

Twenty-seven cases of metastatic castrate resistant prostate cancer were evaluated using histopathologic examination to rule out neuroendocrine differentiation. A combination of a hybridization exon-capture next-generation sequencing-based assay (n = 26), fluorescence in situ hybridization for AR copy number status (n = 16), and immunohistochemistry for AR (n = 27), P53 (n = 24) and RB (n = 25) was used to profile these cases.

Results

Of 27 metastatic castrate resistant prostate cancer cases, 17 had AR amplification and showed positive nuclear expression of AR by immunohistochemistry. Nine cases lacked AR copy number alterations using next-generation sequencing/fluorescence in situ hybridization. A subset of these metastatic castrate resistant prostate cancer cases demonstrated the AR-null phenotype by immunohistochemistry (five cases and one additional case where next-generation sequencing failed). Common co-alterations in these cases involved the TP53, RB1, and PTEN genes and all these patients received prior therapy with androgen receptor signaling inhibitors (abiraterone and/or enzalutamide).

Conclusions

Our study suggests that AR immunohistochemistry may distinguish AR-null from AR-expressing cases in the metastatic setting. AR-null status informs clinical decision-making regarding continuation of therapy with androgen receptor signaling inhibitors and consideration of other treatment options. This might be a relevant and cost-effective diagnostic strategy when there is limited access and/or limited tumor material for molecular testing.

Introduction

Therapeutic options in the management of metastatic castration-resistant prostate cancer include chemotherapeutic agents (cabazitaxel, docetaxel), radionuclide therapy (223radium-dichloride), cellular immunotherapy (sipuleucel-T) as well as next-generation androgen receptor signaling inhibitors (ARSIs: enzalutamide and abiraterone) [1–4]. Given the more widespread use of ARSIs, there is an unmet need to better identify mCRPC patients who no longer clinically benefit from these agents, as well as to better understand mechanisms of resistance [4, 5].

The identification of biomarkers to clinically stratify mCRPC patients is an active area of research and in recent years has involved extensive molecular characterization of such cases [4, 6, 7]. A major barrier to translating current biomarker research findings into direct clinical practice involves access disparities to varied diagnostic testing methodologies along with the interpretation of the results. In this context, routine histopathology and immunohistochemistry (IHC) are widely available with the potential for immediate and widespread adoption. In advanced prostate cancer disease states such as mCRPC, prolonged AR pathway inhibition can alter the typical course of the disease, manifest by histological differentiation to a small cell/high-grade neuroendocrine phenotype in 11–17% of cases [7, 8]. Tumors that have been thus reprogrammed lose their dependence on AR signaling and instead rely on what have been termed AR indifferent compensatory mechanisms for growth and survival [9]. Therefore, documentation of small cell/high-grade neuroendocrine transformation, by morphologic evaluation and IHC, signals disease resistance and drives clinical decision-making with regards to further therapy [5]. It must, however, be noted that uniform definitions of small cell/high-grade neuroendocrine transformation are lacking. Furthermore, as mCRPC lesions are often not biopsied in routine clinical practice, the rarity of this specimen type has further hindered such standardization efforts. While morphologic/IHC evaluation can identify neuroendocrine transformation, the recognition of the recently described non-neuroendocrine AR indifferent disease in routine clinical practice remains a challenge [8].

In this study, we have studied biopsy samples from patients with non-neuroendocrine mCRPC by molecular profiling and IHC to identify biomarkers of resistance which indicate that an ARSI treatment is no longer benefitting the patient.

Materials and methods

Patient specimens

This study was approved by the institutional review board and involved analysis of molecular profiling data for all mCRPC samples profiled by an NGS-based assay, as part of an institutional clinical cancer genomics initiative [10–12]. The cBioPortal.32e34 platform was used to analyze data pertaining to molecular characterization of mCRPC [13]. Twenty-seven consecutive cases of mCRPC that were subjected to molecular profiling and had available material for downstream analysis (FISH/IHC) were selected for this study. Clinical history was reviewed to confirm castration-resistant disease status and all cases were evaluated in an algorithmic manner to exclude neuroendocrine differentiation as part of an institutional clinical practice (see Supplementary methods) [14].

Immunohistochemistry (IHC)

IHC was performed for P53 (clone: DO-7, Leica Biosystems, Buffalo Grove, IL, USA), RB (clone: 13A10, Leica Biosystems, Buffalo Grove, IL, USA), and AR (clone: AR441, Dako, Santa Clara, CA, USA). Validation controls for AR IHC included the PC3 cell line, which lacks AR expression (negative control), while the androgen-sensitive LNCaP cell line was used as a positive control.

Fluorescence in situ hybridization (FISH)

FISH analysis was performed on paraffin sections using a 2-color AR/CenX probe.

Next-generation sequencing (NGS): MSK-IMPACT

Details of the MSK-IMPACT assay have been previously reported (see Supplementary methods) [10–12, 15–17].

Statistical analysis

Continuous clinicopathological variables were analyzed with frequency counts and percentages. Tests to assess statistical significance, where applicable, were two-sided, with P < 0.05 considered to be statistically significant.

Results

AR, P53, and RB expression status in mCRPC

Twenty-seven cases of non-neuroendocrine mCRPC were profiled using a combination of NGS-based large-panel testing (MSK-IMPACT) and IHC (AR, P53, and RB). Some cases of mCRPC (9/26, 35%) did not reveal any copy number changes for the AR gene on chromosome X by either NGS/FISH-based copy number assessment, or loss-of-function alterations (Fig. 1a, b); of which five cases showed a complete absence of AR expression by IHC (Fig. 1c), referred to as the AR “null” phenotype. This AR “null” phenotype was also seen in an additional case where NGS failed due to low tumor content. None of the cases showed cytoplasmic AR expression, in the absence of nuclear expression. This contrasted with 17 cases of mCRPC that showed AR amplification using NGS-based copy number assessment and confirmed by FISH in all 12 cases that were tested (Fig. 1d, e). Importantly, all cases with AR amplifications showed robust nuclear protein expression by IHC (Fig. 1f). Although increased AR protein expression was observed in these cases, semiquantitative scoring was not attempted in this study. Unlike the AR “null” and AR “expressing” states which are highly reproducible phenotypes, in the absence of standardized digital image analysis, the quantification of AR expression based on intensity of staining in cases harboring amplifications is subjective and difficult to standardize.

Fig. 1. Copy number assessment and immunohistochemistry for AR.

Two cases of metastatic castrate resistant prostate cancer with either an AR-null phenotype (a–c) or AR-amplified status (d–f) have been depicted. Copy number plots generated using MSK-IMPACT are shown (a, d) and show relative (Log2) tumor/normal ratios (y-axis) and corresponding chromosomes (x-axis), with each blue dot representing an individual probe region. Probes at amplified loci have been depicted with red dots and the locus for AR (Xq11–12) has been highlighted within a black boxed region. Representative images for AR copy number assessment using fluorescence in situ hybridization (b, e) and AR protein expression status using immunohistochemistry (c, f) have been shown.

Additional IHC-based profiling of these cases involved evaluation of P53 and RB expression status. Previous studies have reported scattered nuclear positivity of P53 by IHC to correlate with TP53 wild-type status, while pathogenic alterations were correlated with either diffuse nuclear overexpression (herein, referred to as mutant pattern#1) or a complete absence of expression (herein, referred to as mutant pattern#2) and these patterns were identified in mCRPC, as well (Supplementary Fig. 1A–D) [18]. Similarly, loss-of-function alterations of the RB1 gene have been reported to be associated with loss of protein expression by IHC and this pattern was seen in mCRPC when truncating alterations of the RB1 gene were present (Supplementary Fig. 1E, F) [19]. Of note, due to the limited nature of the biopsy material, IHC for PTEN was not pursued.

Heterogeneity of AR expression in mCRPC

While variant allele frequencies obtained from NGS-based molecular profiling can be informative regarding the clonality of various alterations, IHC can be used to directly visualize different tumor populations. In routine clinical practice, immunophenotyping using IHC/flow cytometry is used for the assessment of clonality in hematopathology, however, the use of protein expression for similar purposes in solid tumors is limited. Inferences regarding clonality, albeit at a lower resolution, can be made based on spatially restricted, unique patterns of protein expression.

Two cases of mCRPC that exhibit both AR “expressing” and AR “null” populations have been illustrated to demonstrate this. For instance, a patient with mCRPC that progressed on ARSIs (Case 5) presented with cervical lymph node metastasis that lacked morphologic features of neuroendocrine differentiation (Supplementary Fig. 2A). IHC demonstrated a predominant AR “null” phenotype, combined with a mutant pattern of P53 expression and RB loss (Supplementary Fig. 2B–D); the latter two alterations suggesting the presence of pathogenic alterations of TP53 and RB1 genes. However, IHC for AR revealed the presence of a morphologically indistinguishable AR “expressing” clone in the background of an AR “null” phenotype (inset, Supplementary Fig. 2B).

A second case of mCRPC which progressed on ARSIs (abiraterone, Case 3) has been illustrated. Two clonal tumor populations without morphologic features of neuroendocrine differentiation were identified (Supplementary Fig. 3A, B) and were found to share a mutant pattern of P53 expression using IHC, consistent with an underlying TP53 p. A159V pathogenic alteration (Supplementary Fig. 3E, F). However, while one clone showed retained AR and RB expression by IHC (FISH: AR amplified; Supplementary Fig. 3C, G), the second clone revealed an AR “null” phenotype and RB loss (FISH/NGS: no AR copy number alterations; NGS: RB1 p. E492fs; Supplementary Fig. 3D, H), suggesting clonal divergence. These two cases highlight the utility of IHC in identifying separate spatially restricted tumor populations with either an AR “expressing” or “null” phenotype that cannot be reliably identified using either NGS or FISH.

Evolution/Selection of mCRPC based on AR expression status

Prior studies have suggested that in response to selective pressure exerted by newer ARSIs such as abiraterone and enzalutamide, a phenotypic shift occurs in mCRPC characterized by a transition from an AR “expressing” to an AR “null” phenotype [4]. One such case has been illustrated here (Case 6; Fig. 2). Specifically, multiple metastases have been shown documenting disease progression over time in a patient with prostatic adenocarcinoma and include: a lymph node metastasis at the time of radical prostatectomy (NGS: no AR copy number alterations; Fig. 2a, b), a subsequent lymph node metastasis representing a noncastrate disease state (NGS: AR amplification; Fig. 2c, d) and finally, a liver metastasis representing castration-resistant disease (NGS: no AR copy number alterations; Fig. 2e, f). IHC in this case demonstrates a transition from AR “expressing” metastases (Fig. 2a–d) to AR “null” mCRPC that lacks a neuroendocrine phenotype (Fig. 2e, f).

Fig. 2. Histopathology and immunohistochemistry: clonal evolution/selection.

Prostatic adenocarcinoma showing clonal evolution/selection over time, pertaining to AR status (Case 6). Histology (a, c, e) and AR immunohistochemistry (b, d, f) have been shown. Initial lymph node metastasis at radical prostatectomy (a, b) and subsequent noncastrate lymph node metastasis (c, d; documented AR amplification) showed robust AR expression by immunohistochemistry. On follow-up, a metastatic castrate resistant prostate cancer specimen demonstrated AR-null status by immunohistochemistry (e, f).

Molecular and IHC profiling of mCRPC based on AR expression status

Consistent with previous studies, our prior study had documented frequent alterations of AR (52%), TP53 (48%), PTEN (29%), and RB1 (18%) in mCRPC [6, 20–23]. In our study of mCRPC which were selected after excluding those with small cell/high-grade neuroendocrine differentiation, 26 cases were profiled using large-panel NGS (Fig. 3, Supplementary Fig. 4). The frequency of alterations involving these genes was similar: AR (17/26, 65.4%), TP53 (14/26, 53.8%), PTEN (7/26, 26.9%), and RB1 (7/26, 26.9%). While all cases with AR amplification were found to show strong nuclear expression of the protein using IHC, cases without AR copy number alterations were further stratified into AR “expressing” (4/9, 44%) and AR “null” categories (5/9, 56%) and furthermore, revealed clonal heterogeneity based on AR IHC status in two cases, as depicted earlier. All six cases of mCRPC that were found to have developed the AR “null” phenotype had received prior hormonal therapy, including ARSIs (Table 1). Moreover, except for Case 3 (which showed clonal heterogeneity for AR), all cases showed low to undetectable serum prostate-specific antigen (PSA) levels when AR “null” status was documented (Table 1). Case 6 is of particular interest as serum PSA levels closely mirrored AR status over the course of disease progression (no AR alteration at radical prostatectomy: 1.56 ng/mL, AR amplification on follow-up: 11.96 ng/mL, AR-null status: 0.61 ng/mL; Table 1).

Fig. 3. Molecular profiling and immunohistochemistry.

Immunohistochemistry and molecular profiling results for AR, TP53, and RB1 have been depicted for 27 cases of metastatic castrate resistant prostate cancer. This includes fluorescence in situ hybridization and next-generation sequencing-based assessment of AR copy number alterations. Cases with clonal heterogeneity on immunohistochemistry have been indicated with (smaller) orange boxes, while those with discordant immunohistochemistry and next-generation sequencing results have been highlighted with diamond-shaped boxes. No alterations were identified for additional members of the FGF/MAP Kinase signaling pathway (HRAS, NRAS, KRAS, RAF1, ARAF, BRAF, MAP2K1, MAP2K2, FGF3, FGF4, FGFR1, FGFR2, FGFR3, FGFR4) and other genes such as BRCA1.

Table 1.

Clinicopathologic features of AR “Null” prostatic adenocarcinoma.

Case No.	Initial diagnosis	Antiandrogen therapy and emergence of AR “Null” phenotype	Time from initial diagnosis to AR “Null” disease	Outcome post documentation of AR “Null” disease
1	65-year-old with (AR+) Gleason 5 + 5 = 10 (grade group 5) prostate cancer with pelvic/abdominal lymphadenopathy	- Antiandrogen therapy: degarelix, followed by abiraterone - Biopsy confirmation of castrate resistant metastatic disease - Chemotherapy: docetaxel, carboplatin - Biopsy confirmation of AR “null” castrate resistant liver metastasis (molecular alterations: case 1, Fig. 3) PSA (at diagnosis): 32.67 ng/mL PSA (at molecular testing): 0.38 ng/mL	16 months	Dead of disease: 6 months; total follow-up: 22 months
2	61-year-old diagnosed with Gleason 4 + 4 = 8 (grade group 4) prostate cancer	- Initial therapy: antiandrogens (leuprolide and bicalutamide), brachytherapy, external beam radiation therapy - Subsequent therapy: antiandrogens (leuprolide, degarelix, bicalutamide, and enzalutamide) - Biopsy confirmation of AR “null” castrate resistant lung metastasis (molecular alterations: case 2, Fig. 3) PSA (at diagnosis): 5.9 ng/mL PSA (at molecular testing): 1.28 ng/mL	141 months	Dead of disease: 28 months; total follow-up: 169 months
3	55-year-old diagnosed with Gleason 3 + 4 = 7 (grade group 2) prostate cancer	- Initial therapy: brachytherapy, external beam radiation therapy - Subsequent therapy: antiandrogens (bicalutamide, ketoconazole, abiraterone, leuprolide) - Transurethral resection of prostatic tissue showing AR “null” prostate cancer with clonal heterogeneity (molecular alterations: case 3, Fig. 3) - Further antiandrogen therapy: enzalutamide - Further chemotherapy: docetaxel, cabazitaxel, and mitoxantrone PSA (at diagnosis): 4.9 ng/mL PSA (at molecular testing): 6.32 ng/mL	166 months	Dead of disease: 33 months; total follow-up: 199 months
4	61-year-old diagnosed with Gleason 4 + 5 = 9 (grade group 5) prostate cancer	- Initial therapy: antiandrogens (leuprolide, abiraterone) - Subsequent therapy for progressive disease: brachytherapy, external beam radiation therapy - Biopsy confirmation of AR “null” castrate resistant lung metastasis (molecular alterations: case 4, Fig. 3) - Chemotherapy/immunotherapy: docetaxel, carboplatin, pembrolizumab PSA (at diagnosis): 7.09 ng/mL PSA (at molecular testing): <0.05 ng/mL	8 months	Alive with disease: 6 months; total follow-up: 14 months
5	65-year-old diagnosed with Gleason 4 + 5 = 9 (grade group 5) prostate cancer	- Initial therapy: antiandrogens (leuprolide and degarelix) - Subsequent therapy: antiandrogens (abiraterone) and chemotherapy (docetaxel, carboplatin) - Biopsy confirmation of AR “null” castrate resistant cervical lymph node metastasis with clonal heterogeneity (case 5, Fig. 3: molecular profiling failed as the extracted DNA did not meet the minimum quality/quantity criteria) PSA (at diagnosis): 13.9 ng/mL PSA (at molecular testing): <0.05 ng/mL	8 months	Alive with disease; no additional follow-up (total: 8 months)
6	49-year-old diagnosed with Gleason 3 + 3 = 6 (grade group 1) prostate cancer	- Initial therapy: external beam radiation therapy - Salvage radical prostatectomy: pT3aNl disease - Subsequent therapy: antiandrogens (bicalutamide, leuprolide ) - Biopsy confirmation of AR amplified pelvic lymph node metastasis - Subsequent therapy: docetaxel, external beam radiation therapy - Biopsy confirmation of AR “null” castrate resistant liver metastasis (molecular alterations: case 6, Fig. 3) - Subsequent therapy: antiandrogens (enzalutamide) PSA (at radical prostatectomy): 1.56 ng/mL PSA (at confirmation of AR amplification): 11.96 ng/mL PSA (at confirmation of AR-null status): 0.61 ng/mL	135 months	Alive with disease; no additional follow-up (total: 135 months)

IHC for RB and P53 identified several cases with discrepant results when compared with molecular profiling (RB: 2/24, 8%; P53: 5/23, 22%). For instance, while IHC for RB correctly identified 6/7 cases with loss-of-function alterations, two cases showed discrepancies. Similarly, IHC for P53 correctly identified 11/14 cases with TP53 alterations and five discordant cases were identified. This included two cases with wild-type TP53 status and complete loss of P53 by IHC, while three cases with TP53 alterations (missense ×2 and truncating ×1) showed a wild-type staining pattern.

As former studies have suggested that AR “null” non-neuroendocrine mCRPC may show enrichment for FGF and MAPK pathway alterations, molecular profiling for these six cases was evaluated for similar alterations as well as for common alterations seen in prostatic adenocarcinomas (Fig. 3). A single case (of five) revealed a deep deletion of MAP3K1, however, no alterations were identified for other members of these pathways including HRAS, NRAS, KRAS, RAF1, ARAF, BRAF, MAP2K1, MAP2K2, FGF3, FGF4, FGFR1, FGFR2, FGFR3, and FGFR4.

Discussion

mCRPC is defined as prostate cancer progression manifesting as metastatic disease despite low levels of serum testosterone (<50 ng/dL) [24]. Unlike testosterone-driven AR signaling in the noncastrate setting, AR signaling in castration-resistant disease states involves varied pathogenic mechanisms such as amplification of the AR gene/enhancer as well as constitutively active AR signaling [6, 24, 25]. Clinically, monitoring AR-signaling driven PSA expression in the serum serves as a surrogate for assessing sustained AR signaling [4]. Next-generation ARSIs such as enzalutamide and abiraterone are being increasingly used in mCRPC due to their ability to offset continued AR signaling [24].

Currently, clinical evaluation for adverse prognostic indicators in mCRPC involves either histopathologic/immunohistochemical assessment of biopsy specimens or molecular analysis of biopsy material or circulating tumor cells. Some adverse prognostic indicators include small cell/high-grade neuroendocrine transformation on histopathologic/IHC evaluation, alterations such as AR splice variant 7, those involving TP53, RB1, DNA damage response genes, and AR/PI3K pathways on molecular profiling [7, 26–29]. Although obtaining biopsy material for downstream analysis is challenging, particularly from bone metastasis, when available, morphologic and IHC-based evaluation is routinely performed to confirm a diagnosis of metastatic prostate cancer [6]. In addition to confirming the diagnosis, an algorithmic approach is utilized to rule out small cell/high-grade neuroendocrine transformation using morphology-based screening, followed by IHC-based confirmation using markers such as chromogranin and synaptophysin. Documentation of transformation to small cell/high-grade neuroendocrine carcinoma represents evolution to AR-signaling independence (Fig. 4) and provides a rationale for changing therapy [4].

Fig. 4. Schematic representation of hypothesized prostate cancer evolution.

Androgen deprivation therapy (ADT) in prostatic adenocarcinoma with androgen signaling inhibitors (ARSIs), such as abiraterone and enzalutamide, provides a selection pressure that leads to a survival advantage for androgen insensitive clones. This leads to an intermediate state with clonal heterogeneity and ultimately progression to androgen insensitivity manifesting as either high-grade neuroendocrine carcinoma or (non-neuroendocrine) AR-null disease status.

In a recent study, up to 75% of the AR independent neuroendocrine phenotype has been reported to show AR amplification/protein expression and this observation requires independent validation in subsequent studies [8]. This phenotype (AR positive, neuroendocrine) is potentially an intermediate disease state that eventually progresses to complete AR independent disease. On the other hand, selective pressures exerted by hormonal therapies have also been hypothesized to lead to the development of a non-neuroendocrine AR-null phenotype (Fig. 4) [4]. The prevalence of this phenotype has been estimated to have almost quadrupled over the course of the last decade, mirroring the increased adoption of next-generation ARSIs in clinical practice [4, 30–32]. Our results suggest that after excluding mCRPC with small cell/high-grade neuroendocrine transformation, up to 60% (six of ten cases) of cases that lack AR amplification may represent the AR-null phenotype. Prognostically, both the presence of tumor heterogeneity and an AR “null” state have been associated with poor outcomes [33, 34].

In our experience, AR-null mCRPC are morphologically indistinguishable from AR-expressing non-neuroendocrine tumors. However, incorporation of AR IHC in routine clinical practice can reliably identify these cases and can impact clinical decision-making pertaining to scaling back therapy with ARSIs and consideration of other treatment options. The lack of AR protein expression likely occurs secondary to transcriptional regulation and therefore DNA-based evaluation of either copy number, mutational, or structural variant status may be misleading. Although RNA-seq/gene expression profiling-based techniques may identify these cases, such methodologies have significant barriers to routine and widespread clinical implementation both in terms of access/cost as well as in defining algorithms/thresholds to define AR-null status. In addition, although DNA/RNA-based sequencing strategies may fail when a limited amount of tumor is present in biopsy material, a single 4-μm section of tissue can be successfully used to define AR-null status. In either case, sequencing techniques are unlikely to be informative in the presence of heterogeneity in terms of the coexistence of AR-expressing and AR-null clones. Therefore, IHC is likely to be the “gold standard” for the identification of such cases.

Furthermore, our results highlight challenges in the implementation of similar strategies aimed at using P53/RB IHC to predict pathogenic alterations of these genes. A significant number of cases may represent either false positive or false negative results, particularly with regards to TP53 status, where nonpathogenic missense single-nucleotide variants do not necessarily correlate with IHC. In the case of Rb, IHC may not reliably identify subclonal alterations present at low variant allele fractions.

In summary, AR IHC is a CLIA (Clinical Laboratory Improvement Amendments)-validated, robust, cost-effective and easy to access assay. The binary interpretation patterns (absent or positive) are likely to be highly reproducible between laboratories. This lends itself to utilization in routine clinical practice and clinical trials with implications for scaling back therapy with ARSIs and consideration of alternate treatment options in refractory metastatic castration-resistant prostate cancer.