Analytic, Pre-analytic and Clinical Validation of p53 Immunohistochemistry for Detection of TP53 Missense Mutation in Prostate Cancer

Abstract

PURPOSE

TP53 missense mutations may help to identify prostate cancer (PCa) with lethal potential. Here, we pre-analytically, analytically and clinically validated a robust immunohistochemistry (IHC) assay to detect subclonal and focal TP53 missense mutations in PCa.

EXPERIMENTAL DESIGN

The p53 IHC assay was performed in a CLIA-accredited laboratory on the Ventana Benchmark immunostaining system. p53 protein nuclear accumulation was defined as any p53 nuclear labeling in >10% of tumor cells. 54 formalin fixed paraffin embedded (FFPE) cell lines from the NCI-60 panel and 103 FFPE PCa tissues (88 primary adenocarcinomas, 15 metastases) with known TP53 mutation status were studied. DU145 and VCaP xenografts were subjected to varying fixation conditions to investigate the effects of pre-analytic variables. Clinical validation was performed in two partially overlapping radical prostatectomy (RP) cohorts.

RESULTS

p53 nuclear accumulation by IHC was 100% sensitive for detection of TP53 missense mutations in the NCI-60 panel (25/25 missense mutations correctly identified). Lack of p53 nuclear accumulation was 86% (25/29) specific for absence of TP53 missense mutation. In FFPE prostate tumors, the positive predictive value (PPV) of p53 nuclear accumulation for underlying missense mutation was 84% (38/45), while the negative predictive value (NPV) was 97% (56/58). In a cohort of men who experienced biochemical recurrence after RP, the multivariable hazard ratio for metastasis among cases with p53 nuclear accumulation compared to those without was 2.55 (95% CI: 1.1–5.91).

CONCLUSIONS

IHC is widely available method to assess for the presence of deleterious and heterogeneous TP53 missense mutations in clinical PCa specimens.

Introduction

Both older and more recent sequencing studies of castrate resistant prostate cancer (CRPC) metastases have revealed a high frequency of TP53 mutations in advanced disease, making these genomic alterations among the most enriched when comparing primary and advanced metastatic prostate cancer (1–4). Though TP53 aberrations are relatively rare in primary tumors, occurring in less than 10% (5), there is a 4–5-fold increase in alterations in CRPC (4) and TP53 and RB1 mutations are considered key drivers of small cell neuroendocrine carcinomas in the prostate and other organs (6, 7). The reason for this striking and unanticipated enrichment of TP53 loss in aggressive prostate cancer is currently being scrutinized by a number of groups. In two recent studies, combined loss of TP53 and RB1 facilitated prostate cell lineage plasticity and anti-androgen resistance in vitro and in vivo, in part due to downstream SOX2 activation (8, 9), however these mechanistic studies focused on genomic deletion of TP53, rather than mutation. Interestingly, in both primary and metastatic prostate tumors, TP53 alterations are relatively evenly split between those resulting in pure loss-of-function for the tumor suppressor protein (nonsense, frameshift, splice site mutations and homozygous deletion) and missense mutations which result in potential gain-of-function, loss-of-function, and/or dominant negative phenotypes (4, 5). The differing cellular effects of these disparate types of alterations in TP53 remains a hotly debated topic in cancer (10) and whether each has different associations with prostate cancer clinical outcomes remains unknown. Further, though TP53 is thought to be a relatively early occurring mutation in primary prostate cancer (11–14), the chronicity of this alteration with respect to other common genomic changes in prostate cancer, and its potential heterogeneity in primary tumors has not been elucidated.

Underlying these knowledge gaps in prostate cancer is the fact that few validated assays exist to query TP53 status in prostate cancer. Though sequencing will clearly remain a gold-standard technique, recent work has shown that TP53 mutations can be heterogeneous and extremely focal in primary prostate tumors, requiring ultra-deep sequencing to detect (12). Thus, a less expensive in situ assay to rapidly assess TP53 status in large portions of tissue would be useful, both for routine clinical screening of primary tumors where alterations may be focal, and to begin to address some of the biological questions outlined above. Importantly, missense mutations in TP53 generally occur in a few hotspots in the DNA binding domain of the protein, and have long been recognized to result in stabilization and resultant nuclear accumulation of the p53 protein in tumor cells (15–18). Though a complete mechanistic understanding has been elusive (19, 20), this phenomenon is useful since p53 nuclear accumulation by immunohistochemistry (IHC) can be a surrogate to predict TP53 gene status in situ (21). p53 IHC assays have been used previously in other tumor types (22, 23) and prostate cancer (2, 24–31) for some time to screen for TP53 missense mutations, however genetically validated assays for use in prostate cancer have not been developed using contemporary, automated and clinical-grade IHC platforms.

Here, we extensively analytically, pre-analytically and clinically validated a clinical-grade IHC assay to screen for presence of stabilizing TP53 missense mutations in prostate cancer in a CLIA-accredited laboratory environment. We show that this assay is highly sensitive and specific for detection of TP53 missense mutations in cell lines and primary prostate tumors, the assay is robust to pre-analytic fixation conditions and could be of particular use to identify heterogeneous, subclonal TP53 aberrations in the setting of prostate biopsies that may be difficult to detect without ultradeep sequencing. Finally, we show that TP53 status, as determined by this assay, is associated with risk of metastasis in two large cohorts of surgically-treated prostate cancer patients.

Materials and Methods

Patients and tissue samples

In accordance with the US Common Rule and after IRB approval with a waiver of consent, a total of 103 prostate tumors with known TP53 sequencing results (including 90 primary tumors at radical prostatectomy, transurethral resection of prostate or needle biopsy and 13 autopsy specimens) from several different cohorts were utilized for the validation study. The first group consisted of 28 cases from a radical prostatectomy case-cohort design (described below as used for outcomes analysis) (32, 33). These samples were immunostained for p53 on tissue microarrays (TMAs) (see below) and 28 cases (including 4 negative for p53 nuclear accumulation and 24 positive for nuclear accumulation) were selected for sequencing as described below after examination of immunostaining for p53. The second cohort included metastatic tissue from 13 autopsy cases sampled on TMA for immunostaining, for which previously published TP53 sequencing data was available (11, 34). The third cohort included 4 small cell carcinoma of the prostate cases sampled on TMA for which previously collected TP53 sequencing data was available (6). The fourth cohort included 2 radical prostatectomy cases (standard slides) with known homozygous deletions of TP53 from prior studies (35). Finally, the last cohort included 56 radical prostatectomy cases from patients who subsequently developed castration resistant metastatic disease and received enzalutamide and/or abiraterone at Johns Hopkins between 2010 and 2015. These 56 patients included all patients with a radical prostatectomy at Johns Hopkins and available tissue for sequencing, selected from a retrospective analysis of 309 consecutive patients treated with first-line abiraterone or enzalutamide at Johns Hopkins during this time period. (Maughan et al; p53 Status in Primary Tumor Predicts Efficacy of First-Line Abiraterone and Enzalutamide in Castration-Resistant Prostate Cancer Patients; under review). For these cases, all tumors were sequenced irrespective of p53 immunostaining status. The details of each cohort are included in Supplementary Table S1.

To assess the frequency of p53 nuclear accumulation by IHC in prostate needle biopsies, we examined p53 IHC in a cohort of 40 Gleason score 9 or higher needle biopsies described previously (36).

For outcomes analysis, we selected a total of 456 primary prostate tumors from two overlapping and previously published radical prostatectomy cohorts at Johns Hopkins (including 28 tumors with sequencing data as described above) (32, 33). One cohort included 322 patients using a nested case-cohort design for metastasis, where the time to clinical metastasis was measured from the time of diagnosis. An additional and highly overlapping cohort (~55% overlap) contained 195 men who all experienced biochemical recurrence (defined as a PSA measurement of ≥ 0.2 ng/ml on two separate occasions) after radical prostatectomy. Here, time to metastasis was measured from time of biochemical recurrence. These cohorts were originally designed to test for prognostic tumor biomarkers and were accordingly highly enriched for adverse oncologic outcomes. Tumor tissue from the dominant tumor nodule and benign tissue were sampled in quadruplicate on 16 individual TMAs utilizing 0.6 mm cores.

Cell line TMA

56 cell lines from the NCI-60 cell line panel (Developmental Therapeutics Program, NCI) were used to evaluate p53 IHC staining. All cell lines were pelleted, fixed in 10% neutral buffered formalin and processed and cut as tissue. Cell lines were punched and tissue microarrays created as previously described (37). Additional prostate cell lines with known TP53 gene status were also studied, including RWPE-1, VCaP, DU145, and PC-3 (American Type Culture Collection, Manassas, VA). RWPE-1 cells have wild-type TP53. VCaP and PC3 cells harbor one deleted allele of TP53 and one missense mutation or frameshift mutation, respectively (R428W and A138fs) (38, 39). DU145 cells harbor missense mutations on both alleles (V274F and P223L) (38). For all prostate cancer cell lines, cell line authenticity and mycoplasma contamination were routinely assessed in 6- to 10-month intervals by STR genotyping and PCR based testing, respectively (last tested December 2015). For the remaining NCI-60 cell lines, STR genotyping was completed once prior to creation of the cell line TMA.

Xenograft TMA

VCaP and Du145 cell line xenografts were established as previously described (40) and subjected to different fixation conditions after excision to evaluate the effect of pre-analytic variables on the performance of the p53 IHC stain. After harvesting, the xenograft tumors were sectioned into three fragments and then divided in 5 groups, each subjected to varying intervals of cold ischemia at room temperature (0, 2, 4, 12 or 24 hours) prior to fixation. After the appropriate period of cold ischemia elapsed, the thirds of the tumor specimen with up to 4 hours of ischemia were put into 10% Neutral Buffered Formalin and fixed for 4, 12 or 24 hours. The thirds of the specimen with periods of ischemia longer than 12 hours were left in fixation for 12, 24, 36 or 48 hours intervals. The processed samples were arranged in a TMA and stained for p53 as below.

Detection of p53 nuclear accumulation by immunohistochemistry

p53 IHC was performed on the Ventana Benchmark autostaining system using a mouse monoclonal antibody (BP53-11) after antigen retrieval in CC1 buffer followed by detection with the iView HRP system (Roche/Ventana Medical Systems, Tucson, AZ). This protocol was previously validated for detection of p53 mutations in ovarian carcinoma (22) and the BP53-11 clone was tested in prostate and showed very low background staining in benign glands in comparison to another available clone from Ventana (D07), facilitating detection of nuclear p53 accumulation. Each tissue microarray spot containing tumor cells was visually dichotomously scored for presence or absence of nuclear p53 signal by a urologic pathologist blinded to the gene expression data (TLL). A spot was considered to show p53 nuclear accumulation if >10% of tumor nuclei showed p53 positivity. This cutoff value was chosen because it was the most correlated with presence of wild-type TP53 in a prior genomic validation of this staining protocol in ovarian carcinoma (22) and the most commonly chosen cutoff in a meta-analysis of p53 staining in other tumor types (23). A tumor was considered to show p53 nuclear accumulation if any sampled spot was scored as p53-positive, and as p53 negative if all sampled spots were scored as p53 negative.

TP53 sequencing

For samples from the TMAs, a total of five 0.6 μm punches were procured from the same area in the paraffin block sampled on the TMA. DNA was extracted from these FFPE cores using the Qiagen AllPrep DNA extraction kit (Hilden, Germany) according to the manufacturer’s directions. DNA concentrations were quantified with the Qubit fluorometer, using a Quant-iT dsDNA High Sensitivity Assay Kit (Invitrogen, Carlsbad, CA). Mutations were screened by next-generation sequencing (NGS) using multiplexed PCR (AmpliSeq Hotspot Panel) to generate libraries. Adapters were ligated to the PCR products, where the sequences are tagged with a specific barcodes. The barcoded libraries were clonally amplified using emulsion PCR (emPCR). The emPCR product was purified using magnetic bead purification followed by semiconductor-based sequencing on an Ion Torrent PGM (ThermoFisher, Waltham, MA). For each gene, the minimum required coverage was 500 sequence reads based on bidirectional sequencing. The minimum acceptable allelic frequency was 5%, as any lower values may be background noise. Each variant was analyzed manually using variant caller from Ion Torrent software and cross-referenced with Ingenuity^™ software for bioinformatics (Qiagen). The amplicons included for sequencing in the Ampliseq Hotspot panel included 91% (68/75) of the missense mutations detected in the AACR Project GENIE data for prostate cancer (41), though 7 of these detectable mutations (9%) were only detectable at the end of the amplicon so would require sequencing validation for reporting by a CLIA laboratory.

Disruptive versus non-disruptive TP53 mutation categorization

As defined by Poeta et al (42, 43), disruptive mutations include (i) any mutation that originates a stop codon (nonsense, frameshift, and intronic), (ii) missense mutations located inside the L2 or L3 loops (codons 163–195 or 236–251) replacing one residue by another of a different polarity or charge and, (iii), in-frame deletions within the L2 or L3 loops. Non-disruptive mutations are all those not classified as disruptive and include (i) missense mutations and in-frame deletions located outside the L2–L3 loops and (ii) missense mutations within the L2–L3 loops but replacing one residue with another of the same polarity or charge. Disruptive TP53 mutations lead to complete or almost complete predicted loss of activity of the p53 protein, in contrast to non-disruptive mutations that may retain some p53 function or have gain-of-function properties. The Poeta et al criteria were used to categorize all discovered TP53 mutations in FFPE samples by disruptive versus non-disruptive status and this classification is indicated in Supplementary Table S3.

Statistical analysis

The significance level was 0.05 for all statistical tests and analyses were performed in R v3.1 (R Foundation, Vienna, Austria). Primary endpoint of the study was defined as metastases evidenced by axial imaging (CT or MRI) or nuclear medicine bone scan. To determine the association of clinicopathologic variables with p53 status, Fisher’s exact test was used. For the intermediate/high risk cohort, time to event was defined as time from radical prostatectomy to last follow-up or metastasis. For the biochemical recurrence (BCR) cohort, time to event was defined as time from BCR to last follow-up or metastasis. Due to the case-cohort nature of our study, univariable and multivariable Cox proportional hazards model using the Lin-Ying method was used to evaluate the performance of p53 status in predicting metastasis (44, 45). Cumulative incidence curves were constructed using Fine-Gray competing risks (where deaths from other causes was considered as a competing risk) analysis with appropriate weighting of the controls to account for the case-cohort study design (44, 45).

Results

p53 nuclear accumulation by immunohistochemistry correlates with presence of underlying TP53 missense mutation in NCI-60 cancer cell line panel

To begin to determine the sensitivity and specificity of the p53 IHC assay, we first examined a previously described TMA composed of 54 of the 56 cell lines from the NCI-60 cell line panel (IGROV-1 and CAKI-1 were missing). In the cell line TMA, p53 protein nuclear accumulation was 100% sensitive for detection of deleterious TP53 missense mutations, detecting 25 of 25 missense mutations correctly (Figure 1A, Table 1a and 2, Supplementary Table S2). The specificity of the IHC assay was 86%, with lack of p53 nuclear accumulation in 25 of 29 cell lines that lacked missense mutations (ie, wild-type TP53 or loss-of-function mutations in TP53). Of the 4 discordant cell lines, two had possible frameshift mutations in TP53 (COLO-205 and SF-539) with additional missense mutations also reported in some studies. Both of these cell lines were classified as having inconclusive TP53 status (http://p53.free.fr/Database/Cancer_cell_lines/p53_cell_lines.html). The other two discordant lines with p53 nuclear accumulation had reported wild-type TP53 status (UO-31) and a reported splice-site mutation (OVCAR-8, p.Y126_splice) (46). The positive predictive power (PPV) of p53 IHC for detecting TP53 missense mutation was 86% (25/29) and the negative predictive power (NPV) of lack of p53 nuclear accumulation for lack of underlying TP53 mutation was 100% (25/25).

Figure 1. Representative p53 immunostaining in formalin fixed paraffin embedded prostate cell lines and primary and castrate resistant prostate cancer (CRPC) prostate tumors.

(A) RWPE-1 cells, which are TP53 wild-type (WT) show rare positive nuclei comprising less than 10% of the tumor on p53 immunostaining. In contrast, DU145 cells (with two mutations, p.V274 and p.P223L) and VCaP cells (p.R248W) with missense mutations show robust nuclear accumulation of p53. PC3 cells, which have a loss-of-function frameshift mutation (p.A138fs) and loss of heterozygosity (LOH) are entirely negative for p53 expression. All photomicrographs are reduced from 200×. (B) Only rare, weakly positive nuclear immunostaining for p53 is seen in benign prostate tissues and primary and CRPC tumors that are TP53 wild-type (WT), though some cytoplasmic staining is seen specifically in these tumors of unknown significance. The lack of nuclear positivity in benign prostate tissue and TP53 WT tumors meant that the IHC assay cannot distinguish tumors that are TP53 WT from those with loss-of-function alterations (homozygous deletion, frameshift, splice site or nonsense mutations). In contrast, tumors with missense mutations of TP53 (p.R175H, p.R273C, p.V157A, p.131N) are readily distinguishable from WT and show robust nuclear accumulation of p53 protein in the majority of cells. All photomicrographs are reduced from 200×.

Table 1a.

Correlation between p53 nuclear accumulation by immunohistochemistry (IHC) and presence of TP53 missense mutation in NCI-60 cell line panel

	p53 nuclear accumulation	Normal p53 IHC
TP53 missense mutation	25	0
Absence of TP53 missense mutation	4	25

Table 2.

Performance characteristics of p53 immunohistochemistry compared to TP53 sequencing results

	sensitivity	specificity	PPV	NPV
NCI-60 cell lines	100% (25/25)	86% (25/29)	86% (25/29)	100% (25/25)
FFPE prostate tumors	95% (38/40)	89% (56/63)	84% (38/45)	97% (56/58)

Of note, in cell lines, complete lack of p53 expression by IHC generally did distinguish tumors with loss-of-function mutations and/or homozygous deletion in TP53 (n=8) from those with wild-type TP53, which showed nuclear staining for p53 in fewer than 10% of cells (n=13) (Figure 1A, compare RWPE-1 and PC3 cells) (Supplementary Table S2). However, there were a few exceptions to this observation. Two cell lines (CAKI-1 and OVCAR-5) showed complete absence of p53 by IHC despite a reported wild-type TP53 locus (46). In addition, two other cell lines (MOLT-4 and SN12C) showed some weak nuclear positivity for p53 despite the presence of reported underlying nonsense mutations (p.306* and p.E336*, respectively) (46).

p53 nuclear accumulation by immunohistochemistry is highly associated with presence of underlying TP53 missense mutation in formalin-fixed paraffin embedded (FFPE) prostate tumors

The sensitivity and specificity of the p53 IHC assay were then evaluated in 103 FFPE prostate tumors (88 primaries and 15 metastases) with paired DNA sequencing and IHC results in FFPE material. Importantly, for 101 of 108 tumors, p53 immunostaining was initially evaluated on TMA spots, while the remainder were evaluated on standard histologic sections. The sensitivity of p53 IHC nuclear accumulation was 95% for missense mutation detection, with IHC nuclear accumulation detecting 38 out of 40 missense mutations (Figure 1B, Table 1b and 2, Supplementary Table S3). Using IHC, one of two missed mutations was p.R282W in a radical prostatectomy specimen where tissue preservation appeared to be poor (Supplementary Figure S1). Of interest, that same mutation was present and successfully detected by IHC in a separate, unrelated sample, where the p53 immunostaining showed heterogeneous p53 over-expression (Supplementary Figure S2). The other discordant case was an unusual case with 8 individual missense mutations in a single radical prostatectomy specimen, including p.R267W, G226D, M160I, E349K, E339K, A129T, E11K and V10I (Supplementary Figure S1, Supplementary Table S2). This tumor was Gleason Score 3+5=8 (using ISUP 2005 criteria) and pT3a and occurring in a 59 year old white male. The tumor itself did not look morphologically unusual, though did have a very high grade tumor component. It is tempting to speculate that it could be a hypermutated tumor, however it was MSH2 intact by IHC as well. We also cannot rule out the possibility of a sequencing artifact in this case. Indeed, only one other sample had more than one TP53 mutation detected (with 4 total, also a radical prostatectomy). Most mutations detected by the IHC assay were in the DNA binding domain of p53 as expected (Supplementary Figure S3).

Table 1b.

Correlation between p53 nuclear accumulation by immunohistochemistry (IHC) and presence of TP53 missense mutation in formalin fixed paraffin embedded prostate tumors

	p53 nuclear accumulation	Normal p53 IHC
TP53 missense mutation	38	2
Absence of TP53 missense mutation	7	56

Taken together, the specificity of the IHC assay was 89%, with 56/63 tumors lacking TP53 missense mutations by sequencing also lacking p53 nuclear accumulation by IHC. Of the seven discordant cases lacking TP53 missense mutations, but showing nuclear accumulation of p53 on IHC, 5 had wildtype TP53 by sequencing and 2 had in-frame deletions in the DNA binding domain (p.I255delI, and p.V203_D207del15) (Supplementary Figure S1, Supplementary Table S3). Of the 5 cases that showed p53 nuclear accumulation with wild-type TP53, three tumors had p53 immunostaining re-examined on standard sections to evaluate whether tumor heterogeneity could account for discordance. Of note, in all of these cases, heterogeneity of p53 expression could be observed, raising the possibility of focal TP53 mutations in a tumor subclone that may not have been detected by the sequencing assays (Figure 2). Though we cannot entirely exclude the possibility that the Ampliseq sequencing assay failed to detect underlying missense mutations in these cases due to its hotspot design, it is notable that this assay was designed to thoroughly cover the TP53 gene, capable of detecting 91% of the missense mutations detected in the AACR Project GENIE prostate cancer sequencing database (41). Overall, the negative predictive value (NPV) of lack of IHC nuclear accumulation for lack of TP53 missense mutation was 97% (56/58) and the positive predictive value of p53 nuclear accumulation for presence of underlying TP53 mutation was 84% (38/45). In contrast to our observations in cell lines above, complete lack of p53 expression by IHC did not distinguish tumors with loss-of-function mutations and/or homozygous deletion in TP53 (n=12) from those with wild-type TP53 (n=44) since the basal level of nuclear p53 protein expression in most wild-type prostate tumors was extremely low to undetectable (Figure 1, Supplementary Figure S1).

Figure 2. Heterogeneous p53 immunostaining on standard histologic sections was common in primary tumors with discordant TP53 sequencing and p53 immunostaining results.

(A) A tumor that was TP53 WT by sequencing shows focal p53 nuclear accumulation on standard section (arrow, upper left) with an adjacent area lacking nuclear accumulation (arrow, lower right). (B) A tumor that was TP53 WT by sequencing because a p.C135Y mutation was detected in <5% of cells, shows focal p53 nuclear accumulation on standard section (arrow, upper right) with an adjacent area lacking nuclear accumulation (arrow, lower left). (C and D) Gleason score 9 biopsies with clear focal nuclear accumulation of p53 in some, but not all tumor cells (arrows). All photomicrographs are reduced from 200×.

Previous work has categorized TP53 mutations as “disruptive” versus “non-disruptive” (42, 43), Disruptive TP53 mutations lead to complete or almost complete predicted loss of activity of the p53 protein, in contrast to non-disruptive mutations that may retain some p53 function or have gain-of-function properties. Interestingly, the majority of the non-detected alterations by p53 IHC accumulation were disruptive (80% or 8/10) (Supplementary Table S3), while the majority of the detected mutations were non-disruptive by these criteria (76% or 19/25), suggesting that our assay is biased towards mutations that may have gain-of-function properties (as has been previously described for many of these missense mutations).

Frequency of p53 nuclear accumulation among Gleason score 9 prostate biopsies

Finally, because the biopsy setting is likely the most clinically useful context to apply the p53 IHC assay, we assessed the frequency of p53 nuclear accumulation among a cohort of 40 high risk biopsies with Gleason score 9 or higher tumor. Of the 40 tumors selected for IHC, 39 (97.5%) had sufficient tumor for analysis. Of these, 13% (5/39) had p53 nuclear accumulation present. Interestingly, of the positive cases, 80% (4/5) had focal nuclear accumulation at biopsy (Figure 2C, 2D).

p53 nuclear accumulation by immunohistochemistry is minimally affected by different pre-analytic variables of delay of fixation or time in fixation

VCaP and Du145 cell line xenografts, which both have missense mutation of TP53 (38, 39) and nuclear accumulation of p53 protein, were used to validate the robustness of the p53 IHC assay to pre-analytic variables, including time of fixation and cold ischemia time. Time in formalin fixation did not markedly affect intensity of nuclear accumulation of p53 on IHC in DU145 xenografts, though minimal decrease in intensity was noted for VCaP xenografts. Similarly, cold ischemia time did not appreciably affect intensity of the p53 IHC signal or number of cells staining in either xenograft, though increased fixation delay did result in apparent tissue degeneration and autolysis and increased background signaling in both lines (Figure 3).

Figure 3. Effect of pre-analytic tissue fixation conditions (cold ischemia time and time in fixation) on p53 immunostaining in two xenograft models with known TP53 missense mutations.

(A) Effect of variation of fixation duration (4, 12 and 24 hours in 10% neutral buffered formalin with immediate fixation) on p53 immunostaining in DU145 and VCaP xenografts. Minimal variation in p53 nuclear accumulation is seen in DU145 cells (left panels), while VCaP cells show a slight decrease in intensity of staining with longer fixation conditions (right panels), though number of cells staining is not markedly affected. (B) Effect of variation of cold ischemia duration (4, 12, 24 hours without fixation at room temperature with 24 hour fixation in 10% neutral buffered formalin) on p53 immunostaining in DU145 and VCaP xenografts. Minimal variation in intensity of p53 nuclear accumulation is seen in DU145 cells (left panels) or VCaP cells (right panels) with variation of cold ischemia time, though tissue quality clearly decreases with increased autolysis and background staining compared to 0 hours old ischemia time (compare with last row of both panels in A).

p53 nuclear accumulation by immunohistochemistry correlates with risk of metastasis in two cohorts of surgically treated patients

Having fully analytically validated the p53 IHC assay, we next sought to clinically validate it. We assessed whether p53 nuclear accumulation by IHC is associated with metastatic progression in two partially over-lapping and previously reported radical prostatectomy cohorts with clinical follow-up (32, 33). The first cohort included 322 patients in a nested case-cohort design for metastasis. Demographic and clinical-pathologic characteristics of the cohort are reported in Supplementary Table S4. This cohort included 27 tumors with p53 nuclear accumulation (8%). Because the immunostains were performed on TMAs, we first examined whether tumor sampling in TMA format might miss p53 nuclear accumulation in comparison to immunostaining on standard slides. We immunostained 49 cases from the cohort that lacked p53 nuclear accumulation on TMA spots, examining the same dominant nodule that was punched for the TMA using standard sections. None of the cases showed focal p53 nuclear accumulation in standard sections, suggesting that TMA sampling is reasonably sensitive to detect p53 nuclear accumulation and suitable for using in outcome studies. In the TMA cohort, p53 nuclear accumulation was highly enriched among high Gleason score tumors compared to lower Gleason score tumors, with nuclear accumulation in 20% of Gleason score 9–10 tumors (p<0.001). In univariable analyses, Gleason score, positive surgical margin status, tumor stage and p53 IHC status (Figure 4A) were all associated with risk of metastasis, while patient age and pre-operative PSA were not significantly associated with outcome (Table 3). The univariable hazard ratio for metastasis among cases with p53 nuclear accumulation compared to those without was 4.84 (95% CI: 2.44–9.61). In multivariable models, only Gleason score, surgical margin status, seminal vesicle invasion and lymph node involvement remained significantly associated with risk of metastasis. The hazard ratio for metastasis among men with p53 nuclear accumulation compared to those without was 1.87, however this trend did not reach statistical significance (95% CI: 0.83–4.23; p=0.13) (Table 3).

Figure 4. Cumulative incidence of metastasis in patients from intermediate and high risk prostate cancer case-cohort study.

(A) or post-biochemical recurrence case-cohort study (B), stratified by p53 immunostaining status. Note that the case-cohort design makes a random sampling of the cohort and then is enriched with all the metastatic patients (not selected randomly). Thus, to get the full cohort, the controls are re-weighted by the inverse of sampling fraction (32, 33).

Table 3.

Univariable (UVA) and multivariable (MVA) models of association of p53 and clinicopathologic risk factors with metastasis in intermediate/high risk radical prostatectomy cohort

Variables	UVA		MVA
	Hazard Ratio (95% CI)	p-value	Hazard Ratio (95% CI)	p-value
Patient Age at RP	0.99 (0.96–1.02)	0.52	1.00 (0.96–1.04)	0.93
Log2 Preoperative PSA (ng/ml)	1.23 (0.94–1.61)	0.13	1.10 (0.84–1.42)	0.49
RP Gleason Score ≤3+4	ref	1	ref	1
RP Gleason Score 4+3	1.90 (0.89–4.04)	0.1	1.81 (0.83–3.95)	0.14
RP Gleason Score 8	4.70 (2.24–9.89)	<0.001	3.09 (1.26–7.56)	0.01
RP Gleason Score ≥9	13.00 (7.19–23.51)	<0.001	7.12 (3.82–13.29)	<0.001
Positive Surgical Margins	2.08 (1.34–3.24)	0.001	1.79 (1.07–2.99)	0.03
Extraprostatic Extension	3.95 (2.17–7.19)	<0.001	1.43 (0.72–2.86)	0.31
Seminal Vesicle Invasion	7.96 (5.04–12.58)	<0.001	3.32 (1.91–5.77)	<0.001
Lymph Node Invasion	6.66 (4.12–10.75)	<0.001	2.63 (1.43–4.83)	0.002
p53 Normal	ref	1	ref	1
p53 Nuclear Accumulation	4.84 (2.44–9.61)	<0.001	1.87 (0.83–4.23)	0.13

We next examined the association of p53 IHC status with metastasis in a partially over-lapping cohort of 195 men who underwent radical prostatectomy followed by biochemical recurrence (Supplementary Table S5). In this cohort, p53 nuclear accumulation was present in 9% (17/195) of cases, and was again enriched among Gleason score 9–10 tumors (seen in 20%) compared to lower Gleason score tumors. In this cohort, univariable analyses showed that Gleason score, seminal vesicle invasion, lymph node involvement and p53 IHC status (Figure 4B) were associated with risk of metastasis, while patient age, pre-operative PSA, extraprostatic extension and surgical margins were not significantly associated with outcome (Table 4). The univariable hazard ratio for metastasis among cases with p53 nuclear accumulation compared to those without was 4.14 (95% CI: 2.41–7.11). In multivariable analyses, only Gleason score, seminal vesicle invasion and p53 nuclear accumulation remained significantly associated with risk of metastasis. The multivariable hazard ratio for metastasis among cases with p53 nuclear accumulation compared to those without was 2.55 (95% CI: 1.1–5.91; p=0.03) (Table 4).

Table 4.

Univariable (UVA) and multivariable (MVA) models of association of p53 and clinicopathologic risk factors for prediction of metastasis in post-biochemical recurrence cohort

Variables	UVA		MVA
	Hazard Ratio (95% CI)	p-value	Hazard Ratio (95% CI)	p-value
Patient Age at RP	0.99 (0.95–1.02)	0.47	0.99 (0.95–1.03)	0.55
Log2 Preoperative PSA (ng/ml)	1.05 (0.83–1.32)	0.68	0.97 (0.76–1.23)	0.79
RP Gleason Score ≤3+4	ref	1	ref	1
RP Gleason Score 4+3	2.41 (1.18–4.94)	0.02	2.43 (1.15–5.13)	0.02
RP Gleason Score 8	4.27 (2.33–7.81)	<0.001	4.13 (2.24–7.59)	<0.001
RP Gleason Score ≥9	4.84 (2.78–8.41)	<0.001	3.99 (2.27–7)	<0.001
Positive Surgical Margins	0.73 (0.46–1.16)	0.18	0.71 (0.45–1.11)	0.14
Extraprostatic Extension	1.72 (0.93–3.18)	0.09	0.87 (0.45–1.7)	0.68
Seminal Vesicle Invasion	2.91 (1.87–4.53)	<0.001	2.2 (1.35–3.57)	0.001
Lymph Node Invasion	1.78 (1.12–2.81)	0.01	0.95 (0.55–1.63)	0.86
p53 Normal	ref	1	ref	1
p53 Nuclear Accumulation	4.14 (2.41–7.11)	<0.001	2.55 (1.1–5.91)	0.03

Discussion

Herein, we have extensively pre-analytically, analytically and clinically validated a clinical-grade and automated immunohistochemistry assay for identifying the presence of deleterious TP53 missense mutations in primary prostate tumors. In one of the largest studies in prostate cancer to compare an IHC assay with next generation sequencing data, we found that p53 IHC is 95% sensitive and nearly 90% specific for TP53 missense mutation detection compared to TP53 sequencing in over 100 prostate tumors. Importantly, the IHC assay is fairly robust to pre-analytic variables such as cold ischemia time and tissue fixation duration. For clinical validation of the assay, we show that nuclear accumulation of p53 protein is robustly associated with poor outcomes in two overlapping clinical cohorts of surgically-treated primary prostate tumors. Given the high sensitivity and specificity of the IHC assay with respect to sequencing, we propose that p53 IHC is a cost-effective and widely available method to screen primary prostate tumors for prediction of aggressiveness and outcome and is feasible to perform in high risk prostate biopsies. In addition, IHC provides an invaluable in situ technology to screen for focal TP53 mutations. Recent ultra-deep sequencing and clonal evolution studies have suggested that TP53 mutations, though generally occurring early in the primary tumor (11–14), may be quite focal and heterogeneous in primary prostate tumors (12, 13, 47). Indeed, in the current study, we have also observed frequent heterogeneity of p53 nuclear accumulation in as many as 40% of tumors. Though tumor heterogeneity is a challenge for any genomic technology, IHC is perfectly suited to screen large areas of tumor for focal alterations at a low cost, and thus would be quite useful in this regard in heterogeneous tumors.

The TP53 gene is altered in nearly 50% of human tumors overall, giving it the dubious distinction of being the most frequently mutated gene in human cancer (48). However, compared to other epithelial tumors, TP53 mutation is actually relatively infrequent in prostate cancer, with the prevalence hovering around 7% of contemporary surgical cohorts such as the TCGA cohort, with an additional 1% or so of tumors harboring homozygous deletions involving TP53 (5). The low rate of TP53 alteration in prostate cancer may be a reflection of the generally lower mutational burden seen in this tumor type compared to other carcinomas. Interestingly, however, in recent sequencing efforts in castration resistant prostate cancer (CRPC), the rate of TP53 mutations was dramatically increased compared to what was observed in the primary setting, with up to 40% of CRPC tumors showing TP53 mutations and another 10% or so showing homozygous deletions or genomic rearrangements involving TP53 (4). This significant enrichment of TP53 alterations in CRPC compared to primary tumors was rivaled only by that seen with alterations in the androgen receptor (AR), where genomic alterations are directly and heavily selected for after the imposition of hormonal therapies. Given that TP53 mutations generally occur early in tumor progression, occurring in the primary tumor before clonal divergence in the metastatic setting, these data suggest that primary tumors with TP53 alterations very likely have an unusually aggressive clinical course, which could potentially be predicted from an early timepoint in clinical care.

Of interest, in both the primary and metastatic settings, TP53 mutations are fairly evenly split in prostate cancer between missense mutations and nonsense/frameshift/indel mutations (4, 5). This is in stark contrast to other carcinomas, where missense mutations are more heavily favored, comprising over three quarters of catalogued mutations, compared to nonsense/frameshift/indel alterations (49). Although it has been debated for some time, it is increasingly clear in other tumor types that TP53 missense mutations may be a fairly unique class of alteration associated with gain-of-function, rather than simple loss-of-function or dominant negative phenotypes as was originally thought (10, 50–53). Indeed, even the subtype of missense mutation may confer specific transcriptional or other gain-of-function activities (10, 54). However, this remains to be proven for prostate cancer, and it is currently unknown whether the clinical outcomes of patients with TP53 nonsense/frameshift/indel alterations differ significantly from those with missense mutations. p53 missense mutation, as detected by the IHC assay described in the current study, is clearly associated with adverse outcomes in surgical cohorts of prostate cancer patients. Whether TP53 loss-of-function alterations (such as nonsense/frameshift/indel mutations that are not detected by the current IHC assay) have a similar prognosis is unclear. Future studies will examine these questions in cohorts with extensive clinical follow-up.

One clear limitation of the IHC assay described herein is its inability to detect nonsense/frameshift/indel alterations in TP53, likely due to the relatively low overall expression of p53 protein in benign prostate glands and those tumors lacking missense mutations. Of interest, the same assay did successfully detect mutations resulting in loss of protein in ovarian carcinomas, perhaps because the endogenous expression of wild-type p53 protein in the ovary is relatively higher than that seen in the prostate, enabling more sensitive detection of protein loss (22). Along the same lines, we also found that cell lines seemed to have relatively high basal levels of p53 protein expression, enabling us to successfully distinguish those with nonsense/frameshift/indel alterations or homozygous deletions from those with wild-type TP53 in most cases. However, prostate tumor tissue showed low basal expression of p53, making it impossible to detect loss of the protein in intermingled tumor cells. With this limitation in mind, it will be of particular interest to determine whether tumors with p53 protein loss due to nonsense/frameshift/indel alterations or homozygous deletions, have similarly poor outcomes as tumors with TP53 missense mutations which are detected by our IHC assay. If so, then it may be preferable to do next generation sequencing on prostate tumors, either alone or after IHC screening to rule out or enrich for focal alterations. If not, then IHC alone may be a convenient and inexpensive assay that is easily deployed to community pathology practices to test for clinically significant and heterogeneousTP53 alterations in primary prostate tumors.

Our objective in the current study was to analytically, pre-analytically and clinically validate p53 nuclear accumulation as a prognostic biomarker in prostate cancer. Additional large cohorts and future studies are required to examine the clinical utility of p53 nuclear accumulation and these studies will discern whether the IHC assay described herein adds substantially to current clinical-pathologic parameters for prognostication in primary prostate cancer. Because p53 nuclear accumulation was far more frequent in high grade (Gleason score 9–10) carcinomas compared to all other grades, performing p53 immunostaining on all primary prostate cancers at diagnosis is unlikely to be efficient. Rather, we would propose that evaluation by p53 IHC be limited to high grade, high risk tumors at diagnosis. In this group, it will be important to evaluate in future studies and clinical trials whether p53 nuclear accumulation predicts duration of response to hormonal therapies and/or a greater risk of neuroendocrine transdifferentiation, adding to current prognostic algorithms.

Supplementary Material

6. Supplementary Figure S1: Representative examples of primary tumors with discordant TP53 sequencing and p53 immunostaining results on tissue microarray.

Two tumors with TP53 missense mutations (upper left and upper right) did not show appreciable nuclear accumulation of TP53. Note that the mutation in the tumor on the upper left (p.R282W) was detected in another sample from a different patient by IHC and the tumor on the upper right had 8 separate TP53 missense mutations detected, of unknown significance. Two additional tumors without missense mutations (one WT by sequencing on lower left, and one with an in-frame deletion on lower right) both showed nuclear accumulation of p53 protein by IHC in >10% of cells and were considered false positives. All photomicrographs are reduced from 200×.

7. Supplementary Figure S2: p53 immunohistochemistry in radical prostatectomy sample detects p.R282W mutation.

Note that this mutations was missed in a separate, poorly preserved tumor sample (see Figure 3). Reduced from 200× magnification.

8. Supplementary Figure S3: Location of the mutations/deletions in TP53 gene detected by sequencing stratified p53 immunostaining result.

(A) Location of the TP53 missense mutations that were detected by p53 immunostaining, showing a higher occurrence at hotspots in the DNA binding domain. (B) Location of the missense mutations, frameshifts, nonsense and splice site variants that were not detected by p53 IHC.

Statement of Translational Relevance.

TP53 mutations are among the most highly enriched genomic alterations among castrate resistant prostate cancers (CRPC) compared to primary tumors and recent preclinical data suggests that TP53 inactivation may cooperate with other alterations to confer lineage plasticity and androgen independence in the prostate. We validated an in situ assay to sensitively and specifically detect TP53 missense mutations in prostate cancer using immunohistochemistry (IHC). This assay, which is robust to fixation conditions, is particularly useful to screen for focal or subclonal mutations that could be missed by sequencing and which were present in nearly half of all primary tumors. In two overlapping radical prostatectomy cohorts, presence of TP53 missense mutation by IHC was associated with development of metastasis, clinically validating the assay for use in high risk prostate cancer or future clinical trials.