Focal p53 protein expression and lymphovascular invasion in primary prostate tumors predict metastatic progression

TP53 is one of the most frequently altered genes in prostate cancer. The precise assessment of its focal alterations in primary tumors by immunohistochemistry (IHC) has significantly enhanced its prognosis. p53 protein expression and lymphovascular invasion (LVI) were evaluated for predicting metastatic progression by IHC staining of representative whole-mounted prostate sections from a cohort of 189 radical prostatectomy patients with up to 20 years of clinical follow-up. Kaplan–Meier survival curves were used to examine time to distant metastasis (DM) as a function of p53 expression and LVI status. TP53 targeted sequencing was performed in ten tumors with the highest expression of p53 staining. Nearly half (49.8%) of prostate tumors examined showed focal p53 expression while 26.6% showed evidence of LVI. p53(+) tumors had higher pathologic T stage, Grade Group, Nuclear Grade, and more frequent LVI. p53 expression of > 5% and LVI, individually and jointly, are associated with poorer DM-free survival. TP53 mutations were detected in seven of ten tumors sequenced. Four tumors with the highest p53 expression harbored likely pathogenic or pathogenic mutations. High levels of p53 expression suggest the likelihood of pathogenic TP53 alterations and, together with LVI status, could enhance early prognostication of prostate cancer progression.

www.nature.com/scientificreports/ of primary tumors carry TP53 missense, frameshift, or truncation mutations, and at least 1% have homozygous deletions 7,8 . Genomic analysis of non-indolent localized PCa revealed TP53 to be one of six genes with > 2% somatic single nucleotide variants (SNVs) 9 . The higher frequency of TP53 lesions in localized cancers suggests that they arise relatively early in disease progression. In advanced PCa, the rate of TP53 mutations becomes significantly enriched, approaching 40% SNVs, and 10% homozygous deletions or genomic rearrangements [10][11][12] . TP53 mutations that increase the stability and half-life of mutant proteins in cancer cells and enhance protein detection by IHC 4,13,14 characterize a subgroup of biologically aggressive prostate cancers with high risk of progression after prostatectomy. Multiple studies have reported a correlation between IHC detection of p53 and PCa progression [14][15][16][17][18][19][20] . DNA sequencing of p53 positive (p53(+)) prostate tumors from 16 patients by Griewe et al., found a 69% correlation between p53 expression and TP53 mutation 21 . Schlomm et al., reported a low frequency of p53(+) tumors (2.5% or 62/2514) by IHC in a tissue microarray from RP specimens, of which 47% (29/62) were found to harbor mutations associated with more aggressive disease 18 . In another screen of two overlapping RP patient cohorts with primary prostate tumors, Guedes et al. reported a high positive predictive value (84%) of p53 nuclear staining for underlying TP53 missense mutation 4 . Importantly, in a single-patient longitudinal study, p53(+) metastatic lesions that developed years post-surgery could be traced to a low-grade p53(+) tumor focus in the primary tumor 22 . These findings emphasize the biological impact of focal TP53 alterations in the clonal progression of PCa and support p53 IHC detection in primary PCa as a surrogate indicator of TP53 missense mutations.
In PCa, lymphovascular invasion (LVI) has been shown to be associated with aggressive disease and poor prognosis, as defined by reduced biochemical recurrence (BCR) progression-free survival [23][24][25] , increased risk of PCa-specific mortality 26,27 and other pathologic features of aggressive disease 28,29 . LVI has been evaluated with either TP53 mutation or p53 expression in association with gastric 30 , colorectal 31 , bladder 32 and breast cancer 33 prognosis, but not PCa. This study examines the role of p53 protein expression and LVI in predicting distant metastasis (DM) in a RP cohort with long-term follow-up. We further explored the combined effect of p53 expression and LVI status on DM-free survival. To determine if tumors with higher p53 expression also harbor TP53 mutations, targeted TP53 sequencing was performed on a subset of prostate tumors with the highest percent of p53 expression.

Methods
Study design, population, and clinical assessment. Prostate specimens and clinical-pathologic data were collected from patients undergoing treatment at the Walter Reed National Military Medical Center (WRNMMC) from 1993 to 2013 who provided written informed consent for the use of all data and biospecimens obtained. Patients who had biopsy positive, organ-confined PCa and underwent RP as primary treatment (≤ 6 months post-diagnosis) were included, and those who underwent neo-adjuvant hormonal therapy were excluded. Archived, whole-mounted RP specimens from 50 patients who developed DM at least one year following diagnosis and from 139 patients without evidence of BCR or DM after at least 10 years follow-up, were analyzed. The presence of distant metastases was ascertained by the review of each patient's complete radiographic scan history that included bone scan, computed tomography (CT), positron emission tomography (PET), as well as pelvic and bone magnetic resonance imaging (MRI). Subjects who reached the end of the study period without evidence of PCa metastases, had their last known follow-up, or died without evidence of metastasis before the end of the study period (December 31, 2013), were defined as non-metastatic. This work was approved by the Institutional Review Boards of WRNMMC, the Uniformed Services University of the Health Sciences (USU), and the Joint Pathology Center (JPC) (Protocol number DBS.2020.110).
Immunohistochemistry and pathologic assessment. Preparation and histologic evaluation of whole-mounted RP specimens were performed as previously described 34,35 . Adjacent, four-micron sections from a representative tissue block containing the index tumor were stained with hematoxylin and eosin (H&E), anti-p53 mouse monoclonal antibody (DO-7, Biocare Medical, Pacheco, CA), and anti-podoplanin antibody (D2-40, Biocare Medical) to identify p53 and lymphatic vessels, respectively. Slides were reviewed using the 2014 International Society of Urological Pathology (ISUP) guidelines 36 by a single genitourinary pathologist (I.A.S.), who was blinded to clinical outcomes. The p53 status in index tumors was scored as positive based on the detection of nuclear p53 staining, percent area stained, and staining intensity. Cells were recorded as p53 positive or p53(+), when brown chromogen (3,3'-Diaminobenzidine (DAB)) used to stain the DO-7 antibody was detected in the nuclei of any tumor cells, and as negative or p53(−) in the absence of any nuclear staining. Occasional tumor cells with exclusive cytoplasmic staining of any intensity were considered "negative". Percentage of p53(+) staining was estimated as the area of p53(+) tumor cells with nuclear staining divided by total index tumor area 16 , which was categorized as 0%, 1-5%, and > 5% p53(+) expression. p53 staining intensity was also quantified as 1 + (light), 2 + (medium), and 3 + (maximum) intensity 37 . An independent pathologist review of p53 staining was performed by A.P.B. Findings were presented as percentage of p53 expression. LVI status was recorded as positive or LVI(+), when tumor cells were present within spaces lined by lymphovascular endothelium with characteristic podoplanin staining, and as negative or LVI(−) in the absence of any staining.

Statistical analysis.
Overall and p53-stratified (0%, 1-5%, > 5%) distributions for patient demographics, as well as clinical and pathologic features were compared using Student's T-test for continuous variables and Chi-square and ANOVA tests for categorical variables. Fisher's exact test was used when > 20% of expected cell counts had less than five observations. Unadjusted Kaplan-Meier (KM) estimation curves were used to examine time to DM as a function of p53 status. Log-rank test and its associated p-value are reported for KM models. Associations of p53 and LVI with DM-free survival were first evaluated independently and then jointly. www.nature.com/scientificreports/ variable Cox Proportional Hazards analysis was used to model DM-free survival, controlling for demographic and pathologic factors. The assumption of proportional hazards was tested and confirmed for all KM and Cox models. All statistical tests were 2-sided (summary α-error = 0.05), and the decision rule was based on p < 0.05. All statistical analyses were performed using SAS version 9.4.

TP53 mutation analysis.
Index tumors were scraped from two adjacent whole-mounted FFPE sections derived from ten cases that were selected for targeted sequencing. Library preparations and sequencing reactions were conducted at GENEWIZ, Inc. (South Plainfield, NJ). Gene-specific primers targeting the TP53 CDS were multiplexed into three pools. A sequencing library was prepared using the NEBNext Ultra DNA Library Preparation Kit (New England Biolabs, Ipswich, MA), validated using an Agilent TapeStation (Agilent Technologies, Palo Alto, CA), and quantified by Qubit (Invitrogen, Thermo Scientific, Waltham, MA) and real-time PCR (Applied Biosystems, Carlsbad, CA). Multiplexed DNA libraries were loaded on an Illumina MiSeq instrument (Illumina, San Diego, CA) for 2 × 150 bp paired-end sequencing. Image analysis and base calling were performed using MiSeq Control. Raw reads were aligned to the GRCh37 human reference genome using Burrows-Wheeler Aligner-mem. Samtools fixmate was used to correct any flaws in read-pairing introduced during alignment, and duplicate reads were marked using Picard. Alignments were subjected to base quality score recalibration, according to GATK best practices. Variants were identified using GATK Haplotype Caller and FreeBayes and annotated using the Ensembl Variant Effect Predictor toolset that included Sanger Catalogue of Somatic Mutations In Cancer (COSMIC) and ClinVar annotations from March 2021.
Ethics approval and consent to participate. These prostate specimens and clinical-pathologic data in this study were collected from patients undergoing treatment at the WRNNMMC who provided written informed consent for their use.
Multivariable Cox proportional hazard models predict distant metastasis-free survival. Multivariable Cox Proportional hazards analysis was used to examine independent and joint roles of p53 expression and LVI status, together with patient's age at RP and race, on DM-free survival (Table 3). Three approaches were used: (1) Model One shows a strong correlation between p53 expression of > 5% and increased risk of DM, which increases the hazard for this event by three-fold (hazard ratio (HR) = 3.173; p = 0.0006).
(2) Model Two shows that both > 5% p53 expression and LVI(+) are independent predictors of shorter time to DM. When other covariates were held constant, the hazard for DM increased by two-fold (HR = 2.224; p = 0.0225) for > 5% p53 expression, and by four-fold (HR = 4.053; p < 0.0001) for LVI(+) status (3) In Model Three, the joint analysis of p53 expression and LVI status showed that p53 expression of > 5% and LVI(+) status confer incremental risk for DM: the hazard for this event increased by 4.4-fold-fold (HR = 4.428; p = 0.0091) based on > 5% p53 expression alone, and by almost five-fold (HR = 4.839; p < 0.0001) based on LVI(+) status alone, but together they increased the risk by almost eight-fold (HR = 7.976; p < 0.0001). This suggests a strong mutual and additive impact of higher p53 expression and LVI(+) status on DM-free survival. By contrast, in all three models, after adjusting for the patient's race and p53 expression levels or LVI(+) status, an additional year of age at RP was shown to induce hazards of DM only by a factor of 1.03 to 1.04 (or 3-4%). Thus, increasing age at RP contributes little to the differ- www.nature.com/scientificreports/ ence in the risk of DM. Likewise, after adjusting for the patient's age and p53 expression levels, the patient's race had no significant effect on the hazard of DM. The poorest DM-free survival outcomes were observed among patients who exhibited both > 5% p53(+) and LVI(+) status. In all models, significant correlations between GG and pT with p53 and LVI status prevented their inclusion in multivariate models. Moreover, too few patients were observed in lower GG (1-2) and stage categories.
Association between TP53 mutations and p53 expression or LVI status. Tumor specimens of ten patients with the highest p53 staining (20% to 90%) were selected for targeted TP53 sequencing. The high-depth coverage achieved by TP53 targeted sequencing allowed SNVs to be detected at relatively higher alternate allele frequencies of 0.11 to 0.51. Almost all patients sequenced for TP53 developed aggressive disease, represented by GG 4 or 5, pT3, or DM (Table 4). At least one missense or nonsense TP53 mutation was detected in seven patients, and two mutations were detected in one patient. TP53 mutations were detected in all four patients who had both high (> 30%) expression of p53 and LVI(+) status. Interestingly, in all four patients who had both high expression of p53 and LVI(+) status, the TP53 mutations detected were either likely pathogenic or pathogenic alterations, which were also among the most recurrent TP53 mutations in the COSMIC database. In agreement with results showing an association of LVI(+) status with poorer DM-free survival ( Table 2 and Fig. 3B), all five subjects sequenced who were LVI(+) further developed DM. The number of cases sequenced, however, were too small to indicate any association of LVI positivity with specific mutational status.

Discussion
In this study, p53 expression and LVI status were examined as key independent predictors of DM. High p53 expression was significantly associated with DM, the frequency of which was three-fold higher in patients with > 5% p53(+) compared to patients with 0-5% p53(+). By stratifying the data at > 5% cut-off, we were able to distinguish between two clinically relevant p53(+) populations: patients with > 5% p53(+) have significantly shorter DM-free survival than those with 0-5% p53(+). Likewise, the presence of LVI is associated with higher frequency of DM. LVI(+) patients developed DM at a rate that was four-fold higher than those without LVI. Further analysis by unadjusted univariable KM further confirmed that LVI(+) status significantly predicts poorer www.nature.com/scientificreports/ DM-free survival. Subsequent combined examination of p53 expression and LVI by multivariable analyses showed that together, they exerted an additive increase in risk for DM. Although multiple studies have shown an association between p53 expression and TP53 mutation 4,18,21,22 , inconsistencies were noted by others. These discrepancies could be attributed to limitations of the IHC assay, including antibodies used for detection 38,39 , or to study cohort selection 40 . The focality of TP53 alterations in primary PCa can lead to differences in IHC interpretations or DNA sequencing assays 4,16,21 . Since p53 IHC detection depends on the increased half-life of mutant proteins, proteins with destabilizing mutations may escape detection 41 . The lower frequency of TP53 mutations in localized prostate cancers could reduce the likelihood for finding an association with increased p53 expression 18,40 . Since p53 nuclear accumulation is far more frequent in higher grade carcinomas, performing IHC on all primary prostate cancers at diagnosis is unlikely to establish the expected association 4 . By contrast, this study is designed to focus on defining the association of p53 expression and LVI status with DM. Hence, the proportion of subjects with advanced stage (pT3-4) and grade (GG 4 & 5) is greater in this cohort than subjects who undergo RP without neoadjuvant therapy in the general population. One advantage of this study is the availability of primary PCa specimens with associated long-term follow-up (median = 13 years) data obtained from an equal-access military treatment facility. The greater proportion of patients with aggressive disease, who undergo RP without neo-adjuvant therapy in this cohort compared to patients in the general population allowed us to demonstrate the striking association between both focal p53(+) expression and LVI(+) status, and the development of DM. Furthermore, the use of whole-mounted prostate sections augmented the comprehensive evaluation of p53 expression and LVI in index tumors compared to using tissue microarrays 18,42 or biopsy specimens 38,43 . Lastly, the concordant scores of percent p53 expression between two independent pathologists further validated that this approach was more reproducible than by staining intensity alone (92% vs. 72%, respectively). Table 1. Patient demographic and clinico-pathologic features distributed across categories of percent p53 expression (N = 189). Significant values (P-value < 0.05) are in bold. † Three subjects who were neither Caucasian nor African American race were excluded. § Two missing subject data due to treatment effect. || One missing subject data due to capsular incision on whole-mount specimen; appropriate staging not possible. www.nature.com/scientificreports/ Based on earlier reports that p53 positive tumors were likely due to mutations that increased the half-life of the p53 protein, we hypothesized that tumors with the highest percentage p53 expression would have a higher probability of harboring TP53 mutations. To test this notion, we selected ten cases with the highest percentage of p53 expression for targeted TP53 sequencing. TP53 mutations were detected in seven of ten cases analyzed. These mutations, which include the most recurrent hotspot at Arginine 273, were previously reported in advanced or metastatic PCa and annotated in COSMIC 44 and ClinVar 45 databases. Four patients harboring likely pathogenic or pathogenic TP53 mutations had aggressive disease represented by GG 5, pT3 and LVI(+) tumors that progressed to DM. Consistent with earlier reports, concordance between p53 staining and the presence of pathogenic TP53 mutations further supports the prognostic utility of IHC detection as a surrogate read-out for TP53 mutations 4,21 .
Although LVI is known to be associated with aggressive disease and poor prognosis in PCa, no direct comparison to TP53 mutations or its protein expression has been performed [23][24][25][26] . One study reported ERG(+) tumors had higher LVI and lower p53 expression in ERG(+) tumors, but no significant association was detected due to sample size (N = 51) 46 . The most striking finding of this study was that 81.8% of patients with both > 5% p53(+) expression and LVI(+) status developed DM after an extended follow-up period, while 87.7% patients with LVI(−) and < 5% p53(+) were DM-free. The joint interpretation of these two variables is underscored by three key findings: (1) p53(+) expression of > 5% strongly correlates with LVI(+) status, (2) multivariate analysis suggests p53 expression and LVI status to be additive, and (3) tumors with lower p53 expression of 1-5% may represent less aggressive disease as these patients have DM-free survival outcomes that are remarkably similar to patients without p53 staining. www.nature.com/scientificreports/

Conclusion
Our findings validated the association between pathogenic TP53 mutations and higher p53 expression, which support the IHC staining of p53 as a substitute for detecting TP53 mutations. Primary prostate tumors with combined focal p53(+) of > 5% and LVI(+) status are highly predictive of future DM and should be classified as highly aggressive tumors. This subset of patients may require a more rigorous treatment plan and follow-up protocol. Taken together, determination of p53 expression and LVI status in primary PCa has promising potential to improve prognostication and early prevention of metastatic progression.