Increased EZH2 expression in prostate cancer is associated with metastatic recurrence following external beam radiotherapy

Abstract Background Enhancer of zeste 2 (EZH2) promotes prostate cancer progression. We hypothesized that increased EZH2 expression is associated with postradiotherapy metastatic disease recurrence, and may promote radioresistance. Methods EZH2 expression was investigated using immunohistochemistry in diagnostic prostate biopsies of 113 prostate cancer patients treated with radiotherapy with curative intent. Associations between EZH2 expression in malignant and benign tissue in prostate biopsy cores and outcomes were investigated using univariate and multivariate Cox regression analyses. LNCaP and PC3 cell radiosensitivity was investigated using colony formation and γH2AX assays following UNC1999 chemical probe‐mediated EZH2 inhibition. Results While there was no significant association between EZH2 expression and biochemical recurrence following radiotherapy, univariate analysis revealed that prostate cancer cytoplasmic and total EZH2 expression were significantly associated with metastasis development postradiotherapy (P = 0.034 and P = 0.003, respectively). On multivariate analysis, the prostate cancer total EZH2 expression score remained statistically significant (P = 0.003), while cytoplasmic EZH2 expression did not reach statistical significance (P = 0.053). No association was observed between normal adjacent prostate EZH2 expression and biochemical recurrence or metastasis. LNCaP and PC3 cell treatment with UNC1999 reduced histone H3 lysine 27 tri‐methylation levels. Irradiation of LNCaP or PC3 cells with a single 2 Gy fraction with UNC1999‐mediated EZH2 inhibition resulted in a statistically significant, though modest, reduction in cell colony number for both cell lines. Increased γH2AX foci were observed 24 hours after ionizing irradiation in LNCaP cells, but not in PC3, following UNC1999‐mediated EZH2 inhibition vs controls. Conclusions Taken together, these results reveal that high pretreatment EZH2 expression in prostate cancer in diagnostic biopsies is associated with an increased risk of postradiotherapy metastatic disease recurrence, but EZH2 function may only at most play a modest role in promoting prostate cancer cell radioresistance.

Research Centre (Molecular Diagnostics Theme/Multimodal Pathology Subtheme); National Institutes of Health, Grant/Award Numbers: P50 CA092629, P30 CA008748 not reach statistical significance (P = 0.053). No association was observed between normal adjacent prostate EZH2 expression and biochemical recurrence or metastasis.
LNCaP and PC3 cell treatment with UNC1999 reduced histone H3 lysine 27 trimethylation levels. Irradiation of LNCaP or PC3 cells with a single 2 Gy fraction with UNC1999-mediated EZH2 inhibition resulted in a statistically significant, though modest, reduction in cell colony number for both cell lines. Increased γH2AX foci were observed 24 hours after ionizing irradiation in LNCaP cells, but not in PC3, following UNC1999-mediated EZH2 inhibition vs controls.
Conclusions: Taken together, these results reveal that high pretreatment EZH2 expression in prostate cancer in diagnostic biopsies is associated with an increased risk of postradiotherapy metastatic disease recurrence, but EZH2 function may only at most play a modest role in promoting prostate cancer cell radioresistance.  1 Radical treatment options for localized PCa include radical surgery and radical radiotherapy (RT), which have equivalent cure rates at a median follow-up of 10 years. 2 Almost half of the men with high-risk localized PCa currently receive RT with curative intent, 3 and while concomitant androgen deprivation therapy (ADT) 4,5 and advances in external beam RT delivery 6,7 have improved treatment, RT does not cure all patients. In particular, high-risk localized PCa can recur following RT, with 5-year disease-free survival rates of 78% to 94% being reported for RT plus ADT in large series. 4,[8][9][10] The Polycomb Group protein enhancer of zeste 2 (EZH2) promotes PCa development 11 and is implicated in tumor cell proliferation, invasiveness, metastasis, and progression to a castration-resistant phenotype. [12][13][14][15][16][17][18][19][20] EZH2 functions within the Polycomb Repressive Complex 2 (PRC2) in concert with histone deacetylases (HDACs), 21,22 and catalyzes a transcriptionally repressive histone H3 lysine 27 tri-methylation signal. [23][24][25][26] This results in the recruitment of PRC1, heterochromatin formation, DNA methylation, and gene silencing. 23,24,26,27 Epigenetic regulators mediate resistance to anticancer therapies such as RT through several mechanisms, [28][29][30] and HDAC inhibitors can increase radiosensitivity in several cancers 31,32 including PCa. 33 This study tested the hypothesis that increased EZH2 expression in baseline diagnostic PCa biopsy clinical samples may be associated with subsequent post-RT disease recurrence, and that inhibition of EZH2 function might increase PCa cell radiosensitivity in vitro. We report that patients whose PCa tumors expressed high levels of EZH2 at baseline experienced an increased risk of metastatic disease relapse following RT. We also observed that in vitro inhibition of EZH2 function in PCa cells resulted in only a modest increase in sensitivity to RT treatment. Taken together, these results suggest that increased EZH2 function in PCa promotes post-RT metastatic recurrence through mechanisms above and beyond increased intrinsic radioresistance alone.

| Patient cohort and clinical follow-up
The study population comprised 113 men with PCa who received external beam radical RT with curative intent in Oxford between 2000 and 2005 (from a database of approximately 800 such PCa patients), and from whom pretreatment prostate biopsy slides were archived and available, and on whom outcome data were ascertained.
Anonymous clinical data were available from medical chart review, including age at diagnosis, date of RT treatment, prostate-specific antigen (PSA) at diagnosis and during follow-up, initial standard-ofcare contemporary staging imaging where performed (usually comprising isotope bone scan or computed tomography (CT) scan if high-risk disease or PSA greater than 20 ng/ml at diagnosis), clinical tumor (cT) stage, biopsy Gleason grade group, and clinical follow-up data for biochemical recurrence (BCR) and/or distant metastasis.
Patients were reviewed in the clinic at least once every 6 months after RT, for a minimum of 3 years. External beam RT was 3D conformal and CT planned, and typically a 55 Gy dose was delivered to the planned target volume in twenty fractions over 4 weeks with neoadjuvant and concurrent ADT as previously described. 34,35 Assuming α/β ratio for PCa of 1.8 Gy, 36 this dose/fractionation schedule is equivalent to 65.9 Gy in 2 Gy fractions. RT was administered to all other patients in fractions of 2 Gy. Using followup data including serial PSA monitoring, isotope bone scans, CT scans, magnetic resonance imaging (MRI) and positron emission tomography (PET)/CT scans, patients were assigned to one of three mutually exclusive groups: long-term remission, BCR, or radiologically confirmed distant metastatic relapse. BCR was defined using the ASTRO-Phoenix Consensus criteria 37 as a PSA rise greater than 2 ng/ ml above the post-RT nadir, without evidence of metastatic disease, and if this occurred patients would usually be commenced on ADT, unless contra-indicated due to competing comorbidity or frailty.
Metastatic PCa was defined as bony, visceral, or lymph-node metastases on follow-up imaging (isotope bone scan, or MRI, or PET/CT scan), or inferred by a PSA rise to greater than 100 ng/ml.
The study had institutional ethical committee approval (ORB ethics 09/H0606/5 + 5), and appropriate checks for patient consent for anonymous use of tissue for research were undertaken.

| Immunohistochemistry
Archival diagnostic formalin-fixed paraffin-embedded (FFPE) prostate biopsy samples were selected for this study as described previously. 34,35 Sections were deparaffinized and rehydrated in the standard manner, endogenous peroxidase activity was inactivated using 3% H 2 O 2 in methanol, blocked with 5% normal goat serum and incubated with a previously validated anti-EZH2 primary antibody (anti-EZH2, clone AE25, cat. no. MABE362, 1:1000; Merck Millipore, Watford, UK) 38 at 4°C overnight. Following the addition of a biotinylated secondary antibody, an avidin/biotin-based peroxidase solution was added, followed by 3,3′-diaminobenzidine solution hematoxylin counterstaining. Sections were dehydrated and mounted as standard. Stained PCa biopsy samples were scored by a consultant uropathologist blinded to patient-and tumor characteristics, and a malignant epithelium EZH2 expression intensity score was assigned ranging from 0 (no expression) to 3 (maximal expression), which was multiplied by the percentage of stained cells, to yield a total PCa EZH2 expression score (range, 0-300) for each of "nuclear" and "cytoplasmic" EZH2. A PCa "total" EZH2 expression score was calculated as the sum of nuclear plus cytoplasmic staining (range, 0-600). Where "normal adjacent benign prostate" tissue was available within the prostate biopsy samples, a benign "nuclear," "cytoplasmic," and "total" EZH2 expression score was similarly obtained.

| Antibodies
The following antibodies were used in this study: anti-Histone H3 tri-

| Colony formation assays and γH2AX immunofluorescence
LNCaP and PC3 cells were treated with UNC1999 inhibitor or DMSO as a negative control for 96 hours, and then lifted, diluted and plated into six-well plates in triplicate to perform a colonyformation assay (CFA). Approximately 500 PC3 cells and 6000 LNCaP cells were plated ahead of irradiation in medium containing DMSO control or UNC1999 at different doses. Cells were left for 24 hours at 37°C (5% CO 2 ) to settle and adhere, and treatment plates were then irradiated at 2, 4, and 6 Gy using a Caesium-137 irradiator, Gamma Service: GSR D1; dose rate 1.938 Gy/min. At 24 hours postirradiation, cells were changed to medium without UNC1999. Colonies were grown for 10 to 14 days postirradiation and then stained with crystal violet, and colonies were then counted using a GelCount colony counter (Oxford Optronix, Abingdon, UK). Effects of UNC1999 treatment on the number of surviving colonies at 0, 2, 4, and 6 Gy were compared against DMSO-treated control cells using a paired t test. For γH2AX foci analysis, UNC1999 and DMSO control-treated cells were irradiated as described above, and then fixed at different time points using 4% paraformaldehyde, and permeabilized by fresh 0.1% Triton X-100 in 4% fetal calf serum-containing solution. Cells were probed with an anti-γH2AX antibody, and foci were imaged using a Zeiss LSM 710 confocal microscope and quantified using ImageJ software (NIH Image, NIH, Bethesda, MD).

| Statistical analysis
We sought to assess whether EZH2 expression levels, as determined by immunohistochemistry on pre-RT prostate biopsy samples, could predict BCR or distant metastasis after RT when added to standard predictors. Firstly, we assessed the univariate association between EZH2 expression levels (nuclear, cytoplasmic, and total [nuclear plus cytoplasmic]) per 100-unit change, and the outcome (BCR and metastasis respectively), using Cox regression. We then studied the multivariate association, using Cox regression, between EZH2 expression level and the outcomes, adjusting for PSA, cT stage, and biopsy Gleason grade group. Due to the limited number (n = 17) of patients with image-confirmed metastatic disease during follow-up it was not feasible to include these covariates in a single model. Therefore, a risk score was created using PSA (cubic splines were used to account for nonlinearity), cT stage (cT1 vs cT2 vs cT3/4), and biopsy Gleason grade group (1 vs 2 vs ≥ 3) to predict BCR after external beam radical RT treatment. The risk score was then utilized for model adjustment. For models where the EZH2 expression score was significantly associated with the outcome on multivariate analysis, the improvement in discrimination (Harrell's c-index) was reported and corrected for optimism (to attenuate the discrimination estimate slightly, to better estimate the true discrimination) using bootstrap methods. 39 BCR free-and metastasis-free survival was calculated using Kaplan-Meier analysis, and patients who did not recur were censored at the date of last clinical follow-up. All statistical analyses of the clinical cohort and EZH2 expression were performed using Stata 15.0 (StataCorp, College Station, TX).
All statistical tests of in vitro experiment data were performed as two-tailed t tests and differences were considered significant at a P < 0.05. All in vitro colony formation assay data are representative of three independent experiments, each being performed in triplicate, and are presented as the mean ± standard error of the mean (SEM) from these multiple repeat experiments.
While delivery of external beam RT as a curative treatment option for men with PCa has improved, many patients develop disease recurrence despite concurrent ADT. There is an unmet clinical need to identify druggable targets to increase tumor radiosensitivity, 40,41 and to identify markers of RT treatment failure. We provide evidence that EZH2 overexpression in pretreatment PCa biopsies is associated with subsequent metastatic PCa recurrence following radical RT.
EZH2 is overexpressed in castration-resistant PCa, and RT can reduce EZH2 expression in PCa cells. 42   status of these two cell lines (PC3 being p53-null, and LNCaP containing wild-type p53) might also contribute to the different radiosensitivities observed in our experiments, given the important roles of p53 in double-strand break response and DNA repair. 45 While further experiments would be required to test this hypothesis, the available evidence suggests that any radiosensitizing effect of EZH2 inhibition in the PCa cell lines tested is modest.
While it is acknowledged that UNC1999 is a selective inhibitor of both EZH2 and EZH1, 46 we observed the expected reduction in H3K27Me3 levels following treatment of PCa cells with UNC1999.
Additional research is necessary to further understand any potential differential oncogenic properties of EZH1 vs EZH2.
While studies investigating potential correlations between base- While the 10-year BCR-free survival was low at 22% for patients within this cohort, this is comparable with reported rates of 30% for high-risk disease following external beam RT, 54,55 and it is possible that BCR at such a mature length of follow-up does not equate to true disease recurrence. While it is a strength of our study that the cohort was mature with a median follow-up of 7.9 (IQR, 6.8-8.4) years, we acknowledge that these were not consecutive patients from our institution due to inherent constraints acquiring archival tissue with retrospective follow-up. Indeed, our cohort size of 113 patients is modest, though this has been sufficient to identify other potential mediators of radioresistance. 34,35 Data on the larger cohort from which these patients originated was unfortunately not available.
It would be helpful to validate our findings in similar cohorts from independent institutions.
Patients developing metastatic PCa recurrence following external beam RT may include those with occult micrometastases at the time of irradiation, along with others with local disease recurrence within the radiation field due to radioresistant PCa. Thirty percent of post-RT BCR is estimated to be due to local recurrence indicative of clinical radioresistance. 56 A weakness of our study is that it is difficult to differentiate between patients who may have had micrometastases at baseline, and those who may initially have developed local recurrence with a subsequent metastatic phenotype, because post-RT imaging was generally not performed until BCR occurred, and not all patients in the cohort received re-staging imaging. A contemporary cohort may receive PSMA-PET/CT 57 to accurately evaluate local vs metastatic disease recurrence post-RT, but as this is only a recent clinical development such a cohort with accurate recurrence classifications would lack long-term follow-up. It will be valuable for future studies to investigate whether EZH2 expression in pre-RT samples predicts post-RT disease recurrence in cohorts with accurate post-BCR stage classifications based on molecular imaging.

| CONCLUSIONS
In conclusion, patients with a high level of EZH2 expression in the baseline diagnostic PCa biopsy specimens had an increased risk of metastatic disease recurrence following external beam RT with curative intent. Chemical probe-mediated inhibition of EZH2 function only results in a modest increase in radiosensitivity of PCa cells in vitro. Taken together this suggests that EZH2 function promotes WU ET AL.