Tissue microarray analysis indicates hedgehog signaling as a potential prognostic factor in intermediate-risk prostate cancer

Tissue microarrays (TMAs) were constructed from formalin-fixed paraffin-embedded (FFPE) radical prostatectomy specimens from 170 PCa patients, and from trans-urethral resection of the prostate (TURP) samples from 64 patients with benign prostatic hyperplasia (BPH). The TMA consisted of a single tissue core for each of the 234 patients.

TMA sections were obtained from the Oxford Centre for Histopathology Research, Oxford University Hospitals NHS Foundation Trust, Oxford, UK. The TMA contained patient samples from Oxford and was built under the approval of the Oxford Radcliffe Biobank Ethics Committee (reference number 09/H0606/5 + 5). The study itself was approved by the ethics committee of KU Leuven, Leuven, Belgium (reference number S55726).

Immunohistochemistry was performed using primary antibodies against SHH (1/50, Abcam ab53281), PTCH1 (1/300, Santa Cruz sc-6147), SMO (1/100, Abcam, ab72130), SUFU (1/100, Santa Cruz sc-28,847), GLI1 (1/50, Santa-Cruz sc-20,687), GLI2 (1/1000, Rockland 600–401-845), GLI3 (1/50, R&D AF3690), SNAIL (1/50, R&D, AF3639), SNAI3 (1/50, Novus Biologicals, NBP1–90661), CYCLIND1 (Dako, M364229) and CD31 (Dako, IR610 Clone JC70A). After incubation with the appropriate secondary antibody (Vector Laboratories), antigen presence was revealed with 3.3′-diaminobenzidine (DAB) substrate (Vector Laboratories, ImmPACT DAB, SK-4105) and slides were counterstained with hematoxylin. Control TURP resection specimens were used to validate the specificity of the primary antibodies, which was histopathologically assessed based on the expected subcellular localization of each protein and a lack of nonspecific background staining. For the validation of Snail expression, blood vessel staining was used as an internal positive control. To exclude any nonspecific staining of the secondary antibodies, negative controls were performed without the addition of any primary antibody.

Two independent researchers performed evaluation of the immunohistochemical staining for each protein, and each researcher was blinded to clinicopathological and outcome data. The percentage of stained cells (0–100%) and the staining intensity (0 = negative, 1 = weak, 2 = moderate, 3 = strong) were assessed both in the epithelial and stromal cells. In the malignant cores, only the tumoral glands were considered. Semi-quantitative analyses were performed by calculating histoscores (HS) as the product of the percentage stained cells (0–100) and staining intensity (0–3). If multiple staining intensities were present in the same core, the sum of the individual HS was taken to acquire the average HS of the entire core. Binary HS, with low and high expression respectively being defined as below and above the mean HS of 1.5, were used for statistical analyses. Microvessel density (MVD) was determined by counting the number of CD31-positive blood vessels in each core.

The IBM/SPSS Statistics version 23/24 was used for statistical analyses. A Fisher’s exact test was used to compare the binary Hh expression level in the benign and malignant tissue cores, and to evaluate any potential association between the protein expression levels in the tumor and the clinicopathological parameters. One-way ANOVA with contrast analysis was used to assess any potential correlation between MVD and clinicopathological factors.

The patient and tumor characteristics of the 170 PCa patients are shown in Additional file 1: Table S1. The median age at the time of surgery was 61 years (range 45–71). Approximately two-thirds of patients had a pathological T stage (pT) ≤ pT2c, and only a small number of patients had a pathological GS ≥8. Median follow-up was 8.4 years (range 1.1–13.7), and 23% of patients developed biochemical recurrence (BCR) of PCa following radical surgery, as defined as a confirmed post-operative rise in PSA level to >0.2 ng/ml.

Protein expression of the main Hh components in the PCa cores was compared with the expression level in the benign cores (Table 1 and Fig. 2). In general, Hh signaling protein expression was greater in the prostate epithelium than in the stromal tissue. When considering Hh signaling proteins in the epithelium, we observed that low levels of PTCH1 and GLI3 expression were more likely to occur in tumor tissue than in benign (PTCH1 76% PCa versus 26.6% benign, and GLI3 55.2% versus 29%, respectively, p < 0.001).

In contrast, high expression of the SUFU negative Hh regulator was observed in both benign and malignant prostate epithelium, with significantly higher levels being seen in tumors (p = 0.028).. GLI2 was observed to be widely and highly expressed in both benign and malignant prostate tissues. High levels of Cyclin D1 were more commonly seen in PCa than in benign prostate epithelium (p < 0.001). High epithelial SNAIL expression was also mainly observed in PCa compared with benign prostate epithelium (p = 0.005). Few differences in Hh signaling protein expression levels were observed in the stromal compartments of benign and malignant prostate tissue, with only SUFU being more highly expressed in the stromal compartment of PCa tissue than in benign cores (p < 0.007).

We evaluated the clinical importance of Hh signaling protein expression in PCa patients. First, we investigated potential correlations between Hh protein expression and known prognostic clinicopathological factors such as pT stage, pathological GS, and surgical margin status. Higher epithelial GLI2 expression in the tumor was found to correlate with higher pathological GS (p = 0.047, Table 2), indicating that this might be a potential marker for aggressiveness in this PCa cohort. A trend was observed for higher epithelial GLI2 expression in tumors of a more advanced pT stage (p = 0.084).

We also observed a positive correlation between the mean vessel density (MVD) and pathological GS, suggesting that tumors with a more aggressive phenotype may present with a greater number of blood vessels (Fig. 3). The other Hh signaling proteins investigated did not demonstrate significant correlations with clinicopathological factors.

Patients with a higher GS (pGS ≥ 7; p = 0.023), more advanced tumor stage (pT stage > pT2c; p < 0.001) and/or a positive surgical margin status (p = 0.004) had a shorter time to BCR where this occured. Higher epithelial GLI3 expression in the tumor showed a trend towards a better prognosis although this did not reach statistical significance in this cohort (p = 0.092) (Fig. 4).

In a multivariate Cox regression model, pT stage (HR = 3.317; p = 0.002), pathological GS (HR = 2.572; p = 0.4) and epithelial GLI3 expression (HR = 0.418; p = 0.023) were each found to be significant predictors of BCR (Table 3).