Expression of SOCS1 and the downstream targets of its putative tumor suppressor functions in prostate cancer

From archived radical prostatectomy specimens received at the pathology department of the Centre hospitalier universitaire de Sherbrooke (CHUS), 175 consecutive samples collected between 2012 and 2015 were selected for this study. As the priority of this study was to evaluate tissue protein expression, and because archived specimens progressively lose antigenicity, only recent prostatectomy specimens collected between 2012 and 2015 were used. All the prostatectomy specimens were treatment-naïve. The study design also did not include serum PSA, disease outcome or patient follow-up, as only the recent prostatectomy specimens were prioritized to assess marker expression. Following approval by the Ethics committee of the Centre de Recherche du CHUS (Project #2014-734, 13–222), the patients or their families were contacted and consent was obtained from 90 patients. Among them, 12 cases were excluded due to a small tumor size of less than 5 mm in diameter, and 78 were included in this study.

Tumor grading and staging data from the histopathology report were verified. Tumors were re-categorized according to the recommendations of the International Society of Urological Pathology (ISUP) into 5 groups, represented by the sum value of the first and second most prevalent Gleason patterns [32]: group 1 (Gleason score ≤6), group 2 (Gleason score 7 = 3 + 4), group 3 (Gleason score 7 = 4 + 3), group 4 (Gleason scores 8 = 4 + 4, 3 + 5, 5 + 3) and group 5 (Gleason score 9–10) [33]. All 78 cases harboured prostatic acinar adenocarcinoma. Among them, four showed foamy gland features, one displayed mucinous features and two others contained a ductal adenocarcinoma component. Within the study population, lymph node dissection had been performed in 36 cases, of which only 4 (N1) showed lymph node metastasis and the rest (N0, n = 32) did not. The specimens were categorized into two additional groups T2 and T3 stages, respectively representing those with carcinoma confined within the prostatic capsule (organ confined) and those that have breached this confinement (extra-prostatic extension) according to the American Joint Committee on Cancer (AJCC) Staging manual (7^th edition) [34].

Following a thorough review of the haematoxylin and eosin (H&E)-stained slides, the largest tumor foci representative of the Gleason score was identified and correspondingly mapped on the paraffin block using a stereomicroscope (SZ51, Olympus) for precise targeting with the TMA MASTER system (3DHISTECH Ltd, Budapest, Hungary). The areas rich in tumor cells and representing the two prominent Gleason patterns were targeted. One large 2 mm diameter core was extracted from the mapped area and the 78 tumor cores were fitted in three TMAs using the TMA MASTER system. Cores containing clearly defined transitional epithelium, normal prostatic epithelium, low-grade and high-grade prostatic intraepithelial neoplasia (PIN) served as control or reference. PIN was defined according to WHO criteria [33].

TMA sections of 4 μm thickness mounted on charged slides were deparaffinised, re-hydrated and heat-induced antigen retrieval was performed on DAKO PT Link pre-treatment module (Dako Pathology products, Mississauga, ON) using low or high pH Envision FLEX Target Retrieval Solutions (Dako). Subsequent steps were carried out on DAKO Autostainer Link. Endogenous peroxidase activity was blocked using Envision FLEX Peroxidase-blocking Reagent, followed by incubation with the primary antibodies (Ab) listed in Table 1. The EnVision FLEX /HRP (Dako) reagent containing a dextran polymer conjugated to anti-mouse Ig and anti-rabbit Ig and HRP was used as the secondary Ab. All washing steps were carried out using the Envision FLEX wash buffer and positive staining was detected by precipitation of diaminobenzidine (DAB) using the Envision FLEX DAB+ reagent. Slides were counterstained with hematoxylin, and coverslips were mounted using the Faramount mounting medium (Dako). The images were captured using an automated slide scanner (Aperio ScanScope XT, Aperio, Vista, California), analysed and scored in a blinded manner. A semi-objective scoring method (H-score) was used to quantify the staining. H-Score was evaluated on the whole tissue core. For the various markers, only the nuclear (Ki67, p53, AR and p21), the granular cytoplasmic (Prostein, SOCS1) or the membrane (MET) staining was considered for scoring. The staining intensities were graded on the scale of 0, 1, 2 and 3, representing absent, weak, moderate and strong staining. These values were multiplied by the respective percentage of staining to obtain the H-scores ranging from 0 to 300. Thus, the H-score recognises broad ranges of antigen expression and avoids arbitrary cut-offs.

Statistical analyses were carried out using the GraphPad Prism 6 software (GraphPad Software, San Diego, USA). The data points were tested for Gaussian distribution before performing relevant statistical analyses, as indicated in each figure. For each dataset within groups, the distribution, mean and standard deviation within the 95% confidence limits are shown. Correlation between any two parameters was evaluated by non-parametric Spearman correlation test and the correlation coefficient (ρ) and the p value of the correlation are indicated within each figure. The slope was generated by non-linear regression curve-fit analysis. Non-parametric Kruskal-Wallis and Mann–Whitney tests were employed for comparing multiple or two groups, respectively. Statistically significant p values are represented by asterisks (* ≤ 0.05, ** ≤ 0.01), or the actual values are indicated within each figure.

In order to evaluate the protein expression of SOCS1 and its putative downstream targets of tumor suppression in PCa, we constructed TMAs from archived PCa specimens and performed automated IHC staining to ensure staining homogeneity. To limit overlooking potential heterogeneity within individual tumor specimens, precautions were taken to include as much high-grade tumor foci as possible within the TMA core and not to exclude any smaller foci of higher grade. The TMAs were first stained by H&E to ascertain their histological features and distinguish carcinoma, PIN and benign structures, allowing their classification into five ISUP-recommended groups: 1 (n = 16), 2 (n = 39), 3 (n = 1), 4 (n = 8) and 5 (n = 4). The TMA sections were then stained for Ki67, a common cancer-associated marker (Fig. 1b). The high-grade tumors contained many Ki67+ cells compared to intermediate grade tumors, low grade tumors showed occasional Ki67+ cells, and the normal tissue rarely showed any positivity. Staining for prostein showed an opposite pattern compared to Ki67 staining. In general, low-grade tumors displayed intense prostein staining comparable to normal prostatic epithelium, and high-grade tumors showed reduced prostein staining (Fig. 1c). Consistent with these staining patterns, regression analysis showed highly significant positive correlation between the tumor grade and Ki67 staining, whereas prostein showed an inverse correlation (Fig. 1d and e). These results indicated that the study cohort was well represented by samples of different disease severity, despite containing relatively fewer high-grade tumors.

It has been reported that androgen stimulation induces SOCS1 gene expression in PC3 cells [17]. Therefore, we evaluated the expression of androgen receptor (AR) and SOCS1 in PCa TMAs. The representative staining patterns and the corresponding tumor grades are shown in Fig. 2a and b. In contrast to antibodies specific to Ki67, prostein and AR that are approved for pathological diagnosis, the SOCS1 Ab is a research-grade rabbit polyclonal Ab from Santa Cruz Biotechnology (H-93, Catalogue #sc-9021). This SOCS1 Ab has been previously used on gastric and PCa specimens [17, 35] and by the Human Protein Atlas project [36]. This Ab detected a specific band of 27 kDa as expected in western blot analysis of COS-7 cells expressing human SOCS1 protein (Additional file 1: Figure S1). Two additional bands of higher MW estimated at 42 and 50 kDa were also observed only in SOCS1-transfected cells, suggesting that these bands might represent post-translationally modified SOCS1 protein. None of these bands were detected in control vector-transfected cells, indicating the specificity of the SOCS1 Ab used. We have optimized the staining conditions for IHC using this Ab, specifically pH of the antigen retrieval solution, primary Ab dilution and incubation time, using human fallopian tube specimens as documented in the manufacturer’s datasheet (data not shown). We obtained specific SOCS1 staining in PCa specimens as indicated by positive and negative cells within the same field of observation (Fig. 2b). We observed that SOCS1 showed a granular cytoplasmic staining of the prostatic epithelium with intense staining at the perinuclear region, predominantly on the luminal side (Fig. 2c). Whereas AR staining did not correlate with the ISUP grade groups, SOCS1 staining showed a significant negative correlation, similarly to prostein (Fig. 2d and Fig. 1e).

Although SOCS1 showed a staining pattern similar to that of prostein (Fig. 1c and Fig. 2b), it was unlikely that SOCS1 Ab displayed any cross-reactivity with prostein for the following reasons. While the prostein staining is highly specific to the prostatic epithelium and carcinoma as expected [37], the SOCS1 Ab also stained immune and stromal cells surrounding the gland and around the prostatic ducts (Fig. 2e and f). Moreover, in contrast to prostate epithelial cells, SOCS1 staining in immune cells showed a predominant nuclear distribution (Fig. 2e-g), indicating that SOCS1 staining is distinct from that of prostein.

Evaluation of the relationship between SOCS1 staining and that of AR did not show any association, whereas a positive correlation was observed between SOCS1 and prostein (Fig. 3a). However, AR staining correlated positively with Ki67 staining, whereas SOCS1 and prostein displayed significant negative correlation with Ki67 staining (Fig. 3b), suggesting that SOCS1 expression in PCa specimens might have a diagnostic significance.

SOCS1 can mediate tumor suppression by several potential mechanisms that were defined using cell and animal models. These include activation of p53-induced cellular senescence, and inhibition of oncogenic MET and p21 signalling through their ubiquitination and proteasomal degradation [20‐23]. Hence, we postulated that the loss of SOCS1 might result in efficient p53 activation and compensatory upregulation of p53 expression, and increase in the expression of MET and p21. As each of these proteins is implicated in PCa pathogenesis [29, 30, 38], we evaluated their expression and correlation to SOCS1 levels in PCa TMAs. Representative expression patterns of p53, MET and p21 are shown in Fig. 4a-c. Occasionally, MET expression was confined to cyst-like structures and atrophic ducts with true membranous staining (Fig. 4d), although its significance is unclear. While p53 expression showed significant positive correlation with ISUP grade groups, MET and p21 only showed a tendency for positive correlation that were not statistically significant (Fig. 4e). SOCS1 staining did not correlate with the expression of p53, MET or p21 (Fig. 4f). These results suggested that the loss of SOCS1 might affect the downstream mediators of its tumor suppression mechanisms in a heterogeneous manner.

It has been reported that AR signalling represses MET expression, whereas androgen deprivation is associated with increased expression of p21 [30, 39], suggesting that escape to androgen independence would lead to increased expression of MET and p21, promoting their oncogenic functions. Therefore, we examined the relationship between the expression of AR and those of MET and p21. A strong positive association was observed between AR and p21 expression, whereas the positive correlation tendency between AR and MET was not statistically significant (Fig. 5a). MET and p21, which showed a significant positive correlation were also strongly associated with Ki67 staining (Fig. 5b and c). These results indicated that even though AR expression did not correlate with the ISUP grade or MET (Fig. 2d), AR and MET signalling might influence the expression of p21 expression and its oncogenic functions.

We grouped the specimens into tumors that are either confined within (T2, n = 47) or spread beyond (T3, n = 41) the prostatic capsule. Evaluation of the marker expression in these two groups revealed that the invasive tumors showed significantly higher Ki67 and lower prostein staining (Fig. 6). On the other hand, expression of AR, SOCS1, p53, MET or p21 was not significantly different between T2 and T3 groups. A subset of the samples, for which tumor spread to the regional lymph node had been evaluated (n = 36), were grouped into N0 (n = 32) and N1 (n = 4) subgroups, respectively representing cases without or with metastatic spread to the regional lymph node. Even though the latter group contained only a limited number of cases, their Ki67 expression was significantly higher and prostein levels were significantly lower compared to specimens in the former group, while the AR expression was comparable (Fig. 7). Moreover, specimens from cases with lymph node metastasis showed significantly reduced SOCS1 staining and concomitant increase in MET and p21 expression. Although these findings need to be confirmed in larger study populations, they suggest that SOCS1, MET and p21 levels could be useful as predictive markers of metastatic PCa.