Periostin is up-regulated in high grade and high stage prostate cancer

Background Expression of periostin is an indicator of epithelial-mesenchymal transition in cancer but a detailed analysis of periostin expression in prostate cancer has not been conducted so far. Methods Here, we evaluated periostin expression in prostate cancer cells and peritumoural stroma immunohistochemically in two independent prostate cancer cohorts, including a training cohort (n = 93) and a test cohort (n = 325). Metastatic prostate cancers (n = 20), hormone refractory prostate cancers (n = 19) and benign prostatic tissues (n = 38) were also analyzed. Results In total, strong epithelial periostin expression was detectable in 142 of 418 (34.0%) of prostate carcinomas and in 11 of 38 benign prostate glands (28.9%). Increased periostin expression in carcinoma cells was significantly associated with high Gleason score (p < 0.01) and advanced tumour stage (p < 0.05) in the test cohort. Whereas periostin expression was weak or absent in the stroma around normal prostate glands, strong periostin expression in tumour stroma was found in most primary and metastatic prostate cancers. High stromal periostin expression was associated with higher Gleason scores (p < 0.001). There was a relationship between stromal periostin expression and shortened PSA relapse free survival times in the training cohort (p < 0.05). Conclusions Our data indicate that periostin up-regulation is related to increased tumour aggressiveness in prostate cancer and might be a promising target for therapeutical interventions in primary and metastatic prostate cancer.


Background
Periostin (POSTN) is a 93 kDa N-glycoprotein, first described in 1993 in mouse osteoblasts as osteoblast-specific factor 2 (OSF-2). It shows homology with the cell adhesion molecules fasciclin 1 (drosophila) and beta-IgH3 (human), sharing features that are thought to explain some of its functional characteristics [1,2] like involvement in cell adhesion and osteoblast recruitment [3].
Periostin has been found in several, mainly collagenrich and fetal tissues as an extracellular matrix protein and is up-regulated by mechanical stress during tissue repair and (re)generation [4][5][6][7][8]. Periostin expression can be induced by vascular injury which in turn induces vas-cular endothelial growth factor receptor 2 with consequent promotion of angiogenesis [9,10]. After myocardial infarction, periostin up-regulation seems to be important for the healing process [11,12].
So far there is only a single report on periostin expression in prostate cancer [35]. Increased cancer cell expression of periostin compared to normal glands was found during early stages of prostate cancer whereas in advanced stages stromal periostin expression prevailed [35]. The aim of our study was to determine the periostin expression in the stromal and epithelial compartment of the tumour, as well as the correlation with clinical data including patient follow up data in a larger cohort.  25-2007 neu). In the latter, necessity of patients' informed consent was explicitely ruled out, since this is a retrospective study.

Tissue microarray construction
The tissue microarrays (TMA) were constructed as previously described [36]. We used commercially available tissue arrayers (Beecher Instruments, Woodland, CA, USA) and applied a core diameter of 0.6 mm for the tissue samples of Zurich and 1.0 mm for the tissue samples of Berlin. Each tumour was represented by one tissue core.

Evaluation of stainings
The periostin protein expression was evaluated by two clinical pathologists (FRF, GK) on a multi-headed microscope. The test cohort was subsequently re-evaluated by another pathologist (AS).

Statistical analysis
Statistical analyses were performed with SPSS 17.0 (SPSS Inc., Chicago, IL, USA). The median value of the IRS was used as cut-off point to dichotomize the tumours into a "periostin low" and "periostin high" group. Fisher's exact and chi-squared tests were applied to assess associations between categorized periostin expression and clinicopathological parameters. Correlations were computed using Spearman's bivariate rank order correlation. Univariate survival analysis was carried out according to Kaplan-Meier, differences in survival curves were assessed with the Log rank test. P-values < 0.05 were considered significant.

Periostin expression in epithelia and stroma of prostate tissues
Distinct stromal and epithelial staining characteristics allowed an absolutely certain evaluation of the periostin staining ( Figure 1a and 1b). Benign prostate glands expressed high stromal periostin in only 2/38 cases and high epithelial periostin in 11/38 cases. From the 38 benign prostate samples, 24 displayed no periglandular stromal and 14 no epithelial periostin expression. Of the 24 benign cases without stromal periostin expression, 19 showed epithelial periostin expression and vice versa of the 14 benign cases without periostin expression, 9 revealed stromal periostin expression. Both periostin epithelial and stromal negativity occurred in 5 of the benign cases. Five cases were positive for periostin in both epithelia and stroma. Basal cells showed in some cases a slightly stronger staining than the inner secretory cell layer. In the remaining tumour epithelia, periostin was detected in the cytoplasm without luminal or membranous accentuation. Nuclear staining was not observed. The stroma displayed a fibrillary pattern with considerable variation of intensity. Staining intensity differed frequently within a respective case between the stroma and epithelium.
Tumour stroma was positive in all cases of the training cohort and most cases of the test cohort. Only 11 (3.4%) cases of the test cohort were negative. An IRS above 2 was found in the vast majority of primary (82.8%), hormone resistant (78.9%) and metastatic (85%) tumours. Using our periostin score, it was possible to differentiate between low and high periostin expression levels using the median. The median stromal IRS for both primary prostate cancer cohorts was 6 with a mean value of about 5.5.
Sixty (18.5%) primary prostate cancer cases showed no epithelial periostin expression and 189 (58.2%) cases had an IRS equal to or below 3 (median 3). In total, 142/418 prostate cancer cases expressed high levels of epithelial periostin. Only 7.4% of cases exhibited an IRS for epithelial periostin expression above 6. Revalidation of the stainings resulted in the same median IRS values.
For the 19 hormone resistant prostate cancers the median IRS was 8 for the stromal and 4 for the epithelial periostin expression (means: 6.8 and 5.1). In the 20 samples from prostate cancer metastases the mean and median IRS of epithelial and stromal periostin expression did not differ from that in the primary prostate cancers.

Correlations and associations with clinico-pathological parameters
In the training cohort, stromal periostin expression showed no correlation (Spearman rank order) with any of the clinico-pathological parameters (age, pT-status, Gleason score, residual status). A higher pT stage was significantly associated with high epithelial periostin expression (p = 0.026, Table 1). However, Fisher's exact test revealed a significant association of higher periostin stromal expression with positive resection margins (R1) (high periostin expression in R0 versus R1: 14% (n = 7) versus 39% (n = 16); p = 0.008). No other associations were detected (Table 1). In the test cohort, high stromal and epithelial periostin expression were both associated with high Gleason scores (p = 0.011 and 0.007, Table 2). For epithelial expression an additional significant association with advanced pT-status was demonstrated (p = 0.048, Table 2). In the Spearman rank order correlation for the test cohort, the significant associations from above could be confirmed for epithelial expression (p-values: 0.001 and 0.047, Table 3). For stromal expression, the correlation with Gleason score was also significant (p = 0.003, Table 3). Stromal periostin expression was significantly correlated with epithelial periostin expression (p = 0.003, Table 3).

Periostin and PSA relapse free survival
The standard prognosticators (pT-status, Gleason score and residual tumour) correlated significantly with shortened PSA relapse free survival in both cohorts (Table 4, training cohort not shown). In the training cohort, high stromal periostin was significantly associated with shortened PSA relapse free survival times (p < 0.05; Figure 2a).

Discussion
In this study, we provide evidence for periostin up-regulation during prostate cancer progression. Periostin expression was found in both epithelial cancer cells and in peritumoural stroma. Recently, our group has demonstrated that periostin as a marker for the epithelial-mesenchymal-transition (EMT) programme in lung cancer is prognostically relevant [32]. EMT is correlated with tumour progression and represents an important form of tumour-stroma interaction facilitating the stromal invasion of the cancer cells. Periostin seems to play an important part in this prognostically adverse transdifferentiation process. However, the regulation mechanisms of periostin in tumour progression have not been elucidated so far. Our data demonstrate a significant association between periostin and pT-stage, Gleason grade and involvement of prognosis in two different prostate cancer cohorts, suggesting that EMT is of utmost importance for prostate cancer progression. There is only one study by Tsunoda et al. observing a prostate cancer patient cohort of 77 prostate cancers showing increased periostin expression in early prostate cancer stages as well as in the stroma of advanced prostate cancer cases [35]. This is in contrast to our study of 418 prostate carcinomas where we find increased epithelial periostin expression positively correlated to grade and stage and increased stromal periostin positively correlated to grade. Augmentation of both epithelial and stromal periostin in our cohort is a characteristic of the advanced and more aggressive prostate cancer cases. This observation is further supported by the finding that only 2/38 benign prostate tissues expressed stromal periostin. The differences of Tsunoda et al. and our results may be explained by the sample number (77 versus 418) and differences in grade and stage. Our test cohort was represented by 63.1% pT2 and 36.9% pT3/4 tumours whereas Tsunoda's cohort comprised of 18.2% pT2 and 74.0% pT3/4 tumours. Grade in our test cohort was < 7 in 15.4%, 7 in 59.7% and >7 in We do acknowledge that a core diameter of 0.6 mm per prostate cancer case might not be fully representative for a given case. This is especially of importance in small cohorts. However, during the preparation of our tissue microarray, we reviewed each case very carefully to choose a representative area of tumour tissue.  There are several reports on involvement of other EMT markers eg. platelet-derived growth factor-D, hypoxiainducible factor-1α and zinc finger enhancer binding protein 1 in prostate cancer [37][38][39][40]. Meanwhile, periostin has been found up-regulated in several tumour entities, either in the stroma, the epithelial cells or in the serum [17,26,27,30,41,42]. In most of these tumour entities, periostin has been associated with more aggressive tumour characteristics, which is perfectly in line with our findings in prostate cancer. Apart from EMT, periostin is related to other stromal re-modeling and repair processes such as wound healing or formation of heart valves in embryogenesis. It is not clear yet whether periostin upregulation reflects only the stroma re-modeling process per se or whether it is actively induced by the tumour cells themselves. The presence of both mRNA and protein in the cytoplasm of tumour cells favours an active or signal transducting role of periostin, respectively. Further functional studies are needed to shed light on the mechanism of periostin up-regulation in prostate cancer.
In our test cohort we could not reproduce the promising results concerning the prognostic value of periostin deduced from the training cohort. A possible explanation is most likely the selection of the training cohort with a very high number of cases with PSA relapse (46%) whereas the consecutive cases of the test cohort show usual relapse rates (21%). The selection of the training cohort was done to identify biomarker for PSA relapse. It is not uncommon to see that prognostic significances are better in trainings cohorts than in tests cohorts. The differing results for relapse free survival of both cohorts are therefore not too surprising for us, also taking into account that the composition of the cohorts is so different. However, this also demonstrates that the prognostic value of periostin is limited in comparison to well established conventional prognosticators of prostate cancer and other potentially prognostic molecular markers [36,43]. More important than a prognostic value of periostin might be its use as a therapeutic target. Kudo et al. and Castranovo et al. have recently evaluated its therapeutic potential [41,44]. In a chemical proteomics approach, periostin was found accessible by the blood stream, which is an important factor for effective drugability [41,44]. It has been concluded that its expression characteristics and cancer specific up-regulation make periostin a promising target for ligand-based tumour targeting strategies. Considering its high expression in both stroma and tumour cells, this might be an auspicious option for advanced prostate cancer. Also, the diagnostic serological value of periostin might be worth looking at.

Conclusions
This immunohistochemical study describes the periostin protein expression pattern in prostate cancer and benign prostate tissue in a large patient cohort. Its upregulation in primary, metastatic and hormone resistant prostate cancers was related to a more aggressive and advanced tumour biology. These expression characteristics and its proposed drugability make periostin a promising target for an individualized prostate cancer therapy.