KLK6 and KLK13 predict tumor recurrence in epithelial ovarian carcinoma.

Figure 1 illustrates the pattern of KLK6 and KLK13 localisation in ovarian surface epithelium (OSE) and epithelial ovarian carcinoma. Both KLK6 and KLK13 were localised in the cytoplasm of the normal ovarian surface epithelial ( Figures 1A and B ). Both KLK6 and KLK13 showed staining in all types of ovarian adenocarcinoma. Shown here is strong staining for both KLKs in serous adenocarcinoma ( Figures 1C and D ), with KLK6 shown expressed in a mucinous adenocarcinoma ( Figure 1E ) and KLK13 expression in a clear cell tumour ( Figure 1F ). In all cases, staining was most prominent in epithelial cells, whether they were normal surface epithelium or carcinoma. Immunohistological evaluation of protein levels resulted in no significant associations with clinical characteristics. The intensity of staining may not be only because of the variation of kallikrein expression in each pathological specimen, but also be contributed by the variations in fixing and embedding procedures used during the processing of tissues. To quantify the differences in expression more accurately, we extracted total RNA from ovarian samples and analysed by qRT-PCR, comparing KLK expression in ovarian tumours to normal samples. The relationship between KLK6 and KLK13 expression levels and clinical characteristics is summarised in Table 1 . Patient ages range from 20 to 89 years with a mean age of 60 years. Compared to patients with low KLK6 , a significantly larger proportion of patients with high KLK6 had invasive cancer ( P =0.002) and late stage cancers ( P =0.001). Compared to patients with low KLK13 , a larger proportion of patients with high KLK13 had invasive cancer ( P =0.039). Unlike KLK6 , KLK13 was not associated with clinical stage. At diagnosis, 65% of all patients had serous ovarian cancer. In supplementary analysis, when we compared serous to non-serous ovarian cancers (mucinous, endometrioid, clear cell, and unknown), high KLK6 expression was associated with serous carcinomas ( P =0.001; data not shown). Interestingly, all endometrioid ovarian cancers had high KLK13 expression relative to normal ovaries, but the small sample size was unable to provide sufficient power for a conclusive association. Kaplan–Meier survival curves indicated patients with high KLK6 expression were more likely to have a shorter RFS ( P =0.002, Figure 2A ) and OS ( P =0.011, Figure 2B ), when compared to patients with low KLK6 expression. These data are further supported by the Cox regression analysis presented in Table 2 . In univariate analysis, patients with high KLK6 expression had a greater risk of recurrence ( P =0.004) than patients with low KLK6 expressing tumours. As expected, clinical stage ( P <0.001), tumour grade ( P =0.012), and histological type ( P =0.024) were all significant predictors of recurrence. In the multivariate model ( Table 3 ), high KLK6 expression remains significant as a predictor of recurrence ( P =0.040), indicating these patients are approximately three times more likely to have a recurrence than patients with low KLK6 expression. Overall, late clinical stage (stage III/IV) was the strongest predictor of recurrence ( P =0.001). When OS is examined in a Cox univariate model ( Table 2 ), high KLK6 expression is significantly associated with a shorter OS ( P =0.013). As expected, clinical stage ( P <0.001), tumour grade ( P <0.001), and histological type ( P =0.037) are all associated with a shorter OS. Interestingly, patients 50 years of age or older, at the time of diagnosis, also had a significantly shorter OS ( P =0.045) than patients under the age of 50. When these factors were included in a multivariate analysis ( Table 3 ), clinical stage was the strongest predictor of OS, patients with late stage (stage III/IV) cancers had a four-fold increased likelihood of a shorter OS ( P =0.001) than patients with early stage (Stage I/II) cancer. Other clinical characteristics lost their predictive significance of OS when subjected to multivariate analysis, including high KLK6 expression ( P =0.215). When KLK13 expression was analysed with the Kaplan–Meier model, patients with high KLK13 expression had a shorter RFS than patients with low KLK13 expression ( P =0.027; Figure 3A ). The strength of association between KLK13 high expressing tumours and survival outcome is presented in a Cox regression model ( Tables 3 and 4 ). Univariate analysis showed high KLK13 as a significant predictor of recurrence ( P =0.030), indicating a 2.2-fold increased probability of recurrence compared to low KLK13 ( Table 3 ). When KLK13 expression was examined in a multivariate model, it retains the ability to significantly predict a shorter RFS ( P =0.047; Table 4 ). Late clinical stage and serous type cancer also significantly predict a shorter RFS ( P =0.001 and 0.024 respectively). When we examine KLK13 expression along with other factors in a multivariate model, late clinical stage was the only significant predictor of OS ( P <0.001).

The study analysed formalin-fixed, paraffin-embedded ovarian tissues from 106 cases of sporadic ovarian carcinoma diagnosed in the province of Newfoundland and Labrador, Canada between 1983 and 2002. Eight normal ovary samples were also obtained for comparison. Tissues were collected from pathology archives and selected based on hematoxylin and eosin stains reviewed by a pathologist. Clinical staging was performed using the standard International Federation of Gynecology and Obstetrics staging, with tumours graded as borderline, well differentiated (grade I), moderately differentiated (grade II), or poorly differentiated (grade III). Clinical history was obtained by review of patients’ medical records in accordance with Memorial University's human investigation committee protocol. Medical treatment of patients consisted of a total abdominal hysterectomy, bilateral salpingo-oophorectomy, omentectomy, and tumour staging. When cancer was not in stage 1A grade 1 or borderline, patients went on to receive chemotherapy. As the scope of this study spans from 1983 to 2002, chemotherapy regimes changed over this time. Before 1995/96 patients received cisplatin and cyclophosphamide, given for 6–9 cycles, at the discretion of the physician. After 1995/96 with the introduction of taxanes, treatments involved the combining of taxol with either cisplatin or carboplatin. A small number of patients with late stage (3C and 4) disease before 1998 were also treated with chemotherapy as a neoadjuvant pre-surgery. Sections were cut 4  μ m thick and dried on glass slides overnight. Sections were deparaffinised in xylene and re-hydrated through decreasing graded alcohols. Antigen retrieval was performed using a pressure cooker for 10 min in citrate buffer (pH 6.0). Slides were incubated overnight in primary antibody, washed twice with phosphate-buffered saline (pH 7.4) followed by either of two detection methods. Detection of KLK6 immunocomplex was carried out using Envision (Dako, Mississauga, Ontario, Canada), whereas localisation of KLK13 immunocomplex was performed with LSAB+, Link, and streptavidin reagents (Dako). Immune complexes were visualised by incubating with diaminobenzidine and sections were counterstained with hematoxylin. Immunolocalisation of KLK6 was carried out using a rabbit polyclonal antibody ( Diamandis et al , 2000b ), whereas KLK13 localisation was carried out using a mouse monoclonal antibody, clone 33.1 ( Kapadia et al , 2003 ). Five 10  μ M sections were cut from paraffin-embedded tissues for nucleic acid isolation. Total RNA was extracted using the High Pure RNA Paraffin kit (Roche, Indianapolis, IN, USA) according to the manufacturer's protocol. Total RNA concentration was determined spectrophotometrically at 260 n M and samples were stored at −80°C. The presence of high molecular weight total RNA was determined using a NanoDrop 1000 Spectrophotometer (NanoDrop Technologies Inc., Wilmington, DE, USA) and ethidium bromide staining of samples using formaldehyde gel electrophoresis. cDNA was synthesised using 2  μ g total RNA and Superscript First-Strand Synthesis System for RT–PCR (Invitrogen, Carlsbad, CA, USA) with random hexamers to ensure representation of all mRNA independent of polyadenylated tail. Quantitative real-time polymerase chain reaction (qRT-PCR) was performed using the ABI Prism 7000 (Applied Biosystems, Foster City, CA, USA). Primer/probe sets were purchased as pre-made TaqMan Assays on Demand for KLK13 , KLK6 , and glyceraldehyde 3-phosphate dehydrogenase ( GAPDH ; described in Supplementary Table 1 ). Thermal cycling conditions were according to the manufacture's protocol and all reactions were performed in triplicate. Relative quantification was defined as the amount of the specific mRNA normalised to a normal ovary as determined using the comparative cycle threshold ( C T ) method. The relative target gene expression was defined as 2 −ΔΔCT , where ΔΔ C T =Δ C T normal ovary –Δ C T ovarian cancer . A normal ovarian sample having the median level of expression for a specific target gene was chosen as the calibrator sample and used to normalise expression of all other samples (i.e. target gene expression=1). Δ C T is defined as C T target – C T GAPDH , where the target genes were KLK13 and KLK6 . Relative to the calibrator sample, target gene relative expression (RE) was classified as being low or high kallikrein expression. The expression level for determining high expressing samples was defined as one standard deviation above the mean value for all normal ovarian RE, for each specific target gene. Samples with an RE equal to or above this value were classified as high expression, whereas samples below the cut-off were classified as low expression. The cut-off for KLK6 was determined as 5.211 RE and KLK13 as 0.981 RE. Using a standard deviation above the normal mean RE of each target gene allowed us to utilise the inherent variation of each target gene expression to independently determine its cut-off value. All statistical analysis was performed with the SPSS statistical package for PC (version 13.0; SPSS Inc., Chicago, IL, USA). The relationships between KLK6 and KLK13 mRNA expression and patient clinical characteristics were analysed with a χ 2 -test. For survival analysis, two end points were examined; cancer recurrence (defined as either a local recurrence or metastasis) and death. These end points were used to calculate the recurrence-free survival (RFS) and overall survival (OS) respectively. Recurrence-free survival is defined as the time from first diagnosis to the time of first detected recurrence or metastasis. Overall survival is defined as the time from initial diagnosis to the time of death. Two survival models, the Kaplan–Meier and the Cox proportional hazard regression, were used for the analysis. The Kaplan–Meier model was used to examine survival between the patients expressing kallikreins at low or high levels, whereas significance was measured with the log-rank test. The Cox proportional hazard regression model, using both univariate and multivariate models, was used to determine the hazard ratio.

Unlike other reproductive malignancies, such as prostate cancer, ovarian cancer lacks a biomarker that may be used for screening. Currently, CA125, the only marker used in ovarian cancer patients, is reliable only for monitoring response to treatment and disease recurrence. The identification of early biomarkers for ovarian cancer may lead to novel therapeutic applications and potential screening tests. This study was aimed at examining the expression of KLK6 and KLK13 in ovarian cancer to determine their diagnostic or prognostic value. When we examined kallikrein protein expression by immunohistochemistry ( Figure 1 ), we found that although there appeared to be increased KLK6 and KLK13 expression in ovarian cancer tumours relative to normal OSE, there were no significant associations to clinical variables or survival. These results are similar to previous studies that found increased expression of KLK6 in primary pancreatic ductal adenocarcinoma ( Ruckert et al , 2008 ) and salivary gland tumours ( Darling et al , 2006 ), but found no significant association with survival. Kallikreins 6 and 13 are expressed in normal epithelium ( Petraki et al , 2001 , 2003 ), including the OSE as we have shown here. Ovarian cancer is thought to arise from either OSE or OSE cells bordering inclusion cysts ( Kaku et al , 2003 ), therefore it is expected that ovarian cancer cells would express a basal level of kallikrein expression. Given the semi-quantitative nature of immunohistochemistry and the fact that KLKs are secreted proteins, to see a significant difference in the cellular levels between normal OSE and ovarian cancer would require not only changes in KLK protein production, but also a significant change in the rate of kallikrein exocytosis. With these limitations in quantifying protein expression, we used qRT-PCR to assess kallikrein mRNA expression in ovarian cancer tissues. High KLK6 mRNA expression was associated with the presence of serous ovarian cancer and late stage disease. These results are similar to previous studies, which found increased KLK6 expression in ovarian cancers when compared to normal ovarian tissue ( Anisowicz et al , 1996 ; Tanimoto et al , 2001 ; Ni et al , 2004 ). Interestingly, previous studies have also associated high KLK6 serum levels with advanced ovarian cancer (stage III/IV) and serous tumour histology ( Hoffman et al , 2002 ; Shan et al , 2007 ). Ovarian serous tumours of borderline and low grade are thought to arise from a step-wise progression from adenoma to borderline tumour to carcinoma through the Ras-Raf signalling pathway ( Bell, 2005 ). Recently, KLK6 expression and secretion has been shown to be Ras dependent in a colon carcinoma cell line ( Henkhaus et al , 2008 ). Constitutively active mutant K-Ras resulted in enhanced colon cancer cell invasion through both laminin and Matrigel matrixes. Together, these data suggest that Ras-Raf mutations may increase the invasive potential of these borderline tumours through increased expression of KLK6. This may also hold true for advanced ovarian cancers, as we show high KLK6 expression is associated with later stage, more invasive, cancers. High-grade serous carcinomas have a notably different pathogenesis than low-grade serous tumours. High-grade serous tumours commonly have mutations in p53, BRCA 1, and/or BRCA2 ( Christie and Oehler, 2006 ). To date there have be no reports indicating a relationship between KLK6 and p53 or BRCA genes. When we looked at only high- and low-grade serous tumours, we found no significant differences in KLK6 expression ( P =0.498). Despite the apparent differences in the molecular oncogenesis of low- and high-grade serous tumours, their upregulation of KLK6 suggests a common pathway is activated in both types of tumours. Our study is the first to report KLK13 mRNA expression in normal ovary and ovarian cancer patients. The expression of KLK13 mRNA in normal OSE was extremely low, whereas 55% ovarian cancers examined had high KLK13 expression. These findings support Kapadia et al (2003) , who found serum levels of KLK13 were below levels of detection in healthy individuals, yet 50% ovarian cancer patients were positive for KLK13 ( Kapadia et al , 2003 ). Interestingly, 100% of our endometrioid cancer samples were high grade and expressed KLK13 at a high level. Although our sample size is small (six cases), this expression pattern warrants further investigation. Low-grade endometrioid carcinomas have been suggested to arise from endometriosis or borderline endometrioid tumours ( Obata et al , 1998 ), whereas high-grade endometrioid carcinomas have changes similar to high-grade serous carcinomas ( Bell, 2005 ; Giordano et al , 2008 ; Press et al , 2008 ) suggesting endometrioid cancers may represent two separate malignancies. This characteristic expression pattern of KLK13 may represent not only a novel marker to distinguish between high- and low-grade endometrioid cancers, but a unique pathway in which KLK13 may be involved in ovarian carcinogenesis. Since histological types of ovarian cancers present as a spectrum of pathologically and histologically different phenotypes and may represent cancers of different origins ( Bell, 2005 ), the possibility that a panel of markers, rather than a single marker, may improve the sensitivity and specificity of detecting ovarian cancer at an early stage while it is treatable. Zheng et al (2007) found that a group of kallikrein-related peptidases, including KLK6 and KLK13, in multi-parametric combinations with other biomarkers and clinical variables can significantly predict prognosis and response to treatment in ovarian cancer patients. A similar model, including a number of kallikrein-related peptidases, has been proposed for non-small-cell lung carcinoma ( Planque et al , 2008 ). This study confirms both KLK6 and KLK13 are overexpressed in ovarian cancer and are useful predictors of poor prognosis in ovarian cancer patients. This study is the first to report upregulation of KLK13 mRNA in ovarian cancer patients and indicated that KLK13 may represent a specific marker for endometrioid carcinoma. These findings support the potential role as kallikrein-related peptidases 6 and 13 as novel ovarian cancer biomarkers and may, in the future, offer targets for therapeutic applications.