Transcriptional upregulation of human tissue kallikrein 6 in ovarian cancer: clinical and mechanistic aspects

The human tissue kallikrein family (KLK for protein; KLK for gene) includes 15 members. Twelve kallikreins, including KLK6, are concurrently upregulated in ovarian cancer. However, the mechanism of this phenomenon remains unclear. In this study, we measured KLK6 expression in a large series of ovarian tissue cytosols and examined possible mechanisms of KLK6 up-regulation in ovarian cancer. Using a newly developed enzyme-linked immunosorbent assay (ELISA) with two monoclonal antibodies, we quantified KLK6 expression in ovarian tissue cytosols, and confirmed the upregulation of KLK6 in ovarian cancer and its unfavourable prognostic value. We then examined KLK6 mRNA expression using reverse transcription–polymerase chain reaction and established its good concordance with KLK6 protein expression. This finding suggested that the KLK6 gene is under transcriptional regulation. We then scrutinised a few mechanisms that could explain KLK6 upregulation. The relative abundance of two KLK6 mRNA transcripts was studied; we found the same differential expression pattern in all samples, regardless of KLK6 levels. Genomic mutation screening of all exons and the 5′-flanking region of the KLK6 gene identified two linked single-nucleotide polymorphisms in the 5′-untranslated region, but neither correlated with KLK6 expression. Ovarian cell lines were separately treated with five steroid hormones. None of the treatments produced significant effects on KLK6 expression. We conclude that KLK6 is transcriptionally upregulated in ovarian cancer, but probably not through alternative mRNA transcript expression, genomic mutation, or steroid hormone induction.

The human tissue kallikreins (KLKs) are a family of serine proteases that are aberrantly expressed in several cancer types. Most strikingly, 12 kallikreins (KLK2, KLK3, KLK4, KLK5, KLK6, KLK7, KLK8, KLK10, KLK11, KLK13, KLK14, and KLK15) are concurrently upregulated in ovarian cancer . Many of these KLKs hold promise as prognostic and diagnostic biomarkers (Yousef et al, , 2003bDiamandis et al, 2002Diamandis et al, , 2003Borgono et al, 2003;Kishi et al, 2003;Luo et al, 2003;Scorilas et al, 2004). Several of them have also been implicated in cancer progression and metastasis (Bernett et al, 2002;Magklara et al, 2003;Becker et al, 2004). Evidently, kallikrein upregulation in ovarian cancer has significant relevance to cancer detection, management, and treatment. Understanding the mechanisms underlying their upregulation will allow for a better definition of kallikreins' role in ovarian cancer pathophysiology and their clinical utility as prognostic and/or diagnostic biomarkers or therapeutic targets.
The human tissue kallikrein genes (KLKs) are located in tandem on chromosome 19q13.4, and show significant structural homology . Hence, the 12 kallikreins concertedly overexpressed in ovarian cancer may share a common regulation machinery. On this basis, we investigated the regulation of one kallikrein, human tissue kallikrein 6, as a model system. The findings may also prove applicable to other kallikreins. The KLK6 gene spans 10.5 kb of genomic sequence on the kallikrein locus ( Figure 1A) . The KLK6 protein is normally expressed in adult brain, salivary gland, lung, colon, kidney, breast, uterus, fallopian tube and placental tissues, and secreted into numerous biological fluids (Diamandis et al, 2000;Petraki et al, 2001). Ample evidence supports the overexpression/increase of human KLK6 in both ovarian tumours and presurgical serum of patients with ovarian cancer (Hoffman et al, 2002;Diamandis et al, 2003;Yousef et al, 2003a). Nevertheless, few experiments have been conducted to delineate the processes underlying this upregulation.
In this study, we first quantified KLK6 expression, using a newly developed immunoassay with two monoclonal antibodies, in a relatively large series of ovarian cancer tissue specimens, and confirmed the prognostic value of KLK6 for ovarian cancer. We then assessed the correlation between KLK6 mRNA and KLK6 protein expression in order to determine whether KLK6 is under transcriptional or translational upregulation. Lastly, we examined the role of alternative transcripts, genetic aberrations, and steroid hormones in the upregulation of KLK6 in ovarian cancer.
The human tissue kallikreins exhibit considerable transcriptomic complexity. Through usage of alternative promoters and other mechanisms, each KLK gene can generate numerous alternative transcripts (Landry et al, 2003;Pampalakis et al, 2004). Some of these transcript variants have been shown to be specifically overexpressed in ovarian cancer (Dong et al, 2001(Dong et al, , 2003Magklara et al, 2001). KLK6 has four alternative transcripts encoding for the full-length KLK6 protein and four that may encode for truncated proteins (Kurlender et al, 2005). The longest KLK6 transcript (GenBank accession no. NM002774), also known as the 'classical transcript', consists of 1512 nucleotides and seven exons. KLK6 'alternative transcript 1' has 1517 nucleotides (GenBank accession no. AY318867); it lacks exon 1 but contains a unique sequence at the 5 0 -end of exon 2, denoted as exon 2A ( Figure 1B). Christophi et al (2004) found that these two mRNA transcripts were expressed in a tissue-specific manner and were differentially regulated in response to central nervous system injury. Whether or not an analogous situation occurs in ovarian cancer has not been investigated. To this end, we differentially amplified these two transcripts and examined their expression patterns in ovarian tumour tissues.
Another potential regulatory mechanism of the kallikreins is genetic variation. Even though intragenic mutations have not been identified for the kallikrein genes examined to date, KLK3 (Majumdar and Diamandis, 1999) and KLK10 (Liu et al, 1996), a plethora of single-nucleotide polymorphisms (SNPs) exist within KLK coding and promoter/enhancer regions and may have clinical significance. For example, SNPs in the KLK3 promoter may account for individual variation in serum prostate-specific antigen levels and even cancer susceptibility (Tsuyuki et al, 1997;Xue et al, 2000;Cramer et al, 2003). Similarly, a nonsynonymous SNP in exon 3 of KLK10 is associated with a higher prostate cancer risk (Bharaj et al, 2002). The KLK6 locus in ovarian cancer patients has not previously been scrutinised for genetic aberrations. Hence, we sequenced all KLK6 exons and the 5 0 -flanking region for ovarian tumours with various KLK6 levels to determine if genetic aberrations may account for the upregulation of KLK6 in ovarian cancer.
Lastly, various in vitro and in vivo studies collectively demonstrate that most, if not all, kallikreins are under steroid hormone regulation in endocrine-related tissues and cell lines . Whereas some kallikreins, such as KLK3 (Riegman et al, 1991) and KLK4 (Nelson et al, 1999), are regulated by androgens, others, such as KLK5 , are more responsive to oestrogens. A role for the steroid hormones in the regulation of KLK6 was first suggested by the discovery of several hormone-related response elements, including sterol regulatory element binding protein 1 and 2, progesterone receptor binding site and cAMP response elementbinding protein in the 5 0 -flanking region of the KLK6 gene (Christophi et al, 2004). It was subsequently supported by the findings that KLK6 was significantly up-regulated by norgestrel, oestradiol and DHT in the breast cancer cell line BT474 . In this study, we examined whether or not steroid hormones regulated KLK6 expression in ovarian cancer cell lines.
Our results confirmed the value of KLK6 as an unfavorable prognostic biomarker for ovarian cancer. We also found KLK6 to be under transcriptional regulation. Our data have further suggested a minor role for alternative transcripts, genetic mutations and steroid hormones in the regulation of human kallikrein 6 in ovarian cancer.

Tissue sample collection
We examined 259 patients with ovarian cancer, 49 with benign ovarian conditions (including endometriosis, mucinous cystadenomas, serous cystadenomas, dermoid cysts, ovarian benign teratomas and corpus luteum), 43 with nonovarian tumours that metastasised to the ovary, and 34 apparently normal women. Participants' age ranged from 19 to 89 years, with a median of 58 years. All tissue samples were collected between April 1988 and April 2003 at the Gynecologic Oncology and Breast Cancer Unit, University of Turin, Italy. Identical specimen collection and processing protocols were used for all participants. During surgery, histologic examination was performed on the ovarian tissues through intrasurgery frozen section analysis, which allowed representative portions of each tumour containing more than 80%  (Pettersson, 1994) and graded according to the Day et al, (1975) protocol. Tumour histotypes were classified based on the World Health Organization (WHO) and FIGO recommendations (Serov et al, 1973). Patients with ovarian cancer at all stages (I -IV) and grades (1 -3) were represented in our study. Of the 259 ovarian tumours, the majority (110; 42%) were of serous papillary histotype, followed by endometrioid (46; 18%), undifferentiated (33; 13%), mucinous (20; 8%), clear cell (18; 7%), or other nonepithelial types (24; 9%).
After surgery, we monitored the patients with ovarian cancer for clinical response to chemotherapy and survival outcomes for a median duration of 52 months. Follow-up information was available for 232 patients, of whom 147 (63%) had relapsed and 117 (50%) had died.
All investigations were carried out according to the ethical standards of the Helsiniki Declaration of 1975Declaration of (revised in 1983, and were approved by the Institute of Obstetrics and Gynecology, Turin, Italy, and the Institutional Review Board of Mount Sinai Hospital, Toronto, Canada.

Preparation of cytosolic extracts
Snap-frozen tissues (20 -100 mg) were homogenised in liquid nitrogen to a fine powder and added to 10 volumes of extraction buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 5 mM EDTA, and 1% NP-40 surfactant). The resulting mixtures were incubated on ice for 30 min, with repeated vortexing every 10 min, and then centrifuged at 14 000 r.p.m at 41C for 30 min. The supernatant (cytosolic extract) was collected and stored at À801C until further experiments. Total protein concentration of the extracts was determined using the bicinchoninic acid method. Bovine serum albumin was used as standard (Pierce Chemical Co., Rockford, IL, USA).

Measurement of KLK6 and CA125 protein expression in ovarian cytosolic extracts
The concentration of KLK6 was quantified with a highly sensitive and specific 'sandwich-type' immunoassay recently developed in our laboratory. Two KLK6-specific monoclonal mouse antibodies were used: the coating antibody (clone 27-4) and the detection antibody (clone E24). The assay has a detection limit of 0.05 mg l À1 , a dynamic range up to 20 mg l À1 , and does not cross-react with any other members of the kallikrein family. White polystyrene microtiter plates were first coated with 100 ml of the coating antibody solution (5 mg l À1 of the anti-KLK6 monoclonal antibody clone 27-4 in 50 mM Tris-HCl buffer, 0.05% sodium azide, pH 7.8). The plates were incubated overnight at room temperature and then washed three times with the washing buffer (5 mM Tris-HCl buffer, 150 mM NaCl, 0.05% Tween-20, pH 7.8). After washing, assay standards (KLK6 recombinant protein produced in-house (Diamandis et al, 2000)) or ovarian cytosolic extracts were added to the wells in duplicates (100 ml well À1 ) after being diluted two-fold in assay buffer (50 mM Tris-HCl buffer, 6% bovine serum albumin, 0.01% goat globulin, 0.005% mouse globulin, 0.1% bovine globulin, 0.5 M KCl, 0.05% sodium azide, pH 7.8). The plates were incubated for 2 h with shaking at room temperature and washed six times. Subsequently, 100 ml of the biotinylated detection antibody solution (250 mg l À1 anti-KLK6 monoclonal antibody clone E24 in assay buffer) was added to each well. The plates were incubated for 1 h at room temperature with shaking and washed six times. Then, 100 ml of alkaline phosphatase-conjugated (ALP) streptavidin solution (Jackson ImmunoResearch) diluted 20 000-fold in bovine serum albumin buffer (6% bovine serum albumin, 50 mM Tris-HCl buffer, 0.05% sodium azide, pH 7.8) was added to each well. The plates were incubated for 15 min with shaking at room temperature and washed six times. Lastly, 100 ml of substrate buffer (0.1 M Tris-HCl buffer, 0.1 M NaCl, 1 mM MgCl 2 , pH 9.1), containing 1 mM diflunisal phosphate (DFP), was added to each well and incubated for 10 min with shaking at room temperature. Immediately after incubation, 100 ml of developing solution (1 M Tris-HCl buffer, 0.4 M NaOH, 2 mM TbCl 3 , 3 mM EDTA) was added to each well and incubated for 1 min with shaking at room temperature. The fluorescence was measured with a time-resolved fluorometer, the Cyberfluor 615 Immunoanalyzer (MDS Nordion), as described elsewhere (Christopoulos and Diamandis, 1992). The KLK6 concentrations in ng ml À1 were converted to ng of KLK6 mg À1 of total protein to adjust for the amount of tumour tissue extracted.
CA125 levels (KU mg À1 ) of ovarian tissue samples were measured with the Immulite 2000 assay (Diagnostic Products Corporation, Los Angeles, CA, USA).

Statistical analysis
The distribution of KLK6 concentration in the ovarian tumour cytosols was non-Gaussian. Therefore, we used the nonparametric Mann -Whitney U-test to determine differences among the four groups of samples. This test treated KLK6 concentration in the tumour cytosols (ng mg À1 of total protein) as a continuous variable. We also assessed the association between KLK6 and CA125 levels by determining the Spearman rank correlation coefficient (r) and associated P-values. The ovarian cancer patients were classified as KLK6-positive or KLK6-negative using the median (2.83 ng mg À1 ) as the cutoff point. The relationship between KLK6 expression and several clinicopathologic variables was analysed with the w 2 test and the Fisher's exact test, as appropriate.
We then assessed the impact of KLK6 on patient survival by calculating the hazard ratio (HR), which is the relative risk of relapse or death in the KLK6-positive group, using the Cox univariate and multivariate proportional hazard regression models (Cox, 1972). Whereas progression-free survival (PFS) was defined as the time interval between the first surgery and the identification of recurrence or metastatic disease, overall survival (OS) was defined as the time interval between the first surgery and death. The multivariate models were adjusted for KLK6 expression in tumours and other clinicopathologic variables that may affect survival, including age, tumour grade, CA125 level, and histotype. We only included patients for whom the status of all variables was known in the multivariate models. Lastly, we constructed Kaplan and Meier (1958) PFS and OS curves for the KLK6-positive and KLK6-negative patients. The differences between the survival curves were assessed for statistical significance using the log rank test (Mantel, 1966). Patients were also classified as CA125-positive and CA125-negative using the median 1330 U mg À1 as cutoff, and we constructed Kaplan -Meier PFS and OS curves to evaluate the prognostic significance of CA125.

RNA extraction and RT -PCR
We selected eight tumour tissues (H1 -H8) that over-expressed KLK6 (mean: 12.44 ng mg À1 ), eight samples (L1 -L8) that expressed minimal KLK6 (mean: 0.06 ng mg À1 ), four benign ovarian samples (B1 -B4; mean KLK6 expression: 0.05 ng mg À1 ), and one normal ovarian sample (N1; KLK6 expression: 0.02 ng mg À1 ). Total RNA was isolated from these samples using the TRIzol s method (Life Technologies Inc., Gaithersburg, MD, USA) as directed by the manufacturer. The yield and purity were determined spectrophotometrically by measuring the absorbance of aliquots at 260 and 280 nm. Two micrograms of total RNA were then reverse transcribed using the SuperScript t preamplification system (Life Technologies Inc.) in a final volume of 20 ml. As control for the reverse transcription -polymerase chain reaction (RT -PCR), we used commercially available human brain total RNA (Clonetech Inc., Mountain View, CA, USA). To test the success of our reverse transcription, 1 ml of each cDNA product was PCR-amplified using primers specific for the housekeeping gene, b-actin (for primer sequences see Table 1), and visualised on 1.8% agarose gels stained with ethidium bromide.

PCR amplifications
To amplify the mRNA transcript variants that encode the fulllength KLK6 protein (GenBank accession no. NM002774, AY318867, AY318869, and BC015525), we designed primers (for sequences see Table 1; Figure 1B) that anneal to a common region for all four variants (nucleotides 257 -709; numbers refer to GenBank accession no. AY318867).
One microliter of first-strand cDNA template was amplified in a 25 ml reaction mixture containing 10 Â PCR buffer (Qiagen Inc., Mississauga, ON, USA), 100 mM dNTPs (deoxynucleoside triphosphates), 2.5 U of HotStar Taq DNA polymerase (Qiagen Inc.) and 100 ng of the appropriate primers in an Eppendorf thermocycler. Cycling conditions were as follows: enzyme activation at 951C for 15 min; 32 cycles of denaturation at 951C for 30 s, annealing at 651C for 30 s, and extension at 721C for 30 s; followed by a final extension at 721C for 10 min. The KLK6 PCR product was visualised on 1.8% agarose gels stained with ethidium bromide. Gels were photographed under UV light with the Speedlight Gel Documentation System (Lightools Research, Encinitas, CA, USA). Two independent observers examined the images to determine the presence or absence of KLK6 PCR-generated bands. Each amplification experiment was performed twice to ensure reproducibility of data.
To differentially amplify the 'classical transcript' and 'alternative transcript 1' (GenBank accession no. NM002774 and AY318867, respectively), which differ only in their 5 0 -untranslated regions (UTRs), we used two unique forward primers and the same reverse primer (for sequences, see Table 1; Figure 1B). The PCR reactions amplified nucleotides 1 -285 for the 'classical transcript' (numbers refer to the sequence NM002774), and nucleotides 1 -290 for 'alternative transcript 1' (numbers refer to the sequence AY318867). We ensured that the two primer sets had comparable annealing efficiency by conducting parallel PCR reactions with each primer set and equal amounts of plasmid DNA. These reactions generated PCR bands of similar intensity ( Figure 5B). Cycling conditions for both amplifications were the same as described above, except for the annealing temperature, which was 611C for both reactions. Only tumour samples H1 -H8, which expressed high kallikrein 6 protein and mRNA, were used in this experiment. Normal human brain and ovarian cDNA templates were used as controls.

Genomic DNA extraction and mutation screening
An even representation of ovarian tumour tissues expressing high KLK6 (n ¼ 9; mean ¼ 11.72 ng mg À1 ) and low KLK6 (n ¼ 9; mean-0.09 ng mg À1 ), benign ovarian tissues (n ¼ 8; mean: 0.09 ng mg À1 ), non-ovarian tumours that metastasised to the ovary (n ¼ 6; mean: 0.04 ng mg À1 ), and normal ovarian tissues (n ¼ 13; mean: 0.06 ng mg À1 ) were selected for genomic DNA extraction and sequencing. DNA was extracted using the QIAamp DNA mini Kit (Qiagen Inc.) according to the manufacturer's protocols, and its yield and purity were determined spectrophotometrically by measuring the absorbance of aliquots at 260 and 280 nm.
To design primers for mutation screening, we used KLK6 'alternative transcript 1' as template, as it was found to be the dominant transcript in the ovaries. We designed seven sets of primers amplifying all exons and their flanking intronic sequences, as well as the 5 0 -flanking region (601 bp upstream of the putative transcription start site, numbers refer to the KLK6 genomic sequence, GenBank accession no. AF149289). Sequences of all primers are shown in Table 1. Polymerase chain reaction was carried out with 2 mg of genomic DNA in a 25 ml reaction mixture containing 10 Â reaction buffer, 100 mM of dNTP, 1 U of Pfu Turbo Polymerase (Stratagene, La Jolla, CA, USA), and 100 ng of the appropriate primers. Pfu Turbo DNA polymerase was employed because of its high fidelity. The PCR cycling conditions were as follows: 3 min of enzyme activation at 951C, 35 cycles of 1 min denaturation at 951C, 45 s of annealing at a primer-dependent temperature (Table 1), and 1 min of extension at 721C, followed by 10 min of final extension at 721C. The extension time for each cycle was prolonged to 1 min and 30 s when amplifying the 5 0 -flanking region and exon 7, to compensate for the length of these PCR products. The success of the PCR reactions were evaluated by running 5 ml of each PCR product on 1.8% agarose gels stained with ethidium bromide. The remaining of the PCR products was purified by treatment with exonuclease I (0.25 U) and shrimp alkaline phosphatase (0.25 U) (USB, Cleveland, OH, USA). Mixtures were incubated at 371C for 15 min, and then at 851C for 15 min. After purification, PCR products were directly sequenced using an automated DNA sequencer. The PCR product was sequenced again in the opposite direction if sequencing results showed a departure from the published sequence in GenBank (accession no. AF149289).

Cell lines and hormonal stimulation experiments
BG-1, CaOV3, and OVCAR-3 are ovarian cancer cell lines, purchased from the American Type Culture Collection (ATCC, Rockville, MD, USA) and grown in RPMI media (Invitrogen,  (Kruk and Auersperg, 1992). The hOSE cells were cultured in medium 199 : MCDB105 (1 : 1) (Life Technologies Inc.) with 10% FBS (HyClone). All cell lines were cultured in plastic flasks to near confluency, aliquoted into six-well tissue culture plates, and then cultured to 75% confluency. Twenty-four hours before the hormonal stimulation treatments, the culture medium was changed into each cell line's respective medium containing 10% charcoal-stripped fetal bovine serum. Cells were grown in the new media for 24 h.
The steroid hormones used were estroadiol, aldosterone, dihydrotestosterone (DHT), dexamethasone, and norgestrel. Each steroid hormone was dissolved in 100% ethanol, and added to the culture media in each well at a final concentration of 10 À8 M and less than 1% ethanol. Controls were the cells stimulated with an equal amount of ethanol only. All cells were grown for 7 days after stimulation and the cell culture supernatant was collected for KLK6 measurement using the aforementioned immunoassay. These experiments were repeated three times to ensure reproducibility of results.

Distribution of KLK6 concentration in ovarian tissues
The KLK6 protein concentration in normal ovarian tissues ranged from 0.01 to 4.11 ng mg À1 of total protein, with a median of 0.05. Both benign ovarian tissues (median: 0.09; range: 0.01 -16 ng mg À1 ) and nonovarian tumours that metastasised to the ovary (median: 0.09; range: 0.01 -8.56 ng mg À1 ) also had relatively low KLK6 expression. Conversely, the protein concentration of KLK6 in the 259 ovarian tumour cytosols ranged from 0.01 to 65 ng mg À1 of total protein, with a median of 2.83 (Table 2; Figure 2A). We performed the nonparametric Mann-Whitney Utest to analyse the differences in KLK6 expression among the four groups of ovarian tissue samples. KLK6 expression in ovarian tumour cytosols was found to be 57-fold higher than normal and 31-fold higher than benign and nonovarian metastatic tumours (Po0.0001). We used the median as the cutoff value to categorise the ovarian tumours as KLK6-positive or KLK6-negative.

Relationships between KLK6 status and other clinicopathologic variables
We categorised KLK6-positive and KLK6-negative patients according to clinicopathologic variables including tumour stage, grade, histotype, debulking success and response to chemotherapy ( Table 3). The strength of the relationships between KLK6 and these variables was evaluated with either the w 2 test or the Fisher's exact test, as appropriate. We found that patients with KLK6positive tumours more frequently had late stage (stage III/IV) disease (Po0.001) ( Figure 2B), higher tumour grades (P ¼ 0.04), suboptimal debulking (P ¼ 0.002), and serous and undifferentiated histotypes (Po0.001) ( Figure 2C); but KLK6 expression had no relationship with response to chemotherapy. A weak positive correlation between tissue CA125 and KLK6 expression in ovarian cancer (Spearman correlation r s ¼ 0.566; Po0.001) was also observed (Figure 3).

Univariate and multivariate survival analysis
The association between KLK6 protein expression and patient survival is summarised in Table 4. Univariate Cox regression analysis showed that KLK6-positive patients were at an increased risk of relapse (HR ¼ 1.56; P ¼ 0.008) and death (HR ¼ 1.46; P ¼ 0.045) in comparison with KLK6-negative patients. When KLK6 expression was treated as a continuous variable (after logarithmic transformation of data), its correlation with patient survival remained significant (HR ¼ 1.45 for PFS, Po0.001; and 1.37 for OS, P ¼ 0.004). Other variables such as histological type and grade, but not age, had even higher HRs for both PFS and OS. In multivariate Cox regression analysis, the relationship between KLK6 status and survival outcome was no longer significant. The same was true for patient age. Tumour grade remained a significant predictor of both PFS and OS, though histological type was no longer associated with OS. Kaplan -Meier survival curves showed significant association between KLK6 status and both PFS and OS ( Figure 4A and B). On the contrary, Kaplan -Meier survival curves showed that CA125 did not bear prognostic significance ( Figure 4C and D).

Expression of KLK6 mRNA transcripts in ovarian tumours
The collective expression of all KLK6 mRNA transcripts encoding the full-length protein was assessed in 16 ovarian tumour samples, eight with KLK6 overexpression (H1 -H8) and eight with minimal KLK6 expression (L1 -L8). Whereas KLK6 mRNA expression was significant in H1 -H8, it was undetectable in L1 -L8 ( Figure 5A). For some of the high-KLK6 expressing samples, we observed an  additional, shorter, and much weaker band generated by PCR amplification. After sequencing, we confirmed this band to represent a previously identified splice variant (GenBank accession no. AY279383), which lacks exon 4 (Kurlender et al, 2005). Four benign and one normal ovarian tissue samples expressing low KLK6 were used as controls. All had undetectable KLK6 mRNA expression ( Figure 5A). We also examined the relative expression of two most abundant KLK6 mRNA transcripts. For each of tumour samples H1 -H8, we assessed the differential expression of the 'classical transcript' and 'alternative transcript 1'. Normal brain and ovarian cDNA templates were used as controls. For all samples studied, excluding normal ovarian tissues which showed no expression for either transcript, 'alternative transcript 1' was by far the dominant transcript (GenBank accession no. AY317768) ( Figure 5B).
SNPs in the 5 0 -UTR of KLK6 'alternative transcript 1' As 'alternative transcript 1' was found to be the more abundant transcript ( Figure 5B), we used it as the template for our mutation screening experiments. We amplified and sequenced all KLK6 exons and the 5 0 -flanking region. No mutations were found in any of the coding exons of KLK6. However, we identified two previously unpublished single nucleotide polymorphisms (SNPs) located in a (GT) 4 microsatellite in the 5 0 -untranslated region (exon 2A). The SNPs are located at 55 and 60 bp downstream of the putative transcription start site (Pampalakis et al, 2004), and 195 and 190 bp upstream of the start codon, respectively. G was the more frequent allele for both SNPs. The two SNPs were in linkage disequilibrium (the dependence of an allele at one locus on alleles at another locus), as the T allele was found at the second SNP only if the T allele was also present at the first SNP. We observed six different haplotypes across all samples studied. None of the haplotypes significantly correlated with KLK6 expression or sample type.
According to the published GenBank sequence AY318876, the microsatellite in this region consists of four GT repeats (GT) 4 . However, according to our findings, the most prevalent haplotype (51%) has a (GT) 3 microsatellite owing to a single nucleotide change from T to G at basepair 55. As all tumour samples were collected from Italian subjects, we examined if the difference in the microsatellite length was due to race. We sequenced three normal ovarian tissues obtained from Mount Sinai Hospital, Ontario, Canada, and found them to also have Haplotype B.

Hormonal regulation of KLK6 in ovarian cancer
We examined the effects of steroid hormones on KLK6 expression in BG-1, CaOV3 and OVCAR-3 epithelial ovarian cancer cell lines, as well as the hOSE cell line. As illustrated in Figure 6, none of the steroid hormones produced any significant changes in KLK6 expression for the ovarian cancer cell lines. The hOSE cell line did not produce detectable KLK6 under any treatment conditions (data not shown).

DISCUSSION
Ovarian cancer is the most lethal gynaecologic malignancy in developed countries (Mor et al, 2005). The high morality rate of this disease could be attributed to difficulties underlying its early detection, the disease's intrinsic aggressiveness, and our meagre understanding of its aetiology. The development of novel diagnostic and prognostic biomarkers will facilitate early detection and more individualised and efficient treatment plans for ovarian cancer patients. Of the plethora of novel potential biomarkers identified in recent years, those that also have clinical relevance to cancer initiation and progression are the most valuable, as they may serve the dual role as biomarkers as well as therapeutic targets. Serine proteases are among these newly identified tumourassociated biomarkers (Duffy, 1996). Accumulating evidence suggests that the human tissue kallikreins, a family of secreted serine proteases ubiquitously expressed in human tissues and biological fluids, have prognostic and diagnostic value in various cancer types . Some KLKs have also been shown to play a role in extracellular matrix degradation, which in turn facilitates tumour invasion and metastasis . Human kallikrein 6, in particular, is elevated in presurgical serum of ovarian cancer patients , as well as overexpressed in ovarian tumour tissues (Hoffman et al, 2002), and bears prognostic and diagnostic significance. KLK6 is a trypsin-like enzyme that can degade in vitro laminin and fibronectin (Bernett et al, 2002), as well as other basic constituents of the extracellular matrix (ECM) and the basement membrane (Magklara et al, 2003), and could be linked to tumour cell growth and malignant transformation. Hence, KLK6 may also serve as a therapeutic target for ovarian cancer. To further clarify KLK6's clinical utility and its role in ovarian cancer pathophysiology, we aimed to delineate the molecular processes underlying KLK6 overexpression in ovarian cancer.
We first conducted a validation study to confirm the prognostic significance of KLK6 in a large group of ovarian tumour tissues using a new immunoassay of higher sensitivity and specificity than previously reported (Diamandis et al, 2000). KLK6 protein was expressed at low levels in normal and benign ovarian tissues. Conversely, the KLK6 expression in ovarian tumour cytosols was significantly upregulated. This increase correlated with cancer stage ( Figure 2B). Metastatic tumours to the ovary from primary gastrointestinal, endometrial, uterine, or breast cancers expressed low levels of KLK6 that were comparable with benign ovarian tissues. Hence, the up-regulation of KLK6 seems to be specific to ovarian cancer. Figure 2C indicates that most tumours with KLK6 up-regulation were of epithelial origin, and especially of serous and undifferentiated histotypes.
Univariate Cox regression analyses confirmed that KLK6 was significantly correlated with survival (Table 4). Kaplan -Meier survival curves confirmed that KLK6 was a significant and much stronger prognostic biomarker than CA125 for both PFS and OS (Figure 4). In multivariate Cox regression models adjusted for tumour grade, histotype, CA125 and patient age, the relationship between KLK6 expression and patient survival was no longer significant. Even though KLK6 may not be a strong independent prognostic indicator for ovarian cancer, it appears to be a surrogate marker for other clinicopathological variables such as tumour stage, grade, histotype and debulking success. Recent evidence indicates that multiparametric strategies that merge diagnostic and prognostic information provided by several individual biomarkers yield more informative and accurate medical predictions (Woolas et al, 1993(Woolas et al, , 1995Zhang et al, 1999;Mor et al, 2005). Therefore, KLK6 could be used in conjunction with other kallikreins, as well as non-kallikrein biomarkers in future multiparametric prognostic tests for ovarian cancer.
To delineate the processes underlying the overexpression of KLK6 in ovarian cancer, we first determined whether KLK6 was under transcriptional or translational upregulation. We assessed KLK6 mRNA expression in tumours with high or low KLK6 protein concentration and correlated these two parameters. Excellent concordance between KLK6 protein and mRNA expression was observed ( Figure 5), rendering evidence for transcriptional regulation.
We evaluated the differential expression of the two most abundant transcripts of KLK6. Previous studies showed that 'alternative transcript 1' was expressed in both peripheral tissues (though the ovary was not examined) and the central nervous system, whereas the 'classical transcript' was only expressed in the central nervous system (Christophi et al, 2004;Pampalakis et al, 2004). We confirmed that the 'classical transcript' was highly expressed only in the brain, and weakly expressed or absent in all ovarian tissues. KLK6 'alternative transcript 1' was the dominant transcript in all samples examined ( Figure 5). Therefore, differential expression of alternative transcripts cannot account for KLK6 upregulation in ovarian cancer.
Our genomic mutation screening experiments found no mutations except two novel exonic SNPs within a (GT) 4 microsatellite in the 5 0 -UTR of 'alternative transcript 1' ( þ 55 G/T, þ 60 G/T). The SNPs were not located in any putative promoter region or known steroid hormone response elements regulatory regions (Christophi et al, 2004). No consensual changes were observed within any sample types. Previous experiments have indicated a tendency for SNPs in the KLK locus to exhibit strong linkage disequilibrium (Cramer et al, 2003;. Indeed, these two SNPs were linked to each other. Lastly, we evaluated the role of steroid hormones in the regulation of KLK6 in ovarian cancer. Sex steroid hormones have previously been implicated in the pathogenesis and growth regulation of ovarian cancers (Hoover et al, 1977;Langdon et al, 1990;Godwin et al, 1993;Chien et al, 1994). Hormonal regulation of KLKs in ovarian cancer has only been demonstrated in the ovarian cancer cell line, BG-1, where KLK9 was under the regulation of oestrogen and progesterone . In this study, we found that steroid hormones play a minor role in the regulation of KLK6 in ovarian cancer. It remains to be elucidated why steroid hormones regulate KLK6 in breast cancer cell lines such as BT474 , but not in receptor-positive ovarian cancer cell lines.
In conclusion, we confirmed that KLK6 is significantly upregulated in ovarian cancer and serves as an unfavourable prognostic biomarker. We also found that KLK6 is under transcriptional regulation. Nevertheless, the overexpression of KLK6 in ovarian cancer is probably not due to differential expression of alternative transcripts, mutations in the exons and proximal 5 0 -flanking region, or steroid hormone regulation.
Various other transcriptional regulation mechanisms remain to be tested for their roles in the upregulation of the KLKs in ovarian cancer. As subgroups of kallikreins are frequently coexpressed in tissues, transcription of KLK genes may be regulated by cis-acting locus control regions (Li et al, 1999). In fact, such a mechanism has been implicated in controlling the salivary gland-specific expression of the rat kallikrein gene family (Smith et al, 1992). Future studies should investigate whether a similar mechanism is at work   -H8). N: amplification using primers for the 'classical transcript'; A: amplification using primers for the 'alternative transcript 1'. KLK6 c denotes both 'A' and 'N' amplicons, which are 285 and 290 bp, respectively. Brain and ovarian normal cDNA were used as controls. Note predominance of 'alternative transcript 1' expression in all cases, as well as in the brain tissue (control). PN1, PN2 were two samples of plasmid DNA containing 'N' amplicon as insert, reamplified using 'N' primers. PA1, PA2 were two samples of plasmid DNA containing 'A' amplicon as insert, reamplified using 'A' primers. Note that when starting material was the same, the two sets of primers generated PCR bands of similar intensity.
in human ovaries. Furthermore, the promoter region of 'alternative transcript 1' has not been fully elucidated, and the actual transcription start site is not ascertained. Characterisation of the KLK6 promoter would facilitate future efforts to study the regulation of KLK6, including studying the roles of transcription factors. Lastly, chromosome 19q13 is reportedly rearranged in numerous types of solid tumours (Whang-Peng et al, 1984;Tanaka et al, 1989;Bello and Rey, 1990;Pejovic et al, 1991Pejovic et al, , 1992Jenkins et al, 1993). Potential gene amplification of KLK6 in ovarian cancer has also been suggested (Ni et al, 2004). Therefore, the role of chromosomal rearrangements of the KLK locus in cancer should be examined in future investigations. Ovarian cancer cell lines BG-1, CaOV3, OvCAR3 were stimulated with 10 À8 M of estradiol, aldosterone, DHT, dexamethasone, norgestrel, or alcohol (control). KLK6 protein concentration in supernatants was quantified by immunoassay. No major changes were seen with any steroid hormone stimulation. Error bars represent standard errors of the mean.