Variants of the hK2 Protein Gene (KLK2) Are Associated with Serum hK2 Levels and Predict the Presence of Prostate Cancer at Biopsy

Materials and Methods

Study subjects. Patients were drawn from a consecutive series of 1,437 men who were referred to the University of Toronto Prostate Centres, between June 1998 and June 2000, either because of a PSA value ≥ 4.0 ng/mL or because of an abnormal digital rectal exam (DRE). No patient had a prior history of prostate cancer. Study subjects have been described elsewhere (3).

Blood samples were collected before clinical prostate examination and were banked at −70°C. Of the 1,437 eligible men, 1,287 consented to participate in the study (89.6%). The study was approved by the research ethics board.

The primary end point was the histologic presence of adenocarcinoma of the prostate, based on one or more prostate biopsies. All biopsies were done using transrectal ultrasound guidance using an 18-gauge needle core. A minimum of six systematic needle cores was obtained at the time of biopsy (median number of needle cores = 8). The Gleason scoring system was used to grade all cancers.

KLK2 sequencing analysis for SNP discovery. To identify novel SNPs within the KLK2 gene, the entire KLK2 gene and promoter region was sequenced using the 3100 Genetic Analyzer (ABI Prism) using genomic DNA from 20 patients (all Caucasian) having prostate cancer at biopsy and with Gleason score 8 or more prostate cancer (randomly chosen). All DNA samples were extracted from peripheral blood leukocyte using standard protocols. Sequencing strategy and primer sequences are outlined in Table 1 .

We found a total of 12 SNPs among the 20 patients with high-grade prostate cancer. Of the 12 SNPs, 11 were listed in the National Center for Biotechnology Information Genbank and Celera databases (Table 2 ). The original C for T 73C SNP (rs198977) that we previously examined (3) was also present among these patients.

Genotyping analysis. The 12 SNPs were genotyped among all 1,287 patients who underwent a prostate biopsy. Genotyping was conducted using mass spectrometry–based genotyping analysis using matrix-assisted laser desorption ionization-time of flight (MassArray System, Sequenom, Inc., San Diego, CA) following the manufacturer's instructions, except for the original SNP (rs198977), which was analyzed using restriction fragment digest described previously (3). Details of the SNPs regarding location, allele type, location, and primers are outlined in Table 2.

A standard protocol for multiplex homogeneous mass extend assay developed by Sequenom was used and modified according to designed primers. In brief, multiplex PCR was done in 5 μL total volumes that contain 0.15 unit of HotStar Taq polymerase, 500 μmol/L of each of deoxynucleotide triphosphates, and 4 ng of genomic DNA. Thermocycling was at 95°C for 15 minutes followed by 45 cycles of 95°C for 20 seconds, 56°C for 1 minute, and 72°C for 1 minute using a thermal cycler PTC-225 (MJ Research, Inc., Watertown, MA). Unincorporated deoxynucleotide triphosphates were deactivated using 0.3 unit of shrimp alkaline phosphatase (Roche, Indianapolis, IN) followed by primer extension by use of 5.4 pmol of each primer extension probe, 50 μmol of the appropriate deoxynucleotide triphosphate/dideoxynucleotide triphosphate combination, and 0.5 unit of Thermosequenase (Amersham Pharmacia, Piscataway, NJ). Reactions were cycled at 94°C for 2 minutes followed by 100 cycles of 94°C for 5 seconds, 52°C for 5 seconds, and 72°C for 5 seconds. After addition of a cation-exchange resin to remove residual salt from the reactions, 0.7 nL of the purified primer-extension reaction was loaded onto a matrix pad of a spectroCHIP (Sequenom). SpectroCHIPs were analyzed using a Bruker Biflex III matrix-assisted laser desorption/ionization mass spectrometry-time-of-flight mass spectrometer (SpectroREADER, Sequenom) and were spectra processed using SpectroTYPER Analyzer. All equipment, including the robot system and mass spectrometer, were maintained by contracting with Sequenom.

For quality control, we assigned negative controls for each test plates (Microseal TM 384 V2.0). Successful genotyping assays were defined as those in which 90% of all possible genotyping calls were obtained. Although we used 90% as a minimum threshold, we obtained an average 95 % for all genotypes attempted for each successful SNP. Among 1,267 patients, 1,251 were successfully genotyped.

Measurement of serum hK2. Serum hK2 levels were measured using a new, time-resolved immunofluorometric assay described in detail elsewhere (3). In short, ∼90% of hK2 is in the free form, but the assay measures both free and total forms of hK2 (5). The hK2 assay has a detection limit of 0.006 ng/mL for both plasma and serum samples and has <0.2% cross-reactivity to PSA (5).

Data analysis. For each SNP, a χ² goodness of fit test was used to assess deviation from Hardy-Weinberg equilibrium for genotype frequencies. Those found not to be in Hardy-Weinberg equilibrium were excluded from the analysis. Linkage disequilibrium patterns were characterized by calculating pairwise D′ and R values.

We compared the allele frequencies for each SNP between cases and controls. Cases were defined as patients having prostate cancer at biopsy, and controls were those with no evidence of cancer at biopsy. We calculated the OR for the presence of prostate cancer at biopsy for each variant SNP allele using multivariate unconditional logistic regression analysis. Covariates in the model included age at biopsy, ethnic group, family history of prostate cancer, PSA level, and DRE status. Age and PSA were treated as continuous variables. Ethnicity was categorical and divided into Caucasian, Black, Asian, and Other. A positive family history of prostate cancer was defined as having one or more relative with a history of prostate cancer. DRE status was dichotomized into normal or abnormal.

To determine whether there were any differences in the expression of the KLK2 gene for the various SNP alleles, we compared serum levels of hK2 for each SNP by their genotypes. We also calculated the adjusted OR for prostate cancer by each variant allele stratified by serum hK2 levels.

Finally, we conducted haplotype analysis among the SNPs to examine their relationship between prostate cancer and serum hK2 levels. Haplotype frequencies were determined using an expectation-maximization algorithm (8). We used the haplotype trend regression model with binomial response to estimate haplotype assignment to individuals using the HelixTree Genetic Analysis software program V4.3 (http://www.goldenhelix.com, Bozeman, MT). The estimate of the haplotypic OR was defined as the odds of being a case if the patient has the given haplotype in both chromosomes compared with the odds of a being a control having the haplotype in both chromosomes.

To conduct logistic regression analyses using haplotypes and to determine correlations with hK2 levels, individual subjects required a haplotype assignment. Because haplotype assignments cannot be determined unambiguously, and rather probabilistically, we used a cutoff of minimum diplotype probability of ≥80% of having an assigned haplotype pair for each patient, based on the results of the expectation-maximization algorithm method.

Results

Clinical features of study subjects. The mean age at biopsy of the 1,251 men was 65.4 years (range = 41.4-93.8 years). The mean PSA level was 12.9 ng/mL (range = 0.4-732 ng/mL), and 57.9% had a normal DRE. The majority of the patients were White (n = 1,046, 83.6%); however, 107 (8.6%) were Black, and 68 (5.4%) were Asian. Eleven percent of patients had at least one relative with prostate cancer. More than half of the patients (59.9%) reported no obstructive voiding symptoms (i.e., were asymptomatic).

Of the 1,251 men, 645 (51.6%) men were found to have adenocarcinoma of the prostate at biopsy (cases), and 606 (48.4%) men had no evidence of invasive cancer (controls). The mean age of the cases (66.6 years) was higher than controls (64.2 years; P < 0.0001). The majority of both cases (86.4%) and controls (80.7%) were Caucasian Whites, but 9.5% cases and 7.6% controls were Black, and 2.8% cases and 8.3% controls were of Asian origin. Cases were more likely to have an abnormal DRE (P = 0.0001) and higher PSA (P = 0.0001) than controls, and the prevalence of prostate cancer was lower in Asians (Table 3 ).

Selection of KLK2 SNPs. Direct sequencing across the KLK2 gene in 20 individuals with high-grade prostate cancer identified 12 SNPs, the properties of which are shown in Table 2. Genotyping of these SNPs among the 1,251 subjects revealed negligible allele heterozygosity for five SNPs (rs10422897, rs1064676, rs1064703, rs1059712, and rs1802720), which were thus excluded from further analysis. Two other SNPs (rs3760731 and rs11670728) showed a significant deviation from Hardy-Weinberg equilibrium (P = 0.03) in the controls and were also excluded from further analysis.

Individual SNPs and prostate cancer association. When the five remaining SNPs were assessed individually (Fig. 1 ), two were found to significantly predict the presence of prostate cancer (Table 4 ). The crude OR for having prostate cancer for patients with the variant A allele of KLK2-SNP1 (rs2664155; AA or AG) was 1.4 (95% CI, 1.2-1.8; P = 0.0008) compared with patients with the GG genotype. The crude OR for having prostate cancer for patients with the variant T allele of KLK2-SNP5 (rs198977; TT or TC) was 1.3 (95% CI, 1.1-1.6; P = 0.01) compared with patients with the CC genotype.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

KLK2 gene structure located on chromosome 19q13.41. It consists of five exons (black boxes) and is 5,217 bp in length (226 amino acids). SNP1 (rs2664155) is located on intron 1, and SNP5 (rs198977) is located on exon 5.

These associations were still present after multivariate regression analysis incorporating age, PSA, hK2, DRE, family history of prostate cancer, and ethnicity as covariates. The adjusted ORs did not change significantly (OR, 1.4; P = 0.002 for SNP1 and OR, 1.3; P = 0.05 for SNP5). Similar ORs were also observed in analyses restricted to Caucasian White subjects (data not shown). No associations were detected between prostate cancer and the other three KLK2 SNPs included in the analysis (Table 4).

Individual SNPs and correlation with serum hK2 levels. To determine whether the two risk-associated variants showed any association with the expression of hK2, levels of serum hK2 were compared by genotypes among all subjects and between cases and controls. Mean serum levels of hK2 for all patients with the AA and AG genotype of KLK2-SNP1 were 0.42 and 0.48 ng/mL, respectively, and were significantly higher than patients with the GG genotype (0.27 ng/mL; P = 0.001). Patients with the variant A allele of KLK2-SNP1 had significantly higher serum hK2 levels (mean hK2 = 0.47 ng/mL) compared with patients with the GG genotype (mean hK2 = 0.27 ng/mL; P = 0.006). When grouped by cases and controls, patients with the variant A allele of KLK2-SNP1 had a significantly higher hK2 level (mean hK2 = 0.66 ng/mL) than patients with the GG genotype (mean hK2 = 0.35; P = 0.03) among the 645 cases. No differences in hK2 levels were found between the alleles among the controls (mean hK2 for variant A allele = 0.23 ng/mL versus mean hK2 for GG genotype = 0.20 ng/mL; P = 0.53).

The mean serum levels of hK2 for patients with the TT genotype of KLK2-SNP5 was 0.095 ng/mL, which was significantly lower than patients with the CC (mean hK2 level = 0.41 ng/mL) and CT genotype (mean hK2 level = 0.38 ng/mL; P = 0.0001). However, there were no significant differences in hK2 levels associated with the combined (CT and TT) risk (mean hK2 = 0.34 ng/mL) versus non-risk genotype (P = 0.38). Similarly, no differences were found in hK2 levels between the risk and non-risk genotypes when stratified by cases and controls.

Associations between KLK2 SNP haplotypes, prostate cancer and serum hK2 levels. Evaluation of linkage disequilibrium between the 5 KLK2 SNPs revealed a level of association between the SNPs (average D′ = 0.45) consistent with their comprising a single haplotype block across the KLK2 gene. Further evaluation of allele combinations across the study cohort revealed 15 common and 7 rare haplotypes across the region (Table 5 ). Comparison of haplotype frequencies between the cases and controls using a haplotype trend regression model indicated the haplotype carrying both KLK2-SNP1 and SNP-5 risk alleles (haplotype 13) to be positively associated with risk for prostate cancer (OR, 4.2 for prostate cancer compared with patients without the haplotype; P = 0.01), whereas the most common haplotype (haplotype 1) containing neither risk allele to be significantly associated with a protective effect (OR, 0.59 for prostate cancer compared with patients without the haplotype; P = 0.02). Although haplotype 12 (ATCTT), which contained the variant alleles, did not show an increased risk for prostate cancer (OR, 0.96; P = 0.42), when combined with haplotype 13, the OR for prostate cancer remained significantly high (OR, 3.2; P = 0.003).

To examine the effects of these haplotypes within a multivariate model and whether they correlated with a functional change in gene expression with respect to serum hK2 levels, we assigned haplotypes to individuals based on a minimum diplotype probability threshold of 80% for a patient to have the haplotype pair. From this assignment, 20 patients were assigned the ACCTT haplotype, and 496 patients had the GCCCC haplotype. For the 20 patients with the ACCTT haplotype, 16 (80%) had cancer, and for the 496 patients with the GCCCG haplotype, 239 (48.2%) had cancer. After adjusting for age, ethnicity, family history of prostate cancer, PSA level, hK2 level, and DRE status in a multivariate logistic regression model, the OR for having prostate cancer for patients with the ACCTT haplotype compared with patients with the GCCCC haplotype (most common) was 5.1 (95% CI, 1.6-16.5; P = 0.005). The adjusted OR for prostate cancer for patients with the ACCTT haplotype compared with patients without the ACCTT haplotype was 4.2 (95% CI, 1.3-13.2; P = 0.01). The adjusted OR for prostate cancer for patients with the GCCCC haplotype compared with patients without the GCCCC haplotype was 0.77 (95% CI, 0.6-0.9; P = 0.03). In addition, when removing hK2 level from the multivariate model, the ORs did not significantly change.

The two haplotypes also correlated with serum hK2 levels. Patients assigned the ACCTT haplotype had mean hK2 levels of 0.39 ng/mL, which was significantly higher than patients with GCCCC haplotype (mean hK2 level = 0.33 ng/mL; P = 0.05). There were no significant differences in the distribution of hK2 levels by haplotype status when stratified by cases and controls.

We then compared the proportions of prostate cancer between patients with the ACCTT haplotype with high hK2 levels and those with the GCCCC genotype and low hK2 levels. Using a cutoff of 0.205 ng/mL for serum hK2 levels (the median value), all nine patients with the ACCTT haplotype and with high hK2 levels (≥0.205 ng/mL) had prostate cancer (Table 6 ), compared with only 36.4% (n = 74) of patients with the GCCCC haplotype and with low hK2 levels (<0.205 ng/mL) who had prostate cancer (P = 0.0001; Table 6).

Discussion

Among prescreened men who underwent a prostate biopsy, we found strong associations between two SNPs of the KLK2 gene, hK2 serum levels, and prostate cancer risk. Two individual SNPs of the KLK2 gene (Fig. 1) and the haplotype combinations positively correlated with serum levels of its protein product (hK2) and were positively associated with an increased probability of having prostate cancer at the time of biopsy independently of other risk factors for prostate cancer. In addition, in a small subset of men with high hK2 levels and who had the variant haplotype of the KLK2 gene, all patients had prostate cancer. This is the first report that describes this association with the KLK2 gene and prostate cancer risk.

We previously reported a positive association between the C for T polymorphism (SNP5) and prostate cancer risk (3). We also showed that this SNP correlated with serum hK2 levels. However, the variant genotype (TT) correlated with lower levels of serum hK2, whereas both the TT genotype and high serum hK2 levels are associated with increased prostate cancer risk. To examine other SNPs of the KLK2 gene, we chose to sequence patients diagnosed with high-grade prostate cancer rather than examining tagSNPs or other approaches because we wanted to incorporate our previous SNP findings within the haplotype analysis.

In the current study, we identified a new SNP (KLK2-SNP1), which was positively associated with both prostate cancer risk and serum hK2 levels. Furthermore, when the SNPs were combined in haplotype analysis, the correlation between the variant alleles and serum hK2 levels were both positively associated with serum hk2 levels and prostate cancer risk in all subjects. The reason we did not find correlations after stratified by cases and controls among the haplotypes is likely due to samples size because significant differences were found between cases and controls in hK2 levels by specific SNPs. SNP1 (rs2664155) is located on intron 1 of the KLK2 gene. Subsequent investigation into intron 1 of the KLK2 gene did not identify any regulatory transcription factor binding regions containing this SNP. Given its location and strong correlation to serum hK2 levels, it is likely that its effects may occur during splicing events of its mRNA, and it is possible that this SNP may be a new regulatory region for the KLK2 gene. However, this finding could be a result of simple linkage disequilibrium with another unknown SNP involved in regulatory function. In vitro studies will be required for further elucidation.

For reasons that are currently unclear, two of the promoter region SNPs did not show Hardy-Weinberg equilibrium in this study. This result might reflect chance alone or possibly genotyping errors. We do not believe that genotyping errors account for the result (based on our validation of genotypes on selected samples by sequencing) and if such errors did occur randomly, association between the SNPs and prostate cancer would be underestimated.

Among the seven exonic SNPs, four were nonsynonymous (rs6072, rs10422897, rs198977, and rs1059712). SNP rs198977 (SNP5), which was positively correlated to prostate cancer risk in our study, was shown by Herrala et al. to be associated with differential expression of serum hK2 protein in vitro (4). However, no other studies have shown effect on the three other nonsynonymous SNPs on expression or function of the KLK2 gene.

These findings are consistent with the hypothesis that the KLK2 gene could be a prostate cancer susceptibility gene, and that elevated cancer risk is associated with increased hK2 levels. HK2 is a serine protease produced by the secretory cells of the prostate. Frenette et al. showed that hK2 activates the proforms of PSA and urokinase type plasminogen activator (9), and Mikolajczyk et al. showed that hK2 inactivates plasminogen activator inhibitor-1 (10). These observations support the role of a proteolytic cascade associated with prostate cancer growth and metastasis for which the current novel therapy has been targeted to peptide inhibitors of hK2 (11). Indeed, other kallikrein proteins, including PSA, hK4 and hK5 have been shown to up-regulate production of tumor promoters, vascular endothelial growth factor, and Pim-1 oncogene (12) and also affect expression of insulin-like growth factor proteins and E-cadherin, which also have been involved in prostate cancer development and metastasis (13, 14). In addition, Stenman et al. showed that hK2 mRNA expression was significantly higher in prostate cancer tissue compared with benign tissue (15). Further study of mRNA expression of hK2 from prostate cancer cells based on haplotype status of the variant alleles of the KLK2 gene will be required.

Although significant ORs for prostate cancer were found with two variant haplotypes of the KLK2 gene, our study was limited by sample size. Only 20 patients were predicted to have the ACCTT haplotype (probability for haplotype threshold of ≥80%). Furthermore, after stratifying by hK2 levels (according to median values), our sample size was further reduced. Nevertheless, the associations yielded strong associations (P = 0.0001; Table 6). Further study among a larger group of patients will be required to confirm our findings.

Another interesting finding was the possibility of combining serum hK2 levels and KLK2 genotype status to predict prostate cancer. When we used an arbitrary cutoff of the median value of hK2 based on its overall distribution, we were able to identify all patients with prostate cancer with high hK2 levels and the variant haplotype (Table 6). The number of patients in this subgroup analysis was small, and no firm conclusions can be drawn. However, this finding introduces the possibility of potentially using this approach in future clinical applications if further study confirms these results.

In summary, we identified a haplotype block of the KLK2 gene that is strongly associated with the presence of prostate cancer (OR, 5.1; P = 0.005) and is associated with increased production of its protein product hK2. The KLK2 gene may be an important candidate gene for prostate cancer and could be eventually used as marker for prostate cancer if confirmed from larger confirmatory studies.