Genome-wide association studies have identified SNPs associated with prostate cancer. Recently, SNPs in KLK3 were related to prostate cancer risk in a genome-wide association study (GWAS; 1,854 cases)1 and two candidate gene investigations (596 and 209 cases, respectively)2,3, raising the possibility that this gene and its encoded protein PSA (prostate specific antigen; kallikrein-related peptidase), which is widely-used as a biomarker for prostate cancer detection, are etiologically related to this disease.
Understanding the contribution of common genetic variation in KLK3 to prostate cancer risk is daunting because SNPs in KLK3 are determinants of serum PSA concentrations1,2,4, and thus, observed risk relationships could be due to differential identification of cases, in settings where PSA is used in screening or clinical diagnosis of this disease. Therefore, we evaluated the association of 24 tagSNPs in the KLK region (including KLK3 and nearby genes KLK1, KLK2, and KLK15, chromosome 19: 56022744–56073589) and prostate cancer risk in 1,172 prostate cancer cases and 1,157 PSA screened controls in men of European ancestry from the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial5,6, in which serum PSA was used in screening for this disease. We also examined the prostate cancer risk relationship for 12 of these tagSNPs in the KLK3 region (chromosome 19: 56031744–56063231) in four independent studies including 4,020 prostate cancer cases and 4,028 controls (American Cancer Society Cancer Prevention Study II, the Health Professionals Follow-up Study, the CeRePP French Prostate Case-Control Study, and the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study)6, as a component of the Cancer Genetic Markers of Susceptibility (CGEMS)-GWAS6 (http://cgems.cancer.gov; Supplementary Methods).
None of the 24 KLK-region tagSNPs showed strong evidence for association with prostate cancer in PLCO (Table 1: Cases vs Controls in PLCO) and none of the 12 KLK3-region SNPs was significant in the CGEMS combined analyses (Table 1: Cases vs Controls in CGEMS combined). No substantial differences in risk were noted by disease aggressiveness (Supplement Table 1). Two tagSNPs (rs1058205 and rs2735839), located in the 3′ untranslated region of KLK3, in high linkage disequilibrium (LD; r2=0.8; Figure 1), along with four variants 8–23 kb upstream of the gene (rs2659056, rs2569729, rs266849, and rs266870) and several variants about 7 kb downstream (rs1506684 and rs2569739, in high LD, r2=1.0), were associated with strong differentials in serum PSA concentrations (e.g., rs2735839, p trend=4.1×10−9; Table 1, Adjusted Mean PSA Level, and Figure 1-Panel A). However, when these SNPs were simultaneously included in the multivariate model, associations remained for rs2735839, but others were no longer significantly associated with PSA concentrations.
Eeles et al 1 reported strong associations between three KLK3 SNPs (p-values for stage 1 were 1.2×10−7, 1.0×10−16 and 2.4×10−20 for rs2659056, rs266849, and rs2735839) and prostate cancer, in a GWAS, where controls were selected by design for low PSA (<0.5 ng/ml) and no limitations were placed on case-group PSA values. The replications in non-PSA selected groups showed attenuated associations (p-values for stage 2 were 0.424, 0.228 and 0.0002, respectively). Pal et al 2 also reported a strong association of prostate cancer with KLK3 variants, where control subjects were also preferentially selected by PSA levels (< 2.5 ng/ml). When we preferentially limited the men in the control group in the PLCO Trial GWAS to those with PSA <0.5 ng/ml, as an exercise, we show prostate cancer risk associations (Figure 1-Panel B  and Supplement Table 2) greater than 5-fold for some genotypes (rs1058205, OR=0.19, 95%CI=0.10–0.36, p=0.000015), whereas the same SNPs showed no clear association when the full control group was employed (Figure 1-Panel B [●] and Table 1). Furthermore, modest KLK3 associations with prostate cancer risk observed in our full case-control series and in other studies where controls are not selected by design for PSA level3 may still be related to PSA-based case ascertainment, because PSA is widely used as a clinical aid in selecting men for biopsy for suspect prostate cancer. In PLCO, KLK SNPs were also not associated with prostate cancer when only cases and controls with high PSA levels or only cases and controls with low PSA levels were compared (Supplement Table 3). Thus, the observed associations between the KLK SNPs and prostate cancer risk in earlier studies4,5,3 and in our exercise in PLCO may be due to selection for PSA differentials and not causally related to underlying disease risk.
PSA is a serine protease, functioning in the liquefaction of seminal coagulum7; consistent with our data, Pal et al2 showed serum PSA associations with rs266849 and rs1058205, and Cramer et al4 showed that two SNPs in the 5′ region of KLK3 are related to increased KLK3 promoter activity and serum PSA level. In our study, the association with PSA was strongest with rs2735839, which is highly correlated with rs1058205 (r2=0.8); associations observed with the other SNPs in our study appeared to be largely driven by their linkage disequilibrium with rs2735839. Sliding window haplotype analysis (Supplement Figure 1) and close inspection of the most strongly associated haplotype window (Supplementary Table 4) suggested that additional unmeasured risk variant(s) may influence PSA concentrations.
PSA is an early marker predictive for the presence of prostate cancer and its use has contributed to the near doubling of diagnosed cases over the past two decades in countries where use is common8. While trials are underway to determine whether screening with PSA reduces prostate cancer mortality5,9, it is apparent that many men are diagnosed with clinically indolent cancer on the basis of PSA screening; the over-detection rate is estimated to range from 18 to 56%9,10. In addition, the majority of men with an elevated PSA have a negative biopsy11. In our study, KLK3 variant status was related to PSA test-positivity in control men (Supplementary Table 5). While we show that KLK3 SNPs are not risk factors highly correlated with prostate cancer per se, a combination of KLK3 SNP and serum PSA monitoring, possibly including a broader spectrum of protein and DNA tests12, is a potential tool for advancement in prostate cancer early detection.
In conclusion, SNPs in the KLK3 region were not associated strongly with prostate cancer in large series of cases and controls, where controls are not selected by design for low PSA and even the modest KLK3-prostate cancer risk associations observed in non-selected populations may be due to PSA-directed differential identification of prostate cancer cases with particular KLK3-PSA profiles.