Prostate cancer risk associated Single Nucleotide Polymorphism affects PSA glycosylation and its function

Abstract

BACKGROUND:

Genetic association studies have reported single nucleotide polymorphisms (SNPs) at Chromosome 19q13.3 to be associated with prostate cancer (PCa) risk. Recently, the rs61752561 SNP (Asp84Asn substitution) in exon-3 of the kallikrein-related peptidase 3 (KLK3) gene encoding Prostate-Specific Antigen (PSA), was reported to be strongly associated with PCa risk (P=2.3×10⁻⁸). However, biological contribution of the rs61752561 SNP to PCa risk, has not been elucidated.

METHODS:

Recombinant PSA proteins were generated to assess the SNP-mediated biochemical changes by stability and substrate activity assays. PC3 cell-PSA overexpression models were established to evaluate the effect of the SNP on PCa pathogenesis. Genotype-specific correlation of the SNP with total PSA (tPSA) levels and free/total (F/T) PSA ratio were determined from serum samples.

RESULTS:

Functional analysis showed that the rs61752561 SNP impacts on PSA stability, structural conformation and creates an extra-glycosylation site. This PSA variant had reduced enzymatic activity and ability to stimulate proliferation and migration of PCa cells. Interestingly, the minor allele is associated with lower tPSA levels and high F/T PSA ratio in serum samples, indicating that the aminoacid substitution may impact PSA immunoreactivity to antibodies used in the clinical immunoassays.

CONCLUSIONS:

Our assessment on the biological effects of the rs61752561 showed this SNP to have a potential role in PCa pathogenesis by changing the glycosylation, protein stability, PSA activity and may also impact on the clinically measured F/T PSA ratio. Accounting for these effects on the tPSA and F/T PSA ratio may help to improve the accuracy of the current PSA test.

Introduction

Twin studies have suggested that prostate cancer (PCa) has a significant heritable component and a family history of PCa among first degree relatives remains a large risk factor for the disease (1). Indeed, many genome-wide association studies (GWAS) have successfully identified more than 150 risk loci contributing to PCa susceptibility (2). Recently, a GWAS follow-up fine-mapping analysis was performed using summary results from the largest European study to date comprising 82,591 PCa cases and 61,213 controls from 8 GWAS cohorts imputed to 1000 Genomes (3). This study confirmed the existence of multiple independent risk signals at 12 loci previously reported and identified three independent association signals for the first time at the kallikrein-related peptidase 3 (KLK3) gene (Chr19:50840794–51864623) locus. The region was ±500 kb from the previously reported index rs2735839 SNP. Three independent PCa risk associated signals were identified in this region: the most strongly associated signal was the previously reported intronic rs62113212 single nucleotide polymorphism (SNP) (4, 5) odds ratio (OR)=0.74, P=4.26×10⁻⁸¹, followed by a non-synonymous rs61752561 OR=0.89, 95% CI 0.86–0.93, P=2.33×10⁻⁸ and an intergenic rs266863 OR=0.92, P=6.88×10⁻²⁵ (Supplemental Table 1). The rs62113212 SNP was in linkage-disequilibrium (LD) (r²≥0.97) with four other KLK3 SNPs, while the rs61752561 and rs266863 are single independent SNPs associated with PCa risk (Supplemental Figure 1 derived from Locus Explorer). The overall PCa familial relative risk contributed by this region was 0.86 (95% CI 0.76, 0.98), an increase by 1.9 fold partly due to the identification of the two novel signals (3).

KLK3 codes for prostate specific antigen (PSA), a prostatic secretory protein and a serine protease reported to have a bi-directional role involved in both tumor suppression and progression (6, 7). PSA is the most widely used non-invasive biomarker for PCa detection and surveillance (8, 9). However, the sensitivity and specificity of the test for clinically significant cancer is poor and thus the use of PSA as a biomarker is still controversial (10, 11). Accordingly, recent research has focussed on novel interventions to improve the diagnostic specificity such as proPSA, the Prostate Health Index, free/total (F/T) PSA and PSA velocity. Total PSA (tPSA) comprises both the complexed (with inhibitors α−1-antichymotrypsin/ACT/SERPINA3, alpha-2-macroglobulin/A2M, protein C inhibitor/PCI, alpha 1-protease inhibitor (API)) and free PSA (fPSA) forms. A lower %fPSA has been associated with higher likelihood of a diagnosis of PCa, while measuring F/T PSA is considered more sensitive in discriminating PCa from benign disease (12). Association between SNPs in the KLK3 region and PCa risk has been reported previously (4, 5, 13–15); however, no consistent correlation has been observed thus far between the cancer-associated SNPs in the KLK3 gene and protein levels of either PSA or F/T PSA ratio. The location of the three independent signals and their association with PCa risk, raise the possibility that these SNPs may be causally related to disease risk or affect the clinical detection of PSA, although this remains to be proven. In an independent study, we showed rs17632542, which is in LD with the rs62113212 SNP, to have suggestive functional role in PCa aetiology (4, 16, 17). Here, we hypothesised that the second recently GWAS identified non-synonymous SNP may influence PSA function and thereby its role in PCa. Consequently, our current analysis focussed on characterising the functional consequences of the rs61752561 SNP, with a minor-allele frequency (MAF)=0.04, that leads to an amino acid substitution, Asp84Asn (chymotrypsin numbering) of PSA. Further we determined whether the rs61752561 SNP correlated with circulating PSA levels and affected the F/T PSA ratio.

Materials and Methods

Detailed Materials and Methods are described in the Supplemental file and Supplemental Table 2.

Results

In silico analysis suggested differences in stability and function for the rs61752561 SNP

To analyse the possible impact of the rs61752561 non-synonymous SNP on PSA protein stability, protein function and post-translational processing, publicly available in silico tools were used. I-Mutant3.0 and FoldX in silico tools predicted that the rs61752561 SNP had the potential to reduce protein stability (Supplemental Table 3). SIFT, PROVEAN and PANTHER, suggested a neutral or a marginal effect on protein function (Supplemental Table 3). NetNGlyc1.0 tool predicted that rs61752561 may create an additional glycosylation site due to generation of the consensus sequence Asn-X-Ser/Thr.

rs61752561 introduces an additional glycosylation site

To examine the biochemical effects on PSA of Asp84Asn substitution predicted by our in silico analysis, we expressed and purified recombinant PSA from yeast cells. Recombinant Asn⁸⁴ PSA when resolved on SDS PAGE demonstrated a mobility shift (Supplemental Figure 2A) consistent with the gain of a glycosylation site also suggested by the NetNGlyC prediction tool. A catalytically inactive mutant Ala¹⁹⁵ PSA (Ser195Ala substitution) was also generated. To verify that the mobility shift was due to glycosylation, a deglycosylation assay for the recombinant mature PSA variants wild type (WT) and Asn⁸⁴ PSA was performed (Figure 1A). For Asn⁸⁴ PSA, the extra-glycosylated band was also removed on PNGase treatment and aligned around 25–26 kDa (similar to the other PSA proteins), verifying that the glycosylation site created for the rs61752561-derived protein variant is an N-glycosylation (Figure 1A).

Figure 1. Biochemical analysis of the rs62752561 SNP effect on PSA.

A) A representative Western blot analysis using anti-PSA shows deglycosylation of mature recombinant PSA protein variants (WT and Asn⁸⁴) on PNGase treatment, aligning around 25–26 kDa. An extra glycosylated band was observed for the Asn⁸⁴ PSA alone which was removed on PNGase treatment (pink arrows) aligning to the PNGase treated WT PSA indicating an N-glycosylation. B) The Tm values were calculated as the maximum of the first derivative of the melting curves. The fluorescence values are on an arbitrary scale (AU). C) The first derivative values (d(fluorescence)/dT) are normalised and given in %, where 100% corresponds to the highest peak of the entire experiment set. Tm values for WT PSA and Asn⁸⁴ PSA are 68.4 °C and 69.6 °C, respectively. Representative curves of three experiments are shown. D) Rate of hydrolysis by mature PSA proteins (wild type PSA/WT PSA, Asn⁸⁴ PSA, and catalytically inactive mutant control Ala¹⁹⁵ PSA at 0.1 μM) were compared using the substrates MeO-Suc-RPY-MCA (10 μM) and Mu-HSSKLQ-AMC (1 μM) over 4 h at 37 °C. Proteolytic activity derived from assaying a constant amount of PSA with increasing concentration (0–250 mM) for these two substrates were used to estimate K_cat values using the nonlinear regression analysis in Graphpad Prism. Results are shown as the mean ± SEM from two experiments, each with three replicates (Kruskal-Wallis test ****P<0.0001). E) Silver stained gel image of Galectin-3 (blue arrow) proteolysis by WT and Asn⁸⁴ PSA and inactive Ala¹⁹⁵ mutant (0.2 μM of PSA protein variants (orange arrows) and 0.5 μM Galectin-3), after 18 h of incubation at 37°C. A 16 kDa fragment (green arrow) due to the cleavage event is indicated. F) Time (mins) versus relative absorbance (OD) corrected to the substrate alone controls was plotted indicating the activity of pro-MMP2 (0.14 μM) when pre-incubated with PSA protein variants (WT, Asn⁸⁴ and Ala¹⁹⁵ at 0.07 μM) at 37°C and then the activity analysed with the chromogenic substrate (Ac-PLG-[2-mercapto-4-methyl-pentanoyl]-LG-OC2H5, 40 μM) for active MMP2 over 2 h. Results are shown as the mean ± SEM of three experiments analysed using the Kruskal-Wallis test. ***P<0.001, ****P<0.0001.

rs61752561 SNP affects thermostability of Asn⁸⁴ PSA

To determine the effects of the rs61752561 SNP on protein stability (Supplemental Table 3), glycosylated WT and mutant proteins were measured for their thermal stability using differential scanning fluorimetry (DSF). The melting curves of the glycosylated proteins revealed an increase in the T_m value (1.2 °C) of the Asn⁸⁴ PSA variant compared with WT PSA (ΔT_m = 68.5±0.03 °C for WT and 69.8±0.13 °C for Asn⁸⁴) (Figure 1B, 1C).

Asn⁸⁴ PSA had an altered proteolytic activity

The activity of WT and Asn⁸⁴ PSA were first evaluated using two fluorigenic peptide substrates, MeO-Suc-RPY-MCA and Mu-HSSKLQ-AMC (Figure 1D). The Asn⁸⁴ PSA had a significantly reduced K_cat value compared to WT PSA for both peptides (Figure 1D and Supplemental Figures 2B and 2C). Similarly, a reduced enzymatic activity was observed for the Asn⁸⁴ PSA by casein zymography (Supplemental Figure 2D). We analysed if there is any difference in inhibition kinetics for the two PSA isoforms by a commercial monoclonal antibody (mAb) that is suggested to inhibit PSA activity by 98% at 1:2 molar ratio of PSA. For WT PSA, the activity was reduced with increasing mAb concentration suggesting inhibition by the antibody whereas the rate of inhibition of the activity of the Asn⁸⁴ PSA was not dose dependent for the Asn⁸⁴ PSA (Supplemental Figure 2E) as was also observed with two other mAbs 5A10 and 5C7 (Supplemental Figures 2F and 2G) suggesting differences in proteolytic activity. To examine further the effect of amino acid substitution of Asn⁸⁴ on PSA function, we utilized several previously identified substrates of PSA (7, 18, 19). Silver stain analysis after 22 h incubation of recombinant PSA-protein variants with the full-length substrates, fibronectin, laminin α−4, nidogen-1, IGFBP-3 and galectin-3 demonstrated that Asn⁸⁴ PSA had an altered proteolytic activity compared to the WT PSA with all the substrates analysed (Figure 1E and Supplemental Figures 3A–3E). This was most clearly seen with the prominent cleaved 16 kDa band of galectin-3 (Figure 1E) and densitometry analysis (Supplemental Figure 3E). Furthermore, PSA was observed to cleave pro-MMP2 leading to activation of the zymogen to an active MMP2 protease (20, 21) (Figure 1F); Asn⁸⁴ PSA being less active than WT PSA in this respect. As expected, the mutant Ala¹⁹⁵ PSA did not show any activity in our activity assays.

Asn⁸⁴ PSA complexes with serum inhibitors

Next, we explored whether rs61752561 affects the ability of the recombinant PSA to form complexes with its physiological inhibitors by analysing the F/T PSA ratio. The recombinant PSA proteins at three dilutions (10, 25 and 50 μg/L) were incubated in female serum devoid of PSA (Table 1). The total and free PSA detected by the Siemens Immulite assay indicated a linear curve with decreasing recombinant PSA concentration. The F/T PSA ratio ranged from 62–72% for all three concentrations of WT PSA. Interestingly, this differed significantly for the Asn⁸⁴ PSA where the F/T PSA ratio was observed to range from 34–37% for the three concentrations assayed (Table 1). Total PSA detected by the Beckman assay was similar to Siemens tPSA for both WT and Asn⁸⁴ PSA, while the Beckman fPSA was slightly different and ranged from 69–71% for WT and 45–48% for Asn⁸⁴ PSA. Since these two tests primarily detect PSA-ACT complexes, we hypothesized that the difference we observed for the two PSA isoforms may be due to the differences in their binding with ACT and A2M.

Table 1.

Measured free and total PSA levels in the presence of female serum

Protein	Expected Concentration μg/L	Siemens tPSA μg/L (mean ± SD)*	Siemens Immulite fPSA μg/L (mean ± SD)	Siemens F/T PSA ratio (%)	Beckman tPSA μg/L	Beckman fPSA μg/L	Beckman F/T PSA ratio (%)
WT PSA	50	31.0 ± 0.8	19.5 ± 0.5	62.6	23.5	>15.9	NA
WT PSA	25	13.2 ± 0.1	9.6 ± 0.7	72.8	12.7	8.7	69.0
WT PSA	10	4.9 ± 0.1	3.5 ± 0.7	71.7	4.1	2.9	71.0
Asn84 PSA	50	49.5 ± 1.3	18.7 ± 0.6	37.8	47.6	ND	NA
Asn84 PSA	25	28.2 ± 0.7	10.4 ±0.4	36.7	25.0	12.1	48.5
Asn84 PSA	10	9.4 ± 0.08	3.2 ± 0.23	34.2	7.9	3.6	45.6

tPSA = total PSA; fPSA = free PSA; F/T PSA = free/total PSA; NA = Not applicable; ND = not determined. The threshold for detection of fPSA is 25 μg/L for both methods, hence fPSA levels higher than the threshold levels couldn’t be determined for few concentrations.

^*mean ± SD (standard deviation) calculated from two replicates.

However, silver stain analysis and peptide substrate analysis of recombinant PSA proteins with recombinant ACT and A2M indicated a similar complexing ability of recombinant WT PSA and Asn⁸⁴ PSA (Supplemental Figures 4A–C) to these inhibitors.

Prediction of structural and energetic consequences of the Asp84Asn substitution

To gain insight into the dynamics of PSA and the impact of the Asp84Asn substitution, the crystal structure of PSA was obtained from the Protein Data Bank (PDB ID: 2ZCH) and the structure was visualized using PyMol. The solvent-exposed location of Asp⁸⁴ in a flexible loop distant from the active site, and the relatively conservative nature suggests that the rs61752561 SNP could indirectly affect the active site structure and/or accessibility of substrates by altering protein dynamics.

To explore this hypothesis further, we conducted MD simulations in triplicate for 500ns of WT and Asn⁸⁴ PSA. During the WT PSA simulation, the 70/80 loop, the kallikrein loop and the 148 loop were the most mobile regions of the protein (Figure 2A). Simulations of Asn⁸⁴ PSA revealed similar dynamics of loop 70/80, but markedly greater dynamics in both the kallikrein and 148-loops (Figure 2B). The increased flexibility of the kallikrein loop in Asn⁸⁴ PSA may limit its ability to mediate substrate access to the binding pocket, providing an explanation for its known role in substrate binding kinetics (22), and consistent with our findings of decreased proteolytic activity. To assess whether the conformational changes due to the Asn84 substitution could affect the ability of individual antibodies to react with epitopes, we examined the crystal structures of complexes between PSA and three anti-PSA mAbs (8G8F5 (23), 5D5A5 (24) and 5D3D11 (24)). mAb 8G8F5 binds to PSA in close proximity to Asn84, suggesting that glycosylation at this site may impact antibody binding. In contrast, both 5D5A5 and 5D3D11 mAbs bind to PSA at sites remote to Asn84, suggesting that glycosylation at this site would not affect antibody binding (Figure 2C).

Figure 2. Asp84Asn substitution shows increased flexibility in the kallikrein loop, limiting its ability to mediate substrate access to the binding pocket.

(A) Wild-type PSA has minimal fluctuations of kallikrein (green), 148 (magenta) and 70/80 (brown) loops. (B) Increased dynamics of kallikrein- and 148-loop in Asn⁸⁴ PSA protein. The catalytic triad Asp, Ser and His in active site are depicted by orange spheres. Arrows illustrate loop motion that may impede substrate access to the active site, and thus catalytic activity. The protein is illustrated as cartoon snapshots during the simulation. C) Cartoon depicting PSA in complex with 3 Fab molecules derived from monoclonal antibodies (mAbs). PSA is shown as a white surface, with Asn⁸⁴ coloured yellow and labelled. Fab molecules are shown as cartoons, with heavy and light chains coloured differently. mAb 8G8F5 (PDB ID: 2ZCH (23) binds PSA in close proximity to Asn84, such that glycosylation of Asn⁸⁴ may impact antibody binding. Binding of mAbs 5D5A5 and 5D3D11 (PDB ID: 3QUM (24)) to PSA would most likely not be affected by Asn84 glycosylation.

Asn⁸⁴ PSA expression lowered PCa cell proliferation and migration

To evaluate whether the biochemical effects of the rs61752561 SNP dictates effects on PSA function and thus the proliferation of PCa cells, we utilised PC3 cells overexpressing PSA protein isoforms (Figure 3). The expression of the isoforms (Figure 3A) and the efficient activation of the engineered pro-domain with a furin activation site was evident by the activity of the PSA variants (Figure 3B). The peptide substrate hydrolysis was observed to be higher for the WT PSA compared to the Asn⁸⁴ and Ala¹⁹⁵ PSA reflecting their activity profile observed with recombinant protein activity analysis. Proliferation of the PC3 cell clones expressing WT PSA, Asn⁸⁴ PSA, and inactive mutant Ala¹⁹⁵ PSA were compared with the vector alone transfected PC3 cells. In serum starved conditions, a significant increase in the proliferation rate of WT PSA expressing cells was observed compared to Asn⁸⁴ PSA (P<0.001) (Figure 3C). However, no significant difference in proliferation rates for these four different cell types was observed in media containing 5% FBS (presumably due to the presence of protease inhibitors in serum) (Figure 3D). Similarly, the migration of the cells expressing WT PSA was higher than the cells expressing Asn⁸⁴ and Ala¹⁹⁵ PSA in the absence of serum (Figure 3E; Supplemental Figure 5A) and this effect was not observed in the presence of serum (Figure 3F). We did not see any obvious difference in invasion of the cells (Supplemental Figures 5B, 5C) nor for the anti-angiogenic effect of Asn⁸⁴ PSA and WT PSA on HUVEC endothelial cells (Supplemental Figures 5D and 5E).

Figure 3. Functional assessment of the PSA coding variant Asn⁸⁴ in prostate cancer cells.

A) Representative mRNA analysis demonstrating the expression of PSA in PSA transfected PC3 clones B) Fluorescent activity observed for the active-PSA secreted by the overexpressing PC3 cells (WT, Asn⁸⁴, inactive mutant Ala¹⁹⁵ and vector) using a PSA peptide substrate (n=2, mean ± s.e.m. ****P<0.0001). (C-F) Cell-based functional assays to evaluate the effect of the PSA genetic variant on proliferation and migration of PC3 cells. Proliferation (C) and migration (E) of Asn⁸⁴ PSA overexpressing cells was lower than WT PSA expressing cells in serum free conditions but not in the presence of serum inhibitors (D and F). Statistical significance for all these assays were analysed by the Kruskal-Wallis test. (n=3 experiments performed in triplicate, mean ± s.e.m. ** Indicates P<0.01; *** indicates P<0.001).

The rs61752561 SNP is associated with lower total and high F/T PSA ratio

To determine the association of the rs61752561 SNP with PSA levels, we assessed baseline PSA levels and their correlation with the genotype among 2,647 controls without any diagnosis of PCa in the Malmö Diet and Cancer (MDC)-cohort during long-term follow up (25). The carrier status for the SNP was GG=2,415; AG=227 and AA=5. Baseline levels of free PSA were not statistically significant for the three genotypes (P=0.27) (Figure 4A). By contrast however, the baseline levels of tPSA were significantly lower among the men with [AA] and [AG] genotypes (P=0.00042) (Figure 4B). In concordance with these findings there was a significant association between the rs61752561 genotype and the F/T PSA ratio with the F/T PSA ratios being significantly higher for the [AA] and [AG] genotypes (P=3.1×10⁻⁹) compared to the homozygous [GG] genotype (Figure 4C).

Figure 4. Allelic specific expression of total and free PSA.

A) Box-plot graph showing log (fPSA) levels at diagnosis for AA (n=5), AG (n=227) and GG (n=2,415) genotype. B) Box-plot graph showing lower log (tPSA) levels at diagnosis for AA and AG genotype compared to GG (P =0.00042) (C) Correlation of F/T PSA ratio (%) with rs61752561 genotype. Significantly high F/T PSA ratio was observed for AA and AG genotypes (P=3.1×10⁻⁹) compared to GG genotype.

Discussion

GWAS and fine-mapping studies have identified three independent signals that cluster within a narrow region at chr19q13:KLK3 region (3). In this study we conducted functional analysis to investigate the contribution of one of the independent signals, a non-synonymous rs61752561 SNP at this chromosomal region to PCa predisposition. Notably, the rs61752561 SNP appears to be a rare variant with a MAF of 4% with a modest effect size (OR=0.89), which is likely why it has not been reported as a PCa susceptibility variant at genome-wide significance level until recently (3). Interestingly, in a previous small study, the rs61752561 SNP was reported to be associated with PCa-specific mortality (HR = 4.82, 95%CI: 2.33, 9.98, P<0.0005) (n=95 cases), biochemical recurrence (n=351 cases, P<0.0005) and castration-resistant metastasis (n=146, P <0.0005) (26). The biological role of this rs61752561 SNP and its biological role in mediating PCa risk is unclear and is the main focus of our current study.

In silico analysis for the possible function of the rs61752561 SNP discussed here and in previous studies (4, 5) suggested that this SNP can affect PSA stability or may have deleterious or moderate effect on PSA function. We showed that the Asn⁸⁴ PSA isoform was slightly more stable compared to the WT PSA likely due to the additional glycosylation of recombinant PSA. Although the denaturing conditions employed by the DSF technique may not accurately reflect the physiological temperature of 37 °C, they still provide clues that differences in stability for these proteins exist.

In addition to stability differences, the recombinant Asn⁸⁴ PSA had less activity as tested with two fluorescent peptide substrates compared to the WT PSA. The catalytic turnover of the peptide substrate, Mu-HSSKLQ-AMC, was less for Asn⁸⁴ PSA compared to the MeO-Suc-RPY-AMC substrate suggesting they may have different substrate selectivity. Similarly, their activity on full-length protein substrates varied based on the protein substrate analysed. For example, while hydrolysis of laminin α −4 and galectin-3, and activation of pro-MMP2 by Asn⁸⁴ PSA was less efficient as compared to the WT PSA, the activity of Asn⁸⁴ PSA on fibronectin and IGFBP-3 was similar to WT PSA illustrating that their activity on full-length substrates may vary. Of note, the activation of pro-MMP2 is suggested to contribute to the increased migratory ability of PCa cells via an MMP2 activation pathway (27). Degradation of galectin-3 is suggested to have a role in cell cycle control, cell adhesion and PCa progression (19). In line with the effects on various protein substrates and their known function, we observed higher proliferation and migration of PCa cells on overexpression of WT PSA but these effects were not seen for the Asn⁸⁴ PSA expressing cells.

Recombinant proteins when incubated in serum in vitro showed a recovery rate of 50–62% of tPSA for WT compared to 95% recovery rate for Asn⁸⁴ PSA. This may likely, be due to a difference between WT and Asn⁸⁴ PSA in their ability to form complexes to A2M and ACT, which are the major complexing ligands to catalytic PSA added to serum in vitro. A2M-PSA complexes will not be detected by conventional clinical assays for total PSA (28). However, we didn’t observe any difference in the residual peptide substrate hydrolysis of Asn⁸⁴ and WT PSA in the presence of A2M or with ACT. Another explanation for the differential recovery of the WT and Asn⁸⁴ PSA may be high auto-catalytic activity and low stability of WT PSA compared to Asn⁸⁴ PSA. It is also possible that the Asn⁸⁴ PSA may have an increased immunoreactivity to the antibodies used in clinical immunoassays. This could lead to a bias in the assay, the calibrator being not fully adapted for this molecular form.

The substitution of an acidic amino acid (Asp) with a polar amino acid (Asn) at position 84, led to glycosylation at this site, and thus may affect the local structure and dynamics. In addition, the non-synonymous rs61752561 SNP is located near the catalytically important kallikrein loop which likely mediates substrate access into the active site. Our MD simulations indicated that substitution alone increases the dynamics of both the kallikrein- and the 148-loops which are known to be important for enzyme activity. Therefore, it is probable that this substitution and also glycosylation decrease the catalytic activity of PSA and affect the ability of some antibodies to react with their epitope as observed in our assays involving inhibitory antibodies as well as free and total PSA assays. We observed a difference in binding kinetics of the Asn⁸⁴ PSA variant with three mAbs: a commercial Scrippslab antibody, 5A10 and 5C7 in comparison with WT PSA (Supplemental Figure 2E). Inspection of available complexes between PSA and antibodies (Figure 2C) showed that glycosylation may impact binding with antibodies depending on the proximity of epitopes to the Asn84 site. Overall, this data further emphasises that measurement of the Asn⁸⁴ PSA variant in clinical assays will be dependent on both a lowered catalytic activity as well as the epitope of the antibody used. Validation of this hypothesis awaits characterisation using biophysical and/or X-ray crystallographic methods.

The Asn⁸⁴ PSA had differences in glycosylation, stability and proteolytic activity which may impact on the measurement of PSA levels in men. Savblom et al., (15) observed an association of the [GG] genotype of rs61752561 SNP with higher total PSA concentration in serum (GG=188, GA=13, AA=0). Since the PSA level in patients can be confounded by multiple factors such as elevation in PSA levels associated with the malignant progression which is also dependent on “time to event”, stage and grade, we undertook this analysis in healthy controls from the MDC-cohort with long follow-up. Although in our study we haven’t analysed the PSA concentration in patient serum samples, we also observed lower tPSA and fPSA levels with the minor allele [A] similar to Savblom et al. (15). Furthermore, we observed a difference in directionality of our recombinant PSA and genotype correlation in clinical samples for the total and F/T PSA ratio this could be due to the different forms of PSA and their complexed forms that exist in blood which needs further clarification. Nevertheless our study still indicates that the rs61752561 SNP exerts biochemical effects that may impact on clinically measured PSA levels.

Aberrant glycosylation analysis in cancers including PCa has recently evolved as a promising tool to improve cancer diagnostics (29). Glycosylation is often known to also influence intracellular localisation of the protein which was not analysed in-depth in our analysis. Mammalian and yeast cell expression have different glycosylation patterns (30) which may impact on some of our findings. N-glycosylation in yeast is characterized by hypermannosylation (31). A recent quantitative glycoproteomics study revealed that the Asn⁸⁴ is characterised by high-mannose glycans unlike the glycan at Asn⁵¹ that is predominantly fucosylated (32). Although there may be differences in glycans at other positions including Asn⁵¹, we observed the same activity, and functional outcomes, from our recombinant yeast PSA variants as with the transfected human PC3 cell line. Thus, using custom designed lectin binding assays to discriminate mannose and fucosyl side chains (33, 34) will allow accurate measurement of the Asn⁸⁴ PSA isoform and the WT PSA, respectively. Further, the current calibrants used in clinical assays are not appropriate for the Asn⁸⁴ isoform. We anticipate, that in the future, appropriate standards (different PSA isoforms) will need to be used in clinical assays as we move towards assessing an individual’s genotype to improve the accuracy of PSA test.

In summary, this study sought to assess the effect of PCa risk-associated SNPs at the KLK locus identified as an independent signal by a recent fine-mapping study. We investigated the biological function of the KLK3 rs61752561 SNP and identified the SNP isoform to have reduced activity, an extra glycosylation and differential substrate activity that likely impacts on the role of PSA in PCa pathogenesis. This analysis supports the observed association of the SNPs to lower the risk of PCa in patients carrying the minor allele [A] of the SNP. With its potential biological effects on PSA activity and PSA detection, inclusion of this PSA coding SNP in PCR assays or lectin-binding assays that accounts for the differences in glycosylation for the Asn⁸⁴ PSA, in conjunction with the conventional PSA test, may improve the specificity and sensitivity of the current PSA test.

Abbreviations:

SNPs

single nucleotide polymorphisms

PCa

prostate cancer

KLK3

kallikrein related peptidase 3

PSA

Prostate-Specific Antigen

tPSA

total PSA

fPSA

free PSA

F/T PSA

free/total PSA

GWAS

genome-wide association studies

OR

odds-ratio

LD

linkage-disequilibrium

ACT

α−1-antichymotrypsin

A2M

alpha-2-macroglobulin

PCI

protein C inhibitor

API

alpha 1- protease inhibitor

MAF

minor-allele frequency

WT

wild-type

EK

enterokinase

Ala

alanine

PNGase

peptide-N-glycosidase F

DSF

differential scanning fluorimetry

mAb

monoclonal antibody

TBS

tris-buffered saline

MMP2

matrix metalloproteinase 2

IGFBP-3

insulin-like growth factor-binding protein 3

IFMA

immunofluorometric assay

MD

molecular dynamics

PDB

protein data bank

MDC

Malmö Diet and Cancer

DTT

dithiothreitol

Human genes listed in the manuscript

KLK3

kallikrein-related peptidase, prostate-specific antigen

IGFBP-3

insulin-like growth factor-binding protein 3

MMP2

Matrix-metalloproteinase 2

SERPINA3

anti-chymotrypsin/ACT

A2M

alpha-2-macroglobulin

SERPINA5

protein C inhibitor/PCI

SERPINA1

alpha 1- protease inhibitor/API