The role of plasma microseminoprotein-beta in prostate cancer: an observational nested case–control and Mendelian randomization study in the European prospective investigation into cancer and nutrition

Abstract Background Microseminoprotein-beta (MSP), a protein secreted by the prostate epithelium, may have a protective role in the development of prostate cancer. The only previous prospective study found a 2% reduced prostate cancer risk per unit increase in MSP. This work investigates the association of MSP with prostate cancer risk using observational and Mendelian randomization (MR) methods. Patients and methods A nested case–control study was conducted with the European Prospective Investigation into Cancer and Nutrition (EPIC) with 1871 cases and 1871 matched controls. Conditional logistic regression analysis was used to investigate the association of pre-diagnostic circulating MSP with risk of incident prostate cancer overall and by tumour subtype. EPIC-derived estimates were combined with published data to calculate an MR estimate using two-sample inverse-variance method. Results Plasma MSP concentrations were inversely associated with prostate cancer risk after adjusting for total prostate-specific antigen concentration [odds ratio (OR) highest versus lowest fourth of MSP = 0.65, 95% confidence interval (CI) 0.51–0.84, Ptrend = 0.001]. No heterogeneity in this association was observed by tumour stage or histological grade. Plasma MSP concentrations were 66% lower in rs10993994 TT compared with CC homozygotes (per allele difference in MSP: 6.09 ng/ml, 95% CI 5.56–6.61, r2=0.42). MR analyses supported a potentially causal protective association of MSP with prostate cancer risk (OR per 1 ng/ml increase in MSP for MR: 0.96, 95% CI 0.95–0.97 versus EPIC observational: 0.98, 95% CI 0.97–0.99). Limitations include lack of complete tumour subtype information and more complete information on the biological function of MSP. Conclusions In this large prospective European study and using MR analyses, men with high circulating MSP concentration have a lower risk of prostate cancer. MSP may play a causally protective role in prostate cancer.


Introduction
Microseminoprotein-beta (MSP) is a protein secreted by the prostate epithelium into the seminal fluid [1]. In the only previous prospective study, the Multiethnic Cohort (MEC) [2], a 1 ng/ ml increase in circulating MSP concentration was associated with a 2% decrease in prostate cancer risk. MSP concentrations, in both blood and semen samples from healthy males, are 60% higher among CC homozygotes versus TT homozygotes for rs10993994 (r 2 ¼ 0.38 and 0.23, respectively), located 57 basepairs upstream in the 5 0 promoter region of the MSMB gene [3], which encodes the protein MSP. Furthermore, a genome-wide association study (GWAS) has found carriers of the T allele to have an elevated prostate cancer risk (57% higher for TT versus CC) [2,4].
This prospective study investigated whether circulating MSP concentrations were associated with prostate cancer risk in the European Prospective Investigation into Cancer and Nutrition (EPIC). We then investigated the association of rs10993994 with circulating concentrations of MSP in EPIC and used this genetic variant as an instrument for MSP to assess its potential causal role through Mendelian randomization (MR) analyses by combining EPIC-derived estimates with published data from the Prostate Cancer Association Group to Investigate Cancer Associated Alterations in the Genome (PRACTICAL) consortium [5].

Study population
Totally, 137 000 men participating in EPIC provided blood samples at recruitment between 1992 and 2000 [6]. Lifestyle questionnaires, anthropometric data, and food questionnaires were collected at recruitment. All participants provided written informed consent. Approval for the study was obtained from the ethical review boards of the participating institutions and the International Agency for Research on Cancer (IARC). The current study uses data from Germany, Greece, Italy, the Netherlands, Spain and the UK.

Follow-up
Cancer incidence was identified through record linkage to regional or national registries in most countries (see supplementary methods, available at Annals of Oncology online). Follow-up procedures continued to prostate cancer diagnosis or last follow-up completed (31 December 2007 to 14 June 2010).
Cases were men who were diagnosed with incident prostate cancer (International Classification of Diseases 10th revision code C61 [7]) after blood collection and before the end of follow-up. An incidence density sampling protocol was used to select control participants at random from the cohort of men who were alive and free of cancer (excluding non-melanoma skin cancer) at the time of diagnosis of the index case and who matched on study centre, length of follow-up, age at blood collection, time of blood collection and duration of fasting at blood collection. These analyses included 1871 cases with 1871 matched controls. Information on tumour stage and grade at diagnosis was available for 1263 (67.5%) and 1554 (85.1%) of cases, respectively (see supplementary methods, available at Annals of Oncology online).

Assessment of analytes
Immunoassay measurements for prostate-specific antigen (PSA) [8] and MSP [9,10] were conducted on the AutoDelfia V R 1235 automatic immunoassay system in Dr Lilja's laboratory at the Wallenberg Research Laboratories, Department of Translational Medicine, Lund University, Skåne University Hospital, Malmö, Sweden (see supplementary methods, available at Annals of Oncology online).

Statistical analysis
Analyte concentrations below limits of detection were set to half the lowest concentration (PSA, N ¼ 7), and concentrations above the upper limits were set to the highest value for that analyte (MSP, N ¼ 82; PSA, N ¼ 65). Pearson's v 2 tests and paired t-tests were conducted between matched case-control sets for anthropometric and lifestyle characteristics. Analysis of variance was used to assess differences in analyte concentrations in controls by strata of selected characteristics, country and study phase (matched case-control sets were identified after each of three rounds of follow-up and end point data centralisation in EPIC conducted in approximately 2004, 2008 and 2010, and samples from each phase were assayed together). Log transformations were applied to analyte concentrations and results are presented as geometric means adjusted for age at blood collection, body mass index (BMI), recruitment centre and laboratory batch.
Conditional logistic regression models were used to examine the association of MSP with prostate cancer, conditioned on the matching factors and adjusted for BMI, age at blood collection and further adjusted for fourth of PSA concentration (additional adjustment was shown to not materially alter the results, see supplementary Table S1, available at Annals of Oncology online). These analyses were repeated in subgroups according to study phase, time between blood collection and diagnosis, age at blood collection, age at diagnosis, prostate tumour stage and histological grade. Additional unconditional analyses stratified by median PSA concentration and smoking status were adjusted for age, BMI, fourth of PSA concentration and matching factors. Linear trend was tested using a pseudo-continuous variable equal to medians of the fourths of MSP concentration. For subgroup analyses, likelihood ratio tests were used to test for heterogeneity.rs10993994 genotype data were available for a subset of 1068 EPIC cases and 1186 EPIC controls from the iCOGS [11], OncoArray [12] and Breast and Prostate Cancer Cohort Consortium (BPC3) [13] genotyping projects. Logistic regression models were used to investigate the association of rs10993994 with prostate cancer.
We investigated the potential causal role of MSP in prostate cancer risk using MR analyses. A summary estimate of the association of rs10993994 with prostate cancer was taken from the iCOGS genotyping project in the international consortium PRACTICAL with 25 000 cases from 32 studies [5,11], and from EPIC prostate cancer cases and controls genotyped in the OncoArray [12] and BPC3 studies [13]. Summary estimates for the association of rs10993994 with MSP were calculated using these EPIC data [12,13]. We used the MR-Base platform to do a phenome-wide association scan for rs10993994 with 850 traits to check for pleiotropy [14], and also checked the NHGR-EBI catalogue of published GWAS [15]. Two-sample MR estimates were calculated separately using summary estimates for each of PRACTICAL (iCOGS) [5] and EPIC-derived rs10993994-prostate cancer risk estimates with the EPIC-derived rs10993994-MSP estimate, which were then combined using the inverse-variance weighted method. To address possible confounding by PSA, we conducted sensitivity analyses using the summary association of rs10993994 with residuals from a linear regression of log total PSA on MSP, also calculated within EPIC.
All statistical tests are two-sided and were conducted using STATA software version 14 (StataCorp LP, College Station, TX).

Results
Data from 1871 cases and 1871 matched controls were included in the analyses. The median age at blood collection was 58 years, and, for cases, the median time between blood collection and diagnosis was 8.3 years. No significant differences were observed in selected baseline characteristics between cases and controls ( Table 1).
MSP concentration in controls was higher in men older at blood collection, not married, with normal/low BMI or low-alcohol intake, and who had higher educational attainment (P < 0.05 for all). Compared with never smokers, men who smoked more than 15 cigarettes per day had 30% higher MSP concentrations (Ptrend < 0.0001). PSA concentration was positively associated with age at blood collection and educational attainment, and negatively associated with greater BMI and diabetes (Table 2). MSP and PSA concentrations were positively correlated in both cases and controls (partial correlations r ¼ 0.3 and 0.2, respectively, P < 0.0001).
MSP concentration was not associated with prostate cancer risk after adjustment for age at blood collection and BMI [odds ratio (OR) for highest versus lowest fourth ¼ 0.98, 95% CI 0.82-1.19, Ptrend ¼ 0.9)]. However, after adjustment for PSA, MSP concentration was associated with prostate cancer risk (OR ¼ 0.65, 95% CI 0.51-0.84, Ptrend ¼ 0.001) ( Table 3). There was some evidence of heterogeneity in the association by time to diagnosis (with a stronger association in men diagnosed within 8.5 years of baseline, Pheterogeneity ¼ 0.009), age at diagnosis (Pheterogeneity ¼ 0.03); (supplementary Table S2, available at Annals of Oncology online) and recruitment country (Pheterogeneity ¼ 0.02; supplementary Table S3, available at Annals of Oncology online). There was no significant After correction for multiple testing, no significant association of rs10993994 genotype was observed with potential confounders beyond PSA concentrations in controls (supplementary Table S7, available at Annals of Oncology online). PheWAS using published data [14,15], showed that besides prostate cancer risk, rs10993994 is associated only with the prostate cancer biomarkers PSA and prostate cancer antigen 3 (PCA3) at the genome-wide significance level. An inverse-variance weighted MR showed a one unit increase in

Discussion
In this large prospective study, we found a lower prostate cancer risk in men with higher circulating concentrations of MSP after adjustment for circulating PSA concentrations. MSP is a protein in the immunoglobulin-binding factor family primarily secreted by epithelial cells, which may have a role in tumour suppression [16] and pathogen defence [17]. These findings are in agreement with the only other published prospective investigation [2], which found an inverse association between circulating MSP concentration and prostate cancer; in the MEC study, MSP concentration was inversely associated with prostate cancer risk before and after adjustment for PSA, though the association was much stronger after adjustment, as is to be expected due to the strong positive association of PSA concentration with risk and the moderate positive association of PSA with MSP. In accordance with previous findings [2], we found no evidence that the association of MSP with risk differed by tumour stage or grade, although small numbers of cases in subgroups may have limited power to evaluate heterogeneity. We found some modest evidence for heterogeneity by country and age at diagnosis, but the results are difficult to interpret due to small numbers in subgroups and multiple statistical tests. Short follow-up time (3.8 years) and thus reverse causality was previously suggested in MEC as a possible explanation for the observed association. The present study has more than double the average follow-up (8.3 years), and while we found some observational evidence that the inverse association between MSP and prostate cancer is stronger for men diagnosed closer to blood collection, the apparent differences by time to diagnosis may be at least in part due to differences in the case mix, with cases diagnosed closer to baseline being more likely to be younger at diagnosis. Furthermore, our MR findings suggest that reverse causality is unlikely to explain the overall relationship, with genetic variation in MSP affecting lifetime levels of MSP.
MSP is also secreted at lower levels by epithelial cells in the tracheobronchial tree [18,19]. Smoking has been associated with a 2.5fold increase in expression of MSP in the airway epithelium when compared with non-smokers [20]. Therefore, some variation in MSP concentrations may be due to smoking-induced secretory cell hyperplasia in the respiratory tract. To our knowledge, we are the only study to report higher levels of MSP among current smokers compared with non-smokers. We found no strong evidence of heterogeneity in the MSP association by smoking status but more data are needed to examine this and particularly to assess the association in non-smokers in whom any potential masking effect of smoking on circulating MSP is not present.
The strength of the current MR result stems from the use of rs10993994 as an instrumental variable; rs10993994 lies in the promotor region of the MSMB region, the locus that encodes MSP, and rs10993994 is strongly associated with circulating MSP concentrations and prostate cancer [2,5]. In general, the use of variants in the cis-acting protein-encoding locus is one of the most robust scenarios of MR [21] and a recent review of MSP function [22] suggest the rs10993994 genetic association is specific to MSP. An association of rs10993994 has been observed with concentrations of prostate cancer markers PSA and PCA3 in prostate cancer controls [15], and it remains possible that PSA may confound these results. However, given that the associations of rs10993994 with PSA and PCA3 levels are observed only in controls and that MR results were materially robust to adjustment for PSA concentration, the association of rs10993994 with PSA (and PCA3) may arise from collider bias. Such collider bias [23], which induces the association of rs10993994 with PSA and PCA3 when stratifying on prostate cancer disease status, should not invalidate the results of the MR analysis (which is not stratified on disease status). Additionally, for the biological role of PSA to confound these findings, PSA would have to be causal to prostate cancer development for which there is little evidence.

Conclusion
Using observational data from a prospective nested case-control study and MR, this study supports a possible protective role of MSP in the development of prostate cancer. Experimental studies are needed to elucidate the mechanisms through which MSP may influence prostate cancer development. If shown to be true from randomized clinical trials, therapies that raise MSP levels may provide novel opportunities for the treatment and prevention of prostate cancer.
Data availability: For information on how to submit an application for gaining access to EPIC data and/or biospecimens, please follow the instructions at http://epic.iarc.fr/access/index. php