Patients with prostate cancer for this study were selected from the Prostate Clinical Research Information System and Specimen Tracking Inventory Program databases at the Dana-Farber Cancer Institute [21
]. The Prostate Clinical Research Information System is a central repository of patient data, including comprehensive follow-up of all patients. To be eligible for this study, patients had to have a diagnosis of localized/locally advanced prostate cancer (i.e. stage T3 or less, N0 and M0); consented and donated blood for research before undergoing any type of local therapy; and consented to be followed clinically for research purposes. Of 778 patients who fulfilled these study criteria, 753 were selected according to the availability of complete clinical data and genomic DNA. Plasma collected before any type of therapy was available for selenium assessment in 489 of the selected patients.
The prognostic risk at diagnosis was categorized using modified criteria of D’Amico et al
], as: low risk (≤ T2a and PSA level ≤ 10 ng/mL and Gleason sum ≤ 6); intermediate risk (T2b or PSA level 10–20 ng/mL, or Gleason sum 7); and high risk (> T2b or PSA level > 20 ng/mL or Gleason sum > 7). The primary outcome of interest was the presentation of aggressive prostate cancer at diagnosis, defined as stage T2b-T3, or PSA level > 10 ng/mL or biopsy Gleason 7 (corresponding to the intermediate-/high-risk categories).
Blood was withdrawn using EDTA as an anticoagulant and centrifuged at 1000 g for 10 min at 17 °C. Purified plasma was aliquoted and stored at − 80 °C until used for the analysis. This process was complete 2–16 h after sample withdrawal. Genomic DNA was prepared from EDTA-blood using the QIAamp DNA Blood mini kit (Qiagen Inc, Valencia, CA, USA) within 24 h after withdrawal and stored at 4 °C. DNA concentration was assayed using pico-green (Invitrogen, Carlsbad, CA, USA), and adjusted to 5 ng/μL for genotype analysis.
We initially identified 56 SNPs in SOD1, SOD2
, GPX1, GPX4
, and XRCC1
, derived from the National Cancer for Biotechnology Information database by heterozygosity ratio (≥ 0.05). Six SNPs that did not conform to Hardy–Weinberg equilibrium (P
< 0.01) were removed. Any SNP with a minor allele frequency of < 5% (six) was also removed from analysis. Among the remaining 44 SNPs, tagging polymorphisms were selected using the Haploview procedure (http://www.broad.mit.edu/mpg/haploview/
) by setting a pair-wise linkage disequilibrium (LD) mode to (r2
≥ 0.8 and logarithm of odds, LOD, ≥ 3). In all, 26 SNPs that captured most of the haplotypes in a region of LD were selected and examined for their association with risk of aggressive prostate cancer.
Except for the GCG repeat and rs1050450 within GPX1, all other SNPs were analysed by sequential PCR-mass spectrometry systems (Sequenom, San Diego, CA, USA) ‘Increase Plexing Efficiency and Flexibility for MassARRAY System Assay or homogeneous Mass EXTENDED assay’. Methods for assessing the variants in GPX1 are described below.
The GCG repeat SNP within GPX1
was analysed using a previously described procedure [24
] with minor modifications. Genomic DNA (40 ng) was amplified using 17 pmol each of primers (forward: 5′FAM
-GAAACTGCCTGTGCCACGTGACC-3′ and reverse: 5′-CGAGAAGGCATACACCGACTGGGC-3′) in 22 μL PCR buffer (Qiagen) containing 1.5 mM MgCl2
, 1.8 mM dNTP, Q solution, 1.5 units of Taq polymerase (all Qiagen). The PCR reaction had an initial denaturing temperature at 94 °C (2 min) followed by 35 cycles of denaturing (94 °C; 30 s), annealing (62.5 °C; 1 min), and extension (72 °C; 30 s) steps. An 8-min extension at 72 °C followed the final cycle. The 1 μl of PCR product was diluted with water to 50 μL; 2 μL of diluted PCR product were mixed with 10 μL formamide, 0.25 μL Gene Scan-500 LIZ Size Standard (Applied Biosystems, Foster City, CA, USA), and water to adjust the final volume to 20 μL. This mixture was applied to a POP-7 capillary array which was linked to an automated fluorescence detection system, ABI 3730 (Applied Biosystems). Using ‘Genemapper Software v4.0’ and ‘Peak Scanner Software v1.0’ for the analysis, the GCG repeat number was calculated as ((fragment length – 154 bp)/3 = (GCG)n
). This equation was confirmed by sequencing 30 PCR samples.
For rs1050450 in GPX1
, a SNP (CCC/Pro or CTC/Leu) within GPX1
locates in exon-2 at the amino-acid position between 198 and 200. Shifts in amino-acid position depend on the number of GCG repeats (4–6x) in exon-1. Consequently, this SNP is referred either as Pro198Leu or Pro200Leu. This was analysed using a previously described procedure [25
] with minor modifications. Genomic DNA (20 ng) was amplified using 12.5 pmol each of primers (forward: 5′-CTACGCAGGTACAGCCGCCGCT-3′ and reverse: 5′-CAGGTGTTCCTCCCTCGTAGGT-3′) in 12.5 μL 60 mM Tris-HCl (pH 9.5) buffer containing 15 mM ammonium sulphate, 2 mM MgCl2
, 1.6 mM dNTP, and 0.6 units platinum Taq DNA Polymerase (Invitrogen, Carlsbad, CA, USA). The PCR had an initial denaturing temperature at 94 °C (2 min) followed by 35 cycles of denaturing (94 °C; 30 s), annealing (62.5 °C; 1 min), and extension (72 °C; 30 s) steps. An 8-min extension at 72 °C followed the final cycle. 7.5 μL of PCR product were digested by incubating with 25 units of ApaI (New England BioLabs, Beverly, MA, USA) at 25 °C for 2 h. Digested products were visualized on a 2% agarose gel stained with ethidium bromide. Fragment patterns specific for three genotypes were: Pro/Pro (CCC/CCC; 118 bp, 74 bp), Pro/Leu (CCC/CTC; 192 bp, 118 bp, 74 bp), and Leu/Leu (CTC/CTC; 192 bp).
Plasma selenium level was analysed using a previously described procedure in the laboratory of Irena King (Fred Hutchinson Cancer Research Center [26
]). Diluted 99 : 300 in 0.5% Triton X-100, plasma selenium concentration (μg/L) was analysed by flame-less atomic absorption (Perkin-Elmer 5000; Perkin Elmer Corp., Norwalk, CT, USA) using an electrode-less discharge lamp operating at λ
= 196.0 nm and a 1′Vov platform graphite furnace. Twenty-seven of 489 plasma samples were analysed in duplicate with ‘blind’ numbering; the median coefficient of variation was within 5.3%. For these samples, the first measurement was used in the analysis.
Patient disease characteristics at diagnosis were summarized as counts and percentages, or as median (range), and interquartile range of levels. Plasma selenium levels between genotypes were compared using the Wilcoxon rank-sum test. Associations of disease aggressiveness with genotypes were evaluated using a chisquare test. Relative risk (RR) and 95% CI compared to common homozygote were estimated using a generalized linear model for binomial data with a log-link rather than a logit-link function. Associations of disease aggressiveness with selenium levels were evaluated using a Cochran-Armitage test for trend, where selenium levels were categorized to five ordered groups according to the quintile thresholds (108.3, 118.0, 125.5, 139.7 μg/L, respectively; equivalent to 1.08, 1.18, 1.26, 1.40 ppm). The likelihood ratio test from the generalized linear model was used to test for an interaction between genotypes and selenium levels on disease aggressiveness, where selenium levels were evaluated both as quintile groups and continuous values. All analyses were conducted with P < 0.05 (two-sided) considered to indicate statistical significance.