- Email: [email protected]
Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin
EURURO-6132; No. of Pages 9 EUROPEAN UROLOGY XXX (2015) XXX–XXX
available at www.sciencedirect.com journal homepage: www.europeanurology.com
Prostate Cancer
Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin Teemu J. Murtola a,b,*, Tiina Wahlfors c, Antti Haring a, Kimmo Taari d, Ulf-Ha˚kan Stenman e, Teuvo L.J. Tammela a,b, the PRACTICAL Consortiumy, Johanna Schleutker c,f, Anssi Auvinen g a
School of Medicine, University of Tampere, Tampere, Finland;
b
Department of Urology, Tampere University Hospital, Tampere, Finland;
c
Institute of
d
Biosciences and Medical Technology / BioMediTech and Fimlab Laboratories, University of Tampere, Tampere, Finland; Department of Urology, University of Helsinki and Helsinki University Hospital, Helsinki, Finland;
e
Department of Clinical Chemistry, University of Helsinki, Helsinki, Finland; f Institute of
Biomedicine, Medical Biochemistry and Genetics, University of Turku, Turku, Finland; g School of Health Sciences, University of Tampere, Tampere, Finland
Article info
Abstract
Article history: Accepted March 10, 2015
Background: Energy metabolism is important in cancer proliferation and progression, but its role in prostate cancer (PCa) remains unclear. Objective: We explored whether single-nucleotide polymorphisms (SNPs) of genes involved in energy metabolic pathways are associated with PCa risk and prognosis, and whether antidiabetic treatment modifies any such association. Design, setting, and participants: The PRACTICAL Consortium genotyped 397 SNPs among 3241 screened participants (including 801 PCa cases) in the Finnish Prostate Cancer Screening Trial and 1983 hospital-based PCa cases. Information on medication use was obtained from a national prescription database. Outcome measurements and statistical analysis: Genetic risk scores were calculated in terms of SNPs associated with PCa incidence or survival at a significance level of p < 5 103. Hazard ratios for PCa and disease-specific death were calculated via Cox regression modelling. The predictive value of the genetic risk score was evaluated using receiver operating characteristic and Harrell’s c-index analyses. Results and limitations: A total of 30 SNPs were associated with PCa risk and ten SNPs with survival. The genetic risk score was consistently associated with PCa survival. The risk association was non-significantly weaker in metformin users. The genetic risk score did not improve prediction of PCa risk, but slightly improved the ability to predict PCa survival when added to conventional predictors (c-index improved from 87.4 to 87.9; p < 0.001). A limitation is that information on diabetes apart from medication use was unavailable for the study population. Conclusions: SNPs of genes involved in energy metabolic pathways are associated with PCa survival. This suggests an important role of glucose metabolism in PCa progression, which could point to new avenues for prevention of PCa death. Patient summary: Genetic changes in glucose and energy metabolic pathways are associated with a higher risk of high-risk prostate cancer and adverse outcomes. # 2015 European Association of Urology. Published by Elsevier B.V. All rights reserved.
Keywords: Glucose metabolism Prostate cancer Risk Survival
y Members of the PRACTICAL Consortium are listed in Supplementary File 1. * Corresponding author. School of Medicine, Building M, Room 313, University of Tampere, PL 2000, 33521 Tampere, Finland. Tel. +358 3 31165015; Fax: +358 3 31164358. E-mail address: teemu.murtola@uta.fi (T.J. Murtola).
http://dx.doi.org/10.1016/j.eururo.2015.03.026 0302-2838/# 2015 European Association of Urology. Published by Elsevier B.V. All rights reserved.
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 2
EUROPEAN UROLOGY XXX (2015) XXX–XXX
1.
Introduction
Factors affecting prostate cancer (PCa) progression are not well understood, and further knowledge is needed. Cancer progression requires energy, and the central role of metabolic reprogramming towards anaerobic glycolysis in cancer cells has long been recognised [1]. However, the role of glucose and energy metabolism in PCa development and progression is unclear, although PCa mutations are more common in the mitochondrial genome than in autosomal chromosomes [2]. Diabetes mellitus, a condition affecting systemic glucose and energy balance, is inversely correlated with overall PCa risk [3], but diabetic men may have more poorly differentiated cancer [4]. This supports claims of the importance of glucose metabolism in PCa. The antidiabetic drug metformin inhibits PCa cell growth in vitro, and its use may decrease PCa mortality [5,6]. Polymorphisms in genes involved in glucose metabolism have not been evaluated as a PCa risk factor before, although polymorphisms in mitochondrial genes [7] and insulin-like growth factors [8] may not be associated with a higher risk of PCa. We evaluated associations between single-nucleotide polymorphisms (SNPs) in the genes of glucose metabolism pathways and the PCa risk and prognosis. Furthermore, we explored whether these SNPs predict PCa risk and death better than conventional predictive factors do, and we examined the possible modification of effect exerted by metformin and other antidiabetic drugs.
2.2.
SNP genotypes
In all, 397 SNPs distributed across 80 genes (Supplementary Table 1) involved in the glucose metabolic pathway were genotyped as part of an international PRACTICAL Consortium effort using a custom SNP panel (iCOGs). The SNPs were selected via the UCSC genome browser (http:// genome.ucsc.edu/).
2.3.
Use of antidiabetic medication
The screening trial cohort was linked to the national prescription database of the Social Insurance Institution of Finland (SII) [11] via a unique personal identification number. Information on antidiabetic medication use was available for 1995–2009.
2.4.
Statistical analysis
Associations between individual SNPs and PCa risk and death were tested using the genome association analysis software PLINK. Permutation analysis and Bonferroni correction were used to correct p values for multiple testing. SNPs meeting the a priori defined significance criterion (association with PCa risk or death at a significance level of p 0.005) were further tested using multivariate-adjusted regression models. SNPs associated with PCa risk or mortality at p 0.001 in a Cox regression model including all SNPs were selected for further analysis. A genetic risk score was calculated after obtaining of a correlation coefficient for each significant SNP from the multivariate-adjusted Cox regression for PCa risk and death. Individual coefficients were summed to compose a polygenic risk score as described previously [12]. Hazard ratios (HRs) with 95% confidence intervals (CIs) for PCa diagnosis and death were calculated via the Cox proportional hazards regression method. The time metric was years since the screening trial randomisation (incidence analysis in the screening trial cohort) or years since PCa diagnosis (survival analysis considering both cohorts).
2.
Patients and methods
2.1.
Study population
We applied adjustment for age (age-adjusted model; analyses of PCa risk and mortality) and for PSA level at diagnosis, Gleason score, and the presence of distant metastases (multivariate-adjusted model; analyses of disease recurrence and survival after PCa diagnosis). Receiver operating characteristics were used to compare the
Two previously described study cohorts [9] were used. The screening
predictive accuracy of the genetic risk score (area under the curve,
trial cohort consisted of a randomly selected sample of 3241 men among
AUC) for PCa grade and stage together with age and PSA level. Harrell’s
32 000 men randomised to the screening arm of the Finnish Prostate
c-index analysis was used to compare prediction for PCa survival
Cancer Screening Trial (FinPCST) followed during 1996–2013. The trial
between Cox regression models including conventional predictors of PCa
protocol has been described previously [10]. The men were invited to
death (PSA at diagnosis, tumour stage, Gleason grade) and a model also
prostate-specific antigen (PSA) screening at 4-yr intervals. Among the
including the genetic risk score. Graphical model comparison was
genotyped men, 801 new PCa cases were identified during a median
performed using decision curve analysis.
follow-up of 16 yr.
To estimate the effect of antidiabetic drug usage on the SNP-cancer
The hospital cohort comprised 1983 PCa cases diagnosed outside
association, we stratified the analysis by metformin usage overall and for
systematic screening protocols. The patients were treated at
men with cumulative amount of use above the median (>2100 g in
Tampere University Hospital between 1990 and 2010 and followed
total), and by non-metformin antidiabetic drug usage. Interaction
until 2013.
related to metformin use was evaluated via addition of an interaction
The clinical data included diagnosis date, clinical TNM stage, Gleason
term with genetic risk score to the analysis.
grade, and date and cause of death. For the hospital cohort, the date of
Adjusted means for PSA concentrations were determined for the
disease progression after primary treatment was also obtained. This was
baseline value (PSA measured at the first screening) and for all PSA
defined as two consecutive increases in PSA after reaching the nadir or
values measured during the screening trial. Adjusted means were
radiographic disease progression in patients managed with endocrine
calculated by fitting the natural logarithm of PSA into an age-adjusted or
treatment (258 progressions; 33% within the treatment subgroup),
multivariate-adjusted (for age and use of 5a-reductase inhibitors and
PSA >0.2 ng/ml in patients treated with radical prostatectomy (148;
antidiabetic drugs) linear regression model.
34.9%), or a rising PSA level >2 ng/ml above the nadir in radiationtreated men (91; 39.6%).
Cox regression and ROC analyses were performed using SPSS v.20 statistical software (IBM, Chicago, IL, USA). Adjusted means for PSA
Deaths for which PCa (ICD10 code C61) was the official underlying
values were estimated via Stata v.12 (StataCorp, College Station, TX,
cause as obtained from Statistics Finland and death certificates were
USA). Harrells’s c-index and decision curve analysis were performed
considered to be PCa deaths.
using R version 3.1.2 R Project for Statistical Computing, Vienna, Austria).
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 3
EUROPEAN UROLOGY XXX (2015) XXX–XXX
All statistical tests were two-sided. Results with p 0.05 were considered statistically significant.
3.
Results
3.1.
Population characteristics
In the screening trial cohort, 34.2% of the PCa cases had Gleason 7–10 disease and 2.0% had metastatic disease at diagnosis (Table 1). During the median follow-up of 8.4 yr after diagnosis, 44 of the men (1.4%) died from PCa (Table 1). In comparison to the screening trial cohort, PCa cases in the hospital cohort more often had high-grade or metastatic disease (Table 1). Within a median follow-up of 9.1 yr after diagnosis, 623 (31.2%) experienced disease progression and 284 (14.2%) died of PCa. Cases from the hospital cohort showed higher median PSA at diagnosis and were more often managed with endocrine therapy, whereas surgery and radiation therapy were more common in the screening cohort (Table 1). 3.2.
Permutation analysis
In total, 30 SNPs were associated with PCa risk at the significance level defined a priori. In addition, 10 SNPs were associated with PCa death (Supplementary Table 2). The prevalence of risk SNPs was generally comparable between the study cohorts. 3.3.
PSA concentration among SNP carriers
The combined genetic risk for PCa incidence was associated with elevated PSA and the risk of screening positivity (Table 2). The combined genetic risk score for PCa mortality did not predict PSA at baseline, but higher PSA during consecutive screening rounds was observed (Table 2). Of the individual SNPs, 26 were significantly associated with PSA; seven with higher PSA and 19 with lower PSA (data not shown). Polymorphisms of PDK1 and PGM1 showed varying association with PSA: PDK1 C2_POS173159592 carriers had higher mean PSA, whereas 16 other PDK1 SNPs were associated with lower PSA (Table 2). A similar difference was observed for the PGM1 SNPs rs855314 (higher PSA) and rs2269247 (lower PSA). 3.4.
PCa risk
In the screening trial cohort, the genetic risk score was associated with higher PCa risk overall and with high-grade and metastatic disease at diagnosis (Table 3 and Fig. 1A,B). The risk score did not predict disease stage or grade in the hospital cohort (Table 3). Similarly, the risk score was associated with higher risk of Gleason 8–10 PCa in the screening cohort (HR 1.56, 95% CI 1.31–1.86) but not in the hospital cohort (odds ratio [OR] 0.88, 95% CI 0.67–1.17). Of the individual SNPs, PGM1 rs855314 carriers in the hospital cohort had a higher risk of high-grade PCa, while
Table 1 – Characteristics of the study population of 3241 men in the screening arm of the Finnish Prostate Cancer Screening Trial (screening trial cohort) and 1983 prostate cancer cases diagnosed outside screening protocols and treated at Tampere University Hospital (hospital cohort) Screening trial cohort Men (n) Median follow-up after randomisation, yr (range) Median follow-up after PCa diagnosis, yr (range) Median body mass index, kg/m2 (range) a PCa cases, n (% of the cohort) Median PSA in the first screening round (ng/ml) Median PSA at diagnosis (ng/ml) Tumour Gleason grade b 6, n (% of cases) 7–10, n (% of cases) Unknown Tumour stage Localised, n (% of cases) Metastatic, n (% of cases) c Unknown Deaths, n (% of the cohort) PCa deaths, n (% of the cohort) d Disease progression, n (% of all cases) e Primary treatment, n (% of cases) Prostatectomy Radiation therapy (external beam or brachytherapy) Endocrine therapy Active surveillance or watchful waiting Unknown
Hospital cohort
3241 16 (0.6–17)
1983 NA
8.4 (0.3–16.5)
9.1 (0–36.1)
27 (16–50)
NA
801 (24.7) 0.65
1,983 (100) NA
5.36
10.6
523 (65.3) 274 (34.2) 4 (0.5)
865 (43.6) 820 (41.4) 298 (15.0)
687 (85.8) 16 (2.0) 99 (12.4) 431 (13.3) 44 (1.4) NA
1,704 (85.9) 183 (9.2) 96 (4.8) 971 (48.7) 284 (14.2) 483 (24.4)
311 (38.8) 270 (33.7)
407 (20.5) 230 (11.6)
82 (10.2) 138 (17.2)
683 (34.4) 98 (4.9)
0
565 (28.5)
PCa = prostate cancer; PSA = prostate-specific antigen; NA = not applicable. Available for 2378 men (73.4% of the cohort). Gleason score from prostate biopsy sample. c All M1 cases, as judged by clinical M stage or bone scan when available, regardless of the T or N stage. d Prostate cancer (ICD10 code C61) as the primary cause of death. e Either PSA progression (defined as two consecutive increases in PSA after reaching the nadir or radiographic disease progression in patients managed with endocrine treatment; PSA >0.2 ng/ml in patients treated with radical prostatectomy; or a rising PSA level >2 ng/ml above the nadir in radiationtreated men). An additional 140 men had disease progression but were excluded because of missing data on PSA at the time of progression. a
b
PGM2 rs2995948 predicted overall PCa risk in the screening cohort and metastatic disease in the hospital cohort. Finally, HK2 rs965078 was associated with a higher risk of metastatic disease in the hospital cohort (Table 3). PDK1 C2_POS173159592 differed in effect from the other PDK1 SNPs. C2_POS173159592 was positively associated with overall and metastatic PCa risk in the screening cohort, with a suggestion of increase in high-grade PCa risk (p = 0.065) in the hospital cohort (Table 3). Other PDK1 SNPs were associated with lower overall PCa risk, and most were also linked to lower risk of high-grade disease. 3.5.
PCa prognosis
The genetic risk score correlated with increased risk of PCa death in both study cohorts after adjustment for disease
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 4
EUROPEAN UROLOGY XXX (2015) XXX–XXX
Table 2 – Adjusted mean PSA at baseline and from three screening rounds stratified by carrier status for SNPs of genes involved in the energy metabolic pathway among 3241 men participating in the screening arm of the Finnish Prostate Cancer Screening Trial between 1996 and 2009 Carriers (n)
Genetic risk score PCa risk Median or below Above the median p value for trend PCa mortality Median or below Above the median p value for trend SNP PDK1 C2_POS173159592 Wild type One minor allele Two minor alleles All mutation carriers p value for trend PDK1 C2_POS173156629 Wild type One minor allele Two minor alleles All mutation carriers p value for trend GBE1 rs10440071 Wild type One minor allele Two minor alleles All mutation carriers p value for trend GPI rs2910411 Wild type One minor allele Two minor alleles All mutation carriers p value for trend HK2 rs13360277 Wild type One minor allele Two minor alleles All mutation carriers p value for trend
Baseline Mean PSAAA
Mean PSAMVA
1631 1610
0.97 1.07 0.027
0.97 1.07 0.030
1633 1608
0.99 1.05 0.209
2354 1213 152 1365
Combined
Risk of screening positivity OR (95% CI) b
Mean PSAAA
Mean PSAMVA
Risk of screening positivity OR (95% CI)
Reference 1.34 (1.06–1.71)
0.99 1.07 <0.001
1.03 1.08 <0.001
Reference 1.12 (1.10–1.14)
0.99 1.05 0.193
Reference 1.15 (0.90–1.46)
1.01 1.06 0.051
1.01 1.06 0.039
Reference 1.12 (0.98–1.29)
1.03 1.07 1.34 * 1.09 * 0.032
1 1.03 1.3 * 1.06 * 0.032
Reference 1.15 (0.90–1.47)
0.99 1.03 1.19 * 1.05 * 0.003
0.99 1.03 1.19 * 1.05 * 0.003
Reference 1.17 (1.01–1.35)
3359 206 3 209
1.07 0.86 0.75 0.86 0.02
1.03 0.83 * 0.77 0.83 * 0.019
Reference 0.71 (0.39–1.28)
1.08 0.86 0.63 0.86 <0.001
1.06 0.84 0.66 0.83 <0.001
Reference 0.52 (0.35–0.77)
2604 1017 98 1115
1.03 1.14 0.85 1.11 0.317
1.00 1.10 0.82 1.07 0.339
Reference 1.38 (1.07–1.77)
1.03 1.14 * 0.92 1.11 * 0.037
0.99 1.08 * 0.87 1.06 * 0.046
Reference 1.19 (1.03–1.38)
3261 438 20 458
1.05 1.06 1.06 1.06 0.863
1.02 1.03 1.01 1.03 0.896
Reference 0.93 (0.65–1.35)
0.99 1.11 * 1.44 1.12 * <0.001
0.99 1.1 * 1.43 1.12 * <0.001
Reference 1.31 (1.08–1.59)
3482 235 2 237
1.04 1.32 * 0.45 1.32 * 0.004
1.00 1.28 * 0.35 1.27 * 0.005
Reference 1.55 (1.00–2.41)
1.05 1.23 * 0.40 1.23 * 0.001
1.03 1.21 * 0.28 1.20 * 0.001
Reference 1.31 (1.02–1.69)
a
c
*
*
PSA = prostate-specific antigen; SNP = single-nucleotide polymorphism; AA = age-adjusted; MVA = multivariate-adjusted. p < 0.05 for difference compared to wild-type carriers. a Men with serum PSA 4 ng/ml and men with serum PSA >3 ng/ml but <4 ng/ml and s free/total PSA ratio <16% were considered screening-positive. b Age-adjusted estimate calculated via logistic regression. c Similar decreasing associations with PSA were observed for PDK1 C2_POS173170045, C2_POS173170839, C2_POS173177578, rs109305, rs8179584, rs16860693, rs1530864, rs2357637, rs836628, rs2290563, rs12476089, rs12693006, and rs6433368. *
stage, grade, and PSA at diagnosis (Table 4). By contrast, no correlation with the risk of disease progression was observed (Table 4). Of the SNPs associated with increased PCa risk, PDK1 C2_POS173159592 had an association with PCa mortality that was of borderline significance in the screening trial cohort (HR 1.45, 95% CI 0.97–2.16; Table 4). The other PDK1 SNPs were associated with an increased risk of disease progression and a suggestion of an increase in risk of death (C2_POS173177578, C2_POS173170839, rs10930561, rs1530864, rs2357637, and rs12693006) or no difference in
survival (rs12693006 and rs2290563). Two minor alleles of PGM1 rs855314 were linked to lower PCa survival in the hospital cohort (HR 1.95, 95% CI 1.03–3.69); a similar but nonsignificant increase in risk was observed in the screening trial cohort (Table 4). 3.6.
ROC and Harrell’s c-index analysis
When PCa cases from both study cohorts were combined for ROC analysis, the genetic risk score in conjunction with PSA and age at diagnosis did not improve the prediction of
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 5
EUROPEAN UROLOGY XXX (2015) XXX–XXX
Table 3 – PCa risk among carriers of SNPs of genes involved in the energy metabolic pathway for 3241 men in the screening arm of the Finnish Prostate Cancer Screening Trial (screening trial cohort) and 1983 prostate cancer cases detected outside screening protocols and treated at the Tampere University Hospital (hospital cohort) Overall PCa risk
Genetic risk score Median or below Above the median p value for trend c SNP PDK1 C2_POS173159592 Wild type One minor allele Two minor alleles All mutation carriers p value for trend PGM1 rs855314 Wild type One minor allele Two minor alleles All mutation carriers p value for trend PGM2 rs2995948 e Wild type One minor allele Two minor alleles All mutation carriers p value for trend HK2 rs965078 Wild type One minor allele Two minor alleles All mutation carriers p value for trend
High-grade PCa
Screening trial cohort
Screening trial cohort
Cases (n)
Cases (n)
323 478
HR (95% CI)
a
HR (95% CI)
a
Metastatic PCa
Hospital cohort Cases (n)
OR (95% CI)
b
Screening trial cohort Cases (n)
HR (95% CI)
a
Hospital cohort Cases (n)
OR (95% CI)
b
Reference 1.57 (1.37–1.81) <0.001
110 164
Reference 1.66 (1.30–2.11) < 0.001
402 418
Reference 1.05 (0.86–1.27) 0.651
3 13
Reference 5.19 (1.48–18.24) 0.010
95 88
Reference 0.92 (0.67–1.25) 0.740
489 271 41 312
Reference 1.05 (0.91–1.22) 1.39 (1.01–1.91) 1.09 (0.94–1.26) 0.097
164 95 15 110
Reference 1.10 (0.85–1.41) 1.58 (0.93–2.68) 1,15 (0.90–1.46) 0.136
489 293 38 331
Reference 1.15 (0.93–1.41) 1.47 (0.89–2.42) 1.18 (0.97–1.44) 0.065
4 10 2 12
Reference 4.68 (1.47–14.93) 9.46 (1.73–51.79) 5.11 (1.65–15.85) 0.001
120 58 5 63
Reference 0.88 (0.63–1.22) 0.63 (0.25–1.59) 0.85(0.62–1.17) 0.250
601 189 11 200
Reference 1.07 (0.90–1.25) 0.77 (0.42–1.39) 1.04 (0.89–1.22) 0.845
204 66 4 70
Reference 1.11 (0.84–1.47) 0.76 (0.28–2.04) 1.09 (0.83–1.42) 0.725
471 138 15 153
Reference 1.11 (0.84–1.46) 2.81 (1.07–7.38) 1.18 (0.90–1.54) 0.098
11 5 0 5
Reference 1.54 (0.53–4.44) – 1.34 (0.46–3.87) 0.850
110 34 4 8
Reference 1.13 (0.75–1.71) 1.87 (0.62–5.60) 1.18 (0.80–1.76) 0.295
432 326 43 369
Reference 1.18 (1.02–1.36) 0.88 (0.65–1.21) 1.14 (0.99–1.31) 0.341
148 112 4 116
Reference 1.19 (0.93–1.52) 0.83 (0.48–1.43) 1.13 (0.89–1.44) 0.660
338 245 41 286
Reference 1.07 (0.85–1.36) 0.92 (0.58–1.46) 1.05 (0.84–1.32) 0.887
11 5 0 5
Reference 0.66 (0.23–1.90) – 0.56 (0.19–1.60) 0.189
68 68 12 80
Reference 1.5 (1.05–2.14) 1.46 (0.75–2.82) 1.49 (1.06–2.10) 0.040
299 364 135 499
Reference 0.93 (0.80–1.08) 1.08 (0.88–1.32) 0.96 (0.84–1.11) 0.773
109 118 46 164
Reference 0.82 (0.63–1.06) 1.04 (0.74–1.47) 0.87 (0.68–1.11) 0.741
216 294 110 404
Reference 1.12 (0.87–1.44) 1.17 (0.84–1.63) 1.13 (0.90–1.44) 0.307
7 5 4 9
Reference 0.55 (0.17–1.73) 1.35 (0.40–4.62) 0.75 (0.28–2.00) 0.889
40 72 36 108
Ref 1.39 (0.93–2.09) 2.01 (1.24–3.26) 1.55 (1.06–2.27) 0.005
d
PCa = prostate cancer; SNP = single-nucleotide polymorphism; HR = hazard ratio; CI = confidence interval; OR = odds ratio. From Cox regression analysis with model adjustment for age at baseline. From logistic regression analysis with model adjustment for age at diagnosis. c With the risk score as a continuous variable. d Inverse association with total prostate cancer risk in the screening trial cohort, but not with metastatic prostate disease, was observed for other PDK1 SNPs: C2_POS173217127, C2_POS173170839, rs10930561, rs8179584, rs16860693, rs1530864, rs2357637, rs836628, rs2290563, rs12476089, rs12693006, and rs6433368. Of these, C2_POS173170839, C2_POS173217127, C2_POS173177578, rs10930561, rs1530864, and rs2357637 were associated with a lower risk of high-grade prostate cancer. e A similar risk association was observed for PGM2 rs2911874. a
b
high-grade or metastatic PCa relative to PSA and age alone (data not shown). Addition of the genetic risk score to the established risk factors PSA, age, tumour stage, and Gleason grade in the Cox regression model slightly improved the predictive value for PCa death (c-indices of 87.9 and 87.4 for models with and without the genetic risk score, respectively; model improvement p < 0.001; Fig. 2).
The association between genetic risk and PCa death was slightly weaker among metformin users than non-users of antidiabetic drugs, but the confidence intervals overlapped (Fig. 3C). Body mass index did not significantly affect the association between the genetic risk score and PCa risk in users or non-users of antidiabetic drugs (Supplementary Table 3).
3.7.
4.
Impact of metformin and other antidiabetic drugs
Neither metformin nor non-metformin antidiabetic drugs significantly modified the association between genetic risk score and either PCa risk overall or high-grade disease, although the genetic risk score was a significant risk predictor only in men who had not used antidiabetic drugs (Fig. 3A,B).
Discussion
The combined genetic risk score based on risk SNPs of glucose and energy metabolic genes predicted PCa risk, disease stage and grade, and prognosis in men under systematic PCa screening. Our results suggest a role of energy metabolism in the development and progression of PCa. The risk score predicted PCa death in both study
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 6
EUROPEAN UROLOGY XXX (2015) XXX–XXX
(A)
0.4 Age-adjusted HR 1.57; 95% Cl 1.37–1.81
Cumulative hazard for prostate cancer
p < 0.001
Genetic risk score above the median
0.3
Genetic risk score at or below the median
0.2
0.1
0.0 0.00
5.00
10.00
15.00
20.00
Time since FinPCST randomisation (yr)
Cumulative hazard for Gleason 7–10 prostate cancer
(B)
0.12
Age-adjusted HR 1.66; 95% Cl 1.30–2.11 Genetic risk score above the median
p < 0.001 0.10
0.08 Genetic risk score at or below the median
0.06
0.04
0.02
0.00 0.00
5.00
10.00
15.00
20.00
Time since FinPCST randomisation (yr) Fig. 1 – Prostate cancer risk among men stratified by a genetic risk score based on single-nucleotide polymorphisms of genes involved in glucose and energy metabolic pathways. Cohort of 3241 participants from the screening arm of the Finnish Prostate Cancer Screening Trial (FinPCST) between 1995 and 2009. The p value for trends were calculated by adding the genetic risk score as a continuous variable to the age-adjusted Cox regression model. (A) Overall risk of prostate cancer. (B) Risk of Gleason 7–10 prostate cancer. HR = hazard ratio; CI = confidence interval.
cohorts, but not tumour grade or stage in the hospital cohort. The difference can probably be attributed to differences in disease characteristics between the two cohorts: the role of genetic risk factors may be clearer in a screened population with a high proportion of screen-detected low-grade or earlystage cases. Systematic screening and the resulting stage migration of PCa cases at detection could probably modify the role of genetic risk factors. Because the risk prediction ability of the genetic risk score may vary with environmental
factors, our findings needs to be validated in additional study populations [13]. PDK and PGM1 SNPs exhibited consistent associations with both PCa risk and survival rates in both study cohorts. These genes in particular may play a role in PCa initiation and progression. PDK1 (pyruvate dehydrogenase kinase 1) inhibits the activity of PDH (pyruvate dehydrogenase), a mitochondrial gatekeeping enzyme linking cytosolic glycolysis to the
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 7
EUROPEAN UROLOGY XXX (2015) XXX–XXX
Table 4 – PCa progression and mortality among carriers of SNPs of genes involved in the energy metabolic pathway for 3241 men in the screening arm of the Finnish Prostate Cancer Screening Trial (screening trial cohort) and 1983 prostate cancer cases diagnosed outside screening protocols and treated at the Tampere University Hospital (the hospital cohort) Disease progression Cases (n)
HR (95% CI)
PCa mortality Deaths (n)
a
HR (95% CI)
PCa survival, PCa cases only Screening trial cases
a
Deaths (n) Genetic risk score Median or below Above the median p value for trend SNP PDK1 C2_POS173159592 Wild type One minor allele Two minor alleles All mutation carriers p value for trend PGM1 rs855314 Wild type One minor allele Two minor alleles All mutation carriers p value for trend
HR (95% CI)
Hospital cohort cases
a
Deaths (n)
HR (95% CI)
a
All cases combined
Deaths (n)
HR (95% CI)
a
244 239
Reference 1.00 (0.83–1.19) 0.97
16 28
Reference 1.45 (0.97–2.17) 0.015
16 28
Reference 1.71 (0.92–3.17) 0.014
121 163
Reference 1.40 (1.10–1.77) 0.001
137 191
Reference 1.51 (1.15–1.98) <0.001
303 163 17 180
Reference 0.95 (0.78–1.15) 0.90 (0.55–1.46) 0.95 (0.79–1.14) 0.52
23 17 4 21
Reference 1.44 (0.95–2.19) 1.48 (0.64–3.45) 1.45 (0.97–2.16) 0.09
23 17 4 21
Reference 0.92 (0.48–1.76) 1.52 (0.52–4.47) 0.99 (0.53–1.84) 0.74
191 84 9 93
Reference 0.82 (0.63–1.06) 0.78 (0.40–1.52) 0.82 (0.63–1.05) 0.12
214 101 13 114
Reference 0.80 (0.60–1.07) 0.85 (0.42–1.73) 0.81 (0.61–1.06) 0.84
366 109 8 117
Reference 0.98 (0.79–1.22) 0.93 (0.46–1.88) 0.98 (0.79–1.21) 0.81
32 11 1 12
Reference 1.45 (0.94–2.25) 1.49 (0.37–6.11) 1.46 (0.95–2.23) 0.09
32 11 1 12
Reference 1.11 (0.56–2.20) 1.72 (0.23–1.79) 1.14 (0.59–2.22) 0.63
216 58 10 68
Reference 1.02 (0.77–1.37) 1.95 (1.03–3.69) 1.10 (0.84–1.45) 0.23
248 69 11 80
Reference 1.07 (0.78–1.47) 2.10 (0.99–4.48) 1.14 (0.84–1.55) 0.20
PCa = prostate cancer; HR = hazard ratio; CI = confidence interval. From Cox regression analysis with model adjustment for age and PSA at diagnosis, and prostate cancer stage and grade.
a
0.05
tricarboxylic acid (TCA) cycle in mitochondria [14]. PDH is essential for oncogene-induced senescence, a cellular defence mechanism against malignant transformation; PDK1 inhibits this enzyme and is essential for cancer cell growth [14]. PDK1 also has a role in adaptation to a hypoxic cancer environment, and is directly activated by HIF (hypoxia-inducible factor) [1]. Inactivation of PDH via increased PDK1 activity reduces shuttling of pyruvate into
None All
0.03
Harrell’s c-index for Age, Gleason grade, M stage, and PSA: 87.4
0.02
Age, Gleason grade, M stage, PSA, and genec risk score: 87.9 Model improvement: p < 0.001
0.00
0.01
Net benefit
0.04
Age at dg, tumour Gleason grade, M stage and PSA at dg Age at dg, tumour Gleason grade, M stage, PSA at dg and the genetic risk score
0.0
0.2
0.4
0.6
0.8
Threshold probability Fig. 2 – Decision curve analysis for prostate cancer survival after diagnosis using the genetic risk score and known prognostic factors among prostate cancer cases from the combined study cohorts. dg = diagnosis; PSA = prostate-specific antigen.
mitochondria, facilitating its conversion to lactate in the cytosol (the Warburg effect). Because less pyruvate is shuttled into mitochondria, the amount of intermediates produced by the TCA cycle decreases. If further studies confirm that the PDK1 SNPs tested in our study affect enzyme activity, our results are consistent with this phenomenon occurring in PCa. Known cessation of secretion of citrate (a TCA intermediate) from the prostate epithelium during carcinogenesis supports a change in the role of the TCA cycle [15]. PGM1 SNPs were associated with increased PCa risk and poorer survival. This is the first enzyme in glycogenesis. A recent study showed that, like PDK1, PGM1 expression is increased by HIF under hypoxic conditions [16]. Glycogen storage is a survival mechanism of tumour cells for withstanding glucose and oxygen starvation; therefore, it likely has importance for metastatic progression. PDK1 C2_POS173159592 exhibited associations with PSA and PCa risk that are opposite to those observed for other PDK1 SNPs. This suggests that the minor allele for C2_POS173159592 is inherited with the major allele for the rest of the gene. We plan to further investigate this issue using haplotype analysis in future studies. Nevertheless, all PDK1 SNPs exhibited at least a suggestive association with disease outcome, pointing to a role of PDK1 in PCa progression. SNPs of several other genes encoding enzymes that regulate the conversion of glucose to pyruvate (HK2, GPI, and PCK1) and of pyruvate within mitochondria (PC, DLD, and SDHD) were associated with either PCa mortality or risk, but without consistency across cohorts.
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 8
EUROPEAN UROLOGY XXX (2015) XXX–XXX
(A)
2.5
HR (95%Cl)
2 1.5 1 0.5
Interacon with meormin use: p = 0.21
0 Non-users
HR (95% Cl)
(B)
Non-meormin an-DM drug users
Meormin users
2.10 (0.89–4.99)
1.29 (0.92–1.81)
1.63 (1.39–1.91)
Meormin amount in upper 50% 1.42 (0.87–2.33)
2.5
HR (95% Cl)
2 1.5 1 0.5 Interacon with meormin use: p = 0.094 0
HR (95%Cl)
(C)
Non-users
Non-meormin an-DM drug users
Meormin users
Meormin amount in upper 50%
1.85 (1.40–2.43)
1.22 (0.36–4.11)
1.10 (0.62–1.96)
1.14 (0.56–2.33)
3
Interacon with meormin use: p = 0.171
HR (95% Cl)
2.5 2 1.5 1
caused by direct antineoplastic medication effects rather than the underlying diabetes. The strength of our study lies in the ability to evaluate associations between SNPs and multiple PCa outcomes and to validate the findings with two distinct study cohorts. We were also able to evaluate the impact of PCa screening on the gene-risk association and to combine registry-based information on antidiabetic medication use with genotype data. The registry records medication purchases independently of PCa status. Therefore, the results are not affected by recall bias. Our study also has limitations. Diabetes was assessed in terms of antidiabetic drug usage; we did not have information on diabetes managed via diet. We also had no information on glucose control. This could have led to exposure misclassification and subsequent underestimation of the effects of antidiabetic drugs. Furthermore, we had only limited data on body mass index and cardiovascular disease, and no information on lifestyle risk factors that may have differed between diabetic and nondiabetic men, possibly affecting PCa risk or outcome. However, it is likely that such factors are not related to genotype, and thus are unlikely to bias our estimation of SNP associations. Our median follow-up after PCa diagnosis was 8–9 yr. Therefore, we were not able to reliably estimate very long-term effects of genetic risk factors on PCa survival. Our analyses of the effects of metformin on PCa were limited by the small sample size (only 133 medication users among PCa cases), so our confidence intervals were wide. Finally, our study is limited by its retrospective design, with the possibility of residual confounding by unmeasured variables.
0.5
5.
Conclusions
0
HR (95% Cl)
Non-users
Meormin users
2.40 (1.13–5.09)
0.15 (0.01–2.86)
Meormin use in upper 50% 1.00 (0.13–7.62)
Fig. 3 – Effect of antidiabetic drug use on the association between prostate cancer risk and survival and the genetic risk score calculated based on single-nucleotide polymorphisms of genes involved in the glucose and energy metabolic pathways. Cohort of 3241 participants in the screening arm of the Finnish Prostate Cancer Screening Trial between 1995 and 2009. Men with a genetic risk score above the median compared to those with a risk score below the median. (A) Overall prostate cancer risk. Hazard ratio (HR) from Cox regression with model adjustment for age. (B) Risk of high-grade (Gleason 7–10) prostate cancer. HR from Cox regression with model adjustment for age. (C) Risk of prostate cancer death. HR from Cox regression with model adjustment for age, PSA at diagnosis, tumour Gleason grade, and M stage. CI = confidence interval; DM = diabetes mellitus.
We demonstrated that genetic risk in terms of SNPs of genes involved in cellular glucose and energy metabolism is associated with PCa survival. These genes probably participate in metabolic rewiring of cancer cells to enhance disease progression. This points to the possibility of new avenues for prevention of PCa death. The clinical utility of these SNPs in PCa management should be investigated in validation studies in other populations. Author contributions: Teemu J. Murtola had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Murtola, Wahlfors, Schleutker, Auvinen. Acquisition of data: All authors. Analysis and interpretation of data: Murtola, Wahlfors, Schleutker,
Lower PCa mortality has been reported for metformin users [6]. In our analysis, the link between PCa survival and genetic risk was only slightly weaker among metformin users than among non-users and among users of other antidiabetic drugs. Thus, our results do not strongly support the previous results, although the associations did tend to change in the same direction. However, it remains unclear whether the associations found in epidemiologic studies are
Auvinen. Drafting of the manuscript: All authors. Critical revision of the manuscript for important intellectual content: All authors. Statistical analysis: Murtola, Wahlfors, Auvinen. Obtaining funding: Tammela, Auvinen. Administrative, technical, or material support: Taari, Tammela. Supervision: Schleutker, Auvinen. Other: None.
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 EUROPEAN UROLOGY XXX (2015) XXX–XXX
9
Financial disclosures: Teemu J. Murtola certifies that all conflicts of
[4] Abdollah F, Briganti A, Suardi N, et al. Does diabetes mellitus
interest, including specific financial interests and relationships and
increase the risk of high-grade prostate cancer in patients under-
affiliations relevant to the subject matter or materials discussed in the
going radical prostatectomy? Prostate Cancer Prostatic Dis 2011;
manuscript (eg, employment/affiliation, grants or funding, consultan-
14:74–8.
cies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Teemu J.
[5] Brown KA, Samarajeewa NU, Simpson ER. Endocrine-related cancers and the role of AMPK. Mol Cell Endocrinol 2013;366:170–9.
Murtola has received lecture fees from Janssen-Cilag and has received
[6] Spratt DE, Zhang C, Zumsteg ZS, Pei X, Zhang Z, Zelefsky MJ.
financial support for participation in scientific congresses from Ferring
Metformin and prostate cancer: reduced development of castration-
and Janssen-Cilag. Teuvo L.J. Tammela has received consultancy fees
resistant disease and prostate cancer mortality. Eur Urol 2013;
from Astellas, GlaxoSmithKline, Pfizer, Orion Pharma, and Amgen. Kimmo Taari has received consultancy fees from GlaxoSmithKline,
63:709–16. [7] Wang L, McDonnell SK, Hebbring SJ, et al. Polymorphisms in
Pfizer, Orion Pharma, and Amgen. The remaining authors have nothing to
mitochondrial genes and prostate cancer risk. Cancer Epidemiol
disclose.
Biomarkers Prev 2008;17:3558–66. [8] Tsilidis KK, Travis RC, Appleby PN, et al. Insulin-like growth factor
Funding/Support and role of the sponsor: This work was supported by the European Commission’s Seventh Framework Programme (HEALTH-F22009-223175), Cancer Research UK (C5047/A7357, C1287/A10118, C5047/A3354, C5047/A10692, C16913/A6135), and the National Institutes of Health (Cancer Post-Cancer GWAS initiative grant 1 U19 CA 148537-01). The sponsors played a role in data collection. Acknowledgments: We are grateful to Dr. Elizabeth Platz and Dr. William
pathway genes and blood concentrations, dietary protein and risk of prostate cancer in the NCI Breast and Prostate Cancer Cohort Consortium (BPC3). Int J Cancer 2013;133:495–504. [9] Eeles RA, Olama AA, Benlloch S, et al. Identification of 23 new prostate cancer susceptibility loci using the iCOGS custom genotyping array. Nat Genet 2013;45:385–91. [10] Kilpela¨inen TP, Tammela TL, Malila N, et al. Prostate cancer mor-
B. Isaacs, Johns Hopkins University, for helpful comments and advice
tality in the Finnish randomized screening trial. J Natl Cancer Inst
during preparation of the manuscript.
2013;105:719–25. [11] Martikainen J, Rajaniemi S. Drug reimbursement systems in EU member states, Iceland and Norway. Social Security and Health Report 54. Helsinki: Social Insurance Institution of Finland; 2002.
Appendix A. Supplementary data
http://www.kela.fi/in/internet/liite.nsf/ABID/030303101726PN/ $File/Drug_reimbursement.pdf [12] Pashayan N, Duffy SW, Neal DE, et al. Implications of polygenic risk-
Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j. eururo.2015.03.026.
[13] Vickers A. Prediction models in urology: are they any good, and
References
[14] Kaplon J, Zheng L, Meissl K, et al. A key role for mitochondrial
stratified screening for prostate cancer on overdiagnosis. Genet Med. In press. http://dx.doi.org/10.1038/gim.2014.192 how would we know anyway? Eur Urol 2010;57:571–3. gatekeeper pyruvate dehydrogenase in oncogene-induced senes-
[1] Warburg O. On the origin of cancer cells. Science 1956;123:309–14.
cence. Nature 2013;498:109–12.
[2] Lindberg J, Mills IG, Klevebring D, et al. The mitochondrial and
[15] Zadra G, Photopoulos C, Loda M. The fat side of prostate cancer.
autosomal mutation landscapes of prostate cancer. Eur Urol 2013;
Biochim Biophys Acta 2013;1831:1518–32. [16] Pelletier J, Bellot G, Gounon P, Lacas-Gervais S, Pouysse´gur J,
63:702–8. [3] Xu H, Jiang HW, Ding GX, Zhang H, Mao SH, Ding Q. Diabetes
Mazure NM. Glycogen synthesis is induced in hypoxia by the
mellitus and prostate cancer risk of different grade or stage: a
hypoxia-inducible factor and promotes cancer cell survival. Front
systematic review and meta-analysis. Diabetes Res Clin Pract
Oncol 2012;2:18.
2013;99:241–9.
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
available at www.sciencedirect.com journal homepage: www.europeanurology.com
Prostate Cancer
Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin Teemu J. Murtola a,b,*, Tiina Wahlfors c, Antti Haring a, Kimmo Taari d, Ulf-Ha˚kan Stenman e, Teuvo L.J. Tammela a,b, the PRACTICAL Consortiumy, Johanna Schleutker c,f, Anssi Auvinen g a
School of Medicine, University of Tampere, Tampere, Finland;
b
Department of Urology, Tampere University Hospital, Tampere, Finland;
c
Institute of
d
Biosciences and Medical Technology / BioMediTech and Fimlab Laboratories, University of Tampere, Tampere, Finland; Department of Urology, University of Helsinki and Helsinki University Hospital, Helsinki, Finland;
e
Department of Clinical Chemistry, University of Helsinki, Helsinki, Finland; f Institute of
Biomedicine, Medical Biochemistry and Genetics, University of Turku, Turku, Finland; g School of Health Sciences, University of Tampere, Tampere, Finland
Article info
Abstract
Article history: Accepted March 10, 2015
Background: Energy metabolism is important in cancer proliferation and progression, but its role in prostate cancer (PCa) remains unclear. Objective: We explored whether single-nucleotide polymorphisms (SNPs) of genes involved in energy metabolic pathways are associated with PCa risk and prognosis, and whether antidiabetic treatment modifies any such association. Design, setting, and participants: The PRACTICAL Consortium genotyped 397 SNPs among 3241 screened participants (including 801 PCa cases) in the Finnish Prostate Cancer Screening Trial and 1983 hospital-based PCa cases. Information on medication use was obtained from a national prescription database. Outcome measurements and statistical analysis: Genetic risk scores were calculated in terms of SNPs associated with PCa incidence or survival at a significance level of p < 5 103. Hazard ratios for PCa and disease-specific death were calculated via Cox regression modelling. The predictive value of the genetic risk score was evaluated using receiver operating characteristic and Harrell’s c-index analyses. Results and limitations: A total of 30 SNPs were associated with PCa risk and ten SNPs with survival. The genetic risk score was consistently associated with PCa survival. The risk association was non-significantly weaker in metformin users. The genetic risk score did not improve prediction of PCa risk, but slightly improved the ability to predict PCa survival when added to conventional predictors (c-index improved from 87.4 to 87.9; p < 0.001). A limitation is that information on diabetes apart from medication use was unavailable for the study population. Conclusions: SNPs of genes involved in energy metabolic pathways are associated with PCa survival. This suggests an important role of glucose metabolism in PCa progression, which could point to new avenues for prevention of PCa death. Patient summary: Genetic changes in glucose and energy metabolic pathways are associated with a higher risk of high-risk prostate cancer and adverse outcomes. # 2015 European Association of Urology. Published by Elsevier B.V. All rights reserved.
Keywords: Glucose metabolism Prostate cancer Risk Survival
y Members of the PRACTICAL Consortium are listed in Supplementary File 1. * Corresponding author. School of Medicine, Building M, Room 313, University of Tampere, PL 2000, 33521 Tampere, Finland. Tel. +358 3 31165015; Fax: +358 3 31164358. E-mail address: teemu.murtola@uta.fi (T.J. Murtola).
http://dx.doi.org/10.1016/j.eururo.2015.03.026 0302-2838/# 2015 European Association of Urology. Published by Elsevier B.V. All rights reserved.
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 2
EUROPEAN UROLOGY XXX (2015) XXX–XXX
1.
Introduction
Factors affecting prostate cancer (PCa) progression are not well understood, and further knowledge is needed. Cancer progression requires energy, and the central role of metabolic reprogramming towards anaerobic glycolysis in cancer cells has long been recognised [1]. However, the role of glucose and energy metabolism in PCa development and progression is unclear, although PCa mutations are more common in the mitochondrial genome than in autosomal chromosomes [2]. Diabetes mellitus, a condition affecting systemic glucose and energy balance, is inversely correlated with overall PCa risk [3], but diabetic men may have more poorly differentiated cancer [4]. This supports claims of the importance of glucose metabolism in PCa. The antidiabetic drug metformin inhibits PCa cell growth in vitro, and its use may decrease PCa mortality [5,6]. Polymorphisms in genes involved in glucose metabolism have not been evaluated as a PCa risk factor before, although polymorphisms in mitochondrial genes [7] and insulin-like growth factors [8] may not be associated with a higher risk of PCa. We evaluated associations between single-nucleotide polymorphisms (SNPs) in the genes of glucose metabolism pathways and the PCa risk and prognosis. Furthermore, we explored whether these SNPs predict PCa risk and death better than conventional predictive factors do, and we examined the possible modification of effect exerted by metformin and other antidiabetic drugs.
2.2.
SNP genotypes
In all, 397 SNPs distributed across 80 genes (Supplementary Table 1) involved in the glucose metabolic pathway were genotyped as part of an international PRACTICAL Consortium effort using a custom SNP panel (iCOGs). The SNPs were selected via the UCSC genome browser (http:// genome.ucsc.edu/).
2.3.
Use of antidiabetic medication
The screening trial cohort was linked to the national prescription database of the Social Insurance Institution of Finland (SII) [11] via a unique personal identification number. Information on antidiabetic medication use was available for 1995–2009.
2.4.
Statistical analysis
Associations between individual SNPs and PCa risk and death were tested using the genome association analysis software PLINK. Permutation analysis and Bonferroni correction were used to correct p values for multiple testing. SNPs meeting the a priori defined significance criterion (association with PCa risk or death at a significance level of p 0.005) were further tested using multivariate-adjusted regression models. SNPs associated with PCa risk or mortality at p 0.001 in a Cox regression model including all SNPs were selected for further analysis. A genetic risk score was calculated after obtaining of a correlation coefficient for each significant SNP from the multivariate-adjusted Cox regression for PCa risk and death. Individual coefficients were summed to compose a polygenic risk score as described previously [12]. Hazard ratios (HRs) with 95% confidence intervals (CIs) for PCa diagnosis and death were calculated via the Cox proportional hazards regression method. The time metric was years since the screening trial randomisation (incidence analysis in the screening trial cohort) or years since PCa diagnosis (survival analysis considering both cohorts).
2.
Patients and methods
2.1.
Study population
We applied adjustment for age (age-adjusted model; analyses of PCa risk and mortality) and for PSA level at diagnosis, Gleason score, and the presence of distant metastases (multivariate-adjusted model; analyses of disease recurrence and survival after PCa diagnosis). Receiver operating characteristics were used to compare the
Two previously described study cohorts [9] were used. The screening
predictive accuracy of the genetic risk score (area under the curve,
trial cohort consisted of a randomly selected sample of 3241 men among
AUC) for PCa grade and stage together with age and PSA level. Harrell’s
32 000 men randomised to the screening arm of the Finnish Prostate
c-index analysis was used to compare prediction for PCa survival
Cancer Screening Trial (FinPCST) followed during 1996–2013. The trial
between Cox regression models including conventional predictors of PCa
protocol has been described previously [10]. The men were invited to
death (PSA at diagnosis, tumour stage, Gleason grade) and a model also
prostate-specific antigen (PSA) screening at 4-yr intervals. Among the
including the genetic risk score. Graphical model comparison was
genotyped men, 801 new PCa cases were identified during a median
performed using decision curve analysis.
follow-up of 16 yr.
To estimate the effect of antidiabetic drug usage on the SNP-cancer
The hospital cohort comprised 1983 PCa cases diagnosed outside
association, we stratified the analysis by metformin usage overall and for
systematic screening protocols. The patients were treated at
men with cumulative amount of use above the median (>2100 g in
Tampere University Hospital between 1990 and 2010 and followed
total), and by non-metformin antidiabetic drug usage. Interaction
until 2013.
related to metformin use was evaluated via addition of an interaction
The clinical data included diagnosis date, clinical TNM stage, Gleason
term with genetic risk score to the analysis.
grade, and date and cause of death. For the hospital cohort, the date of
Adjusted means for PSA concentrations were determined for the
disease progression after primary treatment was also obtained. This was
baseline value (PSA measured at the first screening) and for all PSA
defined as two consecutive increases in PSA after reaching the nadir or
values measured during the screening trial. Adjusted means were
radiographic disease progression in patients managed with endocrine
calculated by fitting the natural logarithm of PSA into an age-adjusted or
treatment (258 progressions; 33% within the treatment subgroup),
multivariate-adjusted (for age and use of 5a-reductase inhibitors and
PSA >0.2 ng/ml in patients treated with radical prostatectomy (148;
antidiabetic drugs) linear regression model.
34.9%), or a rising PSA level >2 ng/ml above the nadir in radiationtreated men (91; 39.6%).
Cox regression and ROC analyses were performed using SPSS v.20 statistical software (IBM, Chicago, IL, USA). Adjusted means for PSA
Deaths for which PCa (ICD10 code C61) was the official underlying
values were estimated via Stata v.12 (StataCorp, College Station, TX,
cause as obtained from Statistics Finland and death certificates were
USA). Harrells’s c-index and decision curve analysis were performed
considered to be PCa deaths.
using R version 3.1.2 R Project for Statistical Computing, Vienna, Austria).
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 3
EUROPEAN UROLOGY XXX (2015) XXX–XXX
All statistical tests were two-sided. Results with p 0.05 were considered statistically significant.
3.
Results
3.1.
Population characteristics
In the screening trial cohort, 34.2% of the PCa cases had Gleason 7–10 disease and 2.0% had metastatic disease at diagnosis (Table 1). During the median follow-up of 8.4 yr after diagnosis, 44 of the men (1.4%) died from PCa (Table 1). In comparison to the screening trial cohort, PCa cases in the hospital cohort more often had high-grade or metastatic disease (Table 1). Within a median follow-up of 9.1 yr after diagnosis, 623 (31.2%) experienced disease progression and 284 (14.2%) died of PCa. Cases from the hospital cohort showed higher median PSA at diagnosis and were more often managed with endocrine therapy, whereas surgery and radiation therapy were more common in the screening cohort (Table 1). 3.2.
Permutation analysis
In total, 30 SNPs were associated with PCa risk at the significance level defined a priori. In addition, 10 SNPs were associated with PCa death (Supplementary Table 2). The prevalence of risk SNPs was generally comparable between the study cohorts. 3.3.
PSA concentration among SNP carriers
The combined genetic risk for PCa incidence was associated with elevated PSA and the risk of screening positivity (Table 2). The combined genetic risk score for PCa mortality did not predict PSA at baseline, but higher PSA during consecutive screening rounds was observed (Table 2). Of the individual SNPs, 26 were significantly associated with PSA; seven with higher PSA and 19 with lower PSA (data not shown). Polymorphisms of PDK1 and PGM1 showed varying association with PSA: PDK1 C2_POS173159592 carriers had higher mean PSA, whereas 16 other PDK1 SNPs were associated with lower PSA (Table 2). A similar difference was observed for the PGM1 SNPs rs855314 (higher PSA) and rs2269247 (lower PSA). 3.4.
PCa risk
In the screening trial cohort, the genetic risk score was associated with higher PCa risk overall and with high-grade and metastatic disease at diagnosis (Table 3 and Fig. 1A,B). The risk score did not predict disease stage or grade in the hospital cohort (Table 3). Similarly, the risk score was associated with higher risk of Gleason 8–10 PCa in the screening cohort (HR 1.56, 95% CI 1.31–1.86) but not in the hospital cohort (odds ratio [OR] 0.88, 95% CI 0.67–1.17). Of the individual SNPs, PGM1 rs855314 carriers in the hospital cohort had a higher risk of high-grade PCa, while
Table 1 – Characteristics of the study population of 3241 men in the screening arm of the Finnish Prostate Cancer Screening Trial (screening trial cohort) and 1983 prostate cancer cases diagnosed outside screening protocols and treated at Tampere University Hospital (hospital cohort) Screening trial cohort Men (n) Median follow-up after randomisation, yr (range) Median follow-up after PCa diagnosis, yr (range) Median body mass index, kg/m2 (range) a PCa cases, n (% of the cohort) Median PSA in the first screening round (ng/ml) Median PSA at diagnosis (ng/ml) Tumour Gleason grade b 6, n (% of cases) 7–10, n (% of cases) Unknown Tumour stage Localised, n (% of cases) Metastatic, n (% of cases) c Unknown Deaths, n (% of the cohort) PCa deaths, n (% of the cohort) d Disease progression, n (% of all cases) e Primary treatment, n (% of cases) Prostatectomy Radiation therapy (external beam or brachytherapy) Endocrine therapy Active surveillance or watchful waiting Unknown
Hospital cohort
3241 16 (0.6–17)
1983 NA
8.4 (0.3–16.5)
9.1 (0–36.1)
27 (16–50)
NA
801 (24.7) 0.65
1,983 (100) NA
5.36
10.6
523 (65.3) 274 (34.2) 4 (0.5)
865 (43.6) 820 (41.4) 298 (15.0)
687 (85.8) 16 (2.0) 99 (12.4) 431 (13.3) 44 (1.4) NA
1,704 (85.9) 183 (9.2) 96 (4.8) 971 (48.7) 284 (14.2) 483 (24.4)
311 (38.8) 270 (33.7)
407 (20.5) 230 (11.6)
82 (10.2) 138 (17.2)
683 (34.4) 98 (4.9)
0
565 (28.5)
PCa = prostate cancer; PSA = prostate-specific antigen; NA = not applicable. Available for 2378 men (73.4% of the cohort). Gleason score from prostate biopsy sample. c All M1 cases, as judged by clinical M stage or bone scan when available, regardless of the T or N stage. d Prostate cancer (ICD10 code C61) as the primary cause of death. e Either PSA progression (defined as two consecutive increases in PSA after reaching the nadir or radiographic disease progression in patients managed with endocrine treatment; PSA >0.2 ng/ml in patients treated with radical prostatectomy; or a rising PSA level >2 ng/ml above the nadir in radiationtreated men). An additional 140 men had disease progression but were excluded because of missing data on PSA at the time of progression. a
b
PGM2 rs2995948 predicted overall PCa risk in the screening cohort and metastatic disease in the hospital cohort. Finally, HK2 rs965078 was associated with a higher risk of metastatic disease in the hospital cohort (Table 3). PDK1 C2_POS173159592 differed in effect from the other PDK1 SNPs. C2_POS173159592 was positively associated with overall and metastatic PCa risk in the screening cohort, with a suggestion of increase in high-grade PCa risk (p = 0.065) in the hospital cohort (Table 3). Other PDK1 SNPs were associated with lower overall PCa risk, and most were also linked to lower risk of high-grade disease. 3.5.
PCa prognosis
The genetic risk score correlated with increased risk of PCa death in both study cohorts after adjustment for disease
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 4
EUROPEAN UROLOGY XXX (2015) XXX–XXX
Table 2 – Adjusted mean PSA at baseline and from three screening rounds stratified by carrier status for SNPs of genes involved in the energy metabolic pathway among 3241 men participating in the screening arm of the Finnish Prostate Cancer Screening Trial between 1996 and 2009 Carriers (n)
Genetic risk score PCa risk Median or below Above the median p value for trend PCa mortality Median or below Above the median p value for trend SNP PDK1 C2_POS173159592 Wild type One minor allele Two minor alleles All mutation carriers p value for trend PDK1 C2_POS173156629 Wild type One minor allele Two minor alleles All mutation carriers p value for trend GBE1 rs10440071 Wild type One minor allele Two minor alleles All mutation carriers p value for trend GPI rs2910411 Wild type One minor allele Two minor alleles All mutation carriers p value for trend HK2 rs13360277 Wild type One minor allele Two minor alleles All mutation carriers p value for trend
Baseline Mean PSAAA
Mean PSAMVA
1631 1610
0.97 1.07 0.027
0.97 1.07 0.030
1633 1608
0.99 1.05 0.209
2354 1213 152 1365
Combined
Risk of screening positivity OR (95% CI) b
Mean PSAAA
Mean PSAMVA
Risk of screening positivity OR (95% CI)
Reference 1.34 (1.06–1.71)
0.99 1.07 <0.001
1.03 1.08 <0.001
Reference 1.12 (1.10–1.14)
0.99 1.05 0.193
Reference 1.15 (0.90–1.46)
1.01 1.06 0.051
1.01 1.06 0.039
Reference 1.12 (0.98–1.29)
1.03 1.07 1.34 * 1.09 * 0.032
1 1.03 1.3 * 1.06 * 0.032
Reference 1.15 (0.90–1.47)
0.99 1.03 1.19 * 1.05 * 0.003
0.99 1.03 1.19 * 1.05 * 0.003
Reference 1.17 (1.01–1.35)
3359 206 3 209
1.07 0.86 0.75 0.86 0.02
1.03 0.83 * 0.77 0.83 * 0.019
Reference 0.71 (0.39–1.28)
1.08 0.86 0.63 0.86 <0.001
1.06 0.84 0.66 0.83 <0.001
Reference 0.52 (0.35–0.77)
2604 1017 98 1115
1.03 1.14 0.85 1.11 0.317
1.00 1.10 0.82 1.07 0.339
Reference 1.38 (1.07–1.77)
1.03 1.14 * 0.92 1.11 * 0.037
0.99 1.08 * 0.87 1.06 * 0.046
Reference 1.19 (1.03–1.38)
3261 438 20 458
1.05 1.06 1.06 1.06 0.863
1.02 1.03 1.01 1.03 0.896
Reference 0.93 (0.65–1.35)
0.99 1.11 * 1.44 1.12 * <0.001
0.99 1.1 * 1.43 1.12 * <0.001
Reference 1.31 (1.08–1.59)
3482 235 2 237
1.04 1.32 * 0.45 1.32 * 0.004
1.00 1.28 * 0.35 1.27 * 0.005
Reference 1.55 (1.00–2.41)
1.05 1.23 * 0.40 1.23 * 0.001
1.03 1.21 * 0.28 1.20 * 0.001
Reference 1.31 (1.02–1.69)
a
c
*
*
PSA = prostate-specific antigen; SNP = single-nucleotide polymorphism; AA = age-adjusted; MVA = multivariate-adjusted. p < 0.05 for difference compared to wild-type carriers. a Men with serum PSA 4 ng/ml and men with serum PSA >3 ng/ml but <4 ng/ml and s free/total PSA ratio <16% were considered screening-positive. b Age-adjusted estimate calculated via logistic regression. c Similar decreasing associations with PSA were observed for PDK1 C2_POS173170045, C2_POS173170839, C2_POS173177578, rs109305, rs8179584, rs16860693, rs1530864, rs2357637, rs836628, rs2290563, rs12476089, rs12693006, and rs6433368. *
stage, grade, and PSA at diagnosis (Table 4). By contrast, no correlation with the risk of disease progression was observed (Table 4). Of the SNPs associated with increased PCa risk, PDK1 C2_POS173159592 had an association with PCa mortality that was of borderline significance in the screening trial cohort (HR 1.45, 95% CI 0.97–2.16; Table 4). The other PDK1 SNPs were associated with an increased risk of disease progression and a suggestion of an increase in risk of death (C2_POS173177578, C2_POS173170839, rs10930561, rs1530864, rs2357637, and rs12693006) or no difference in
survival (rs12693006 and rs2290563). Two minor alleles of PGM1 rs855314 were linked to lower PCa survival in the hospital cohort (HR 1.95, 95% CI 1.03–3.69); a similar but nonsignificant increase in risk was observed in the screening trial cohort (Table 4). 3.6.
ROC and Harrell’s c-index analysis
When PCa cases from both study cohorts were combined for ROC analysis, the genetic risk score in conjunction with PSA and age at diagnosis did not improve the prediction of
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 5
EUROPEAN UROLOGY XXX (2015) XXX–XXX
Table 3 – PCa risk among carriers of SNPs of genes involved in the energy metabolic pathway for 3241 men in the screening arm of the Finnish Prostate Cancer Screening Trial (screening trial cohort) and 1983 prostate cancer cases detected outside screening protocols and treated at the Tampere University Hospital (hospital cohort) Overall PCa risk
Genetic risk score Median or below Above the median p value for trend c SNP PDK1 C2_POS173159592 Wild type One minor allele Two minor alleles All mutation carriers p value for trend PGM1 rs855314 Wild type One minor allele Two minor alleles All mutation carriers p value for trend PGM2 rs2995948 e Wild type One minor allele Two minor alleles All mutation carriers p value for trend HK2 rs965078 Wild type One minor allele Two minor alleles All mutation carriers p value for trend
High-grade PCa
Screening trial cohort
Screening trial cohort
Cases (n)
Cases (n)
323 478
HR (95% CI)
a
HR (95% CI)
a
Metastatic PCa
Hospital cohort Cases (n)
OR (95% CI)
b
Screening trial cohort Cases (n)
HR (95% CI)
a
Hospital cohort Cases (n)
OR (95% CI)
b
Reference 1.57 (1.37–1.81) <0.001
110 164
Reference 1.66 (1.30–2.11) < 0.001
402 418
Reference 1.05 (0.86–1.27) 0.651
3 13
Reference 5.19 (1.48–18.24) 0.010
95 88
Reference 0.92 (0.67–1.25) 0.740
489 271 41 312
Reference 1.05 (0.91–1.22) 1.39 (1.01–1.91) 1.09 (0.94–1.26) 0.097
164 95 15 110
Reference 1.10 (0.85–1.41) 1.58 (0.93–2.68) 1,15 (0.90–1.46) 0.136
489 293 38 331
Reference 1.15 (0.93–1.41) 1.47 (0.89–2.42) 1.18 (0.97–1.44) 0.065
4 10 2 12
Reference 4.68 (1.47–14.93) 9.46 (1.73–51.79) 5.11 (1.65–15.85) 0.001
120 58 5 63
Reference 0.88 (0.63–1.22) 0.63 (0.25–1.59) 0.85(0.62–1.17) 0.250
601 189 11 200
Reference 1.07 (0.90–1.25) 0.77 (0.42–1.39) 1.04 (0.89–1.22) 0.845
204 66 4 70
Reference 1.11 (0.84–1.47) 0.76 (0.28–2.04) 1.09 (0.83–1.42) 0.725
471 138 15 153
Reference 1.11 (0.84–1.46) 2.81 (1.07–7.38) 1.18 (0.90–1.54) 0.098
11 5 0 5
Reference 1.54 (0.53–4.44) – 1.34 (0.46–3.87) 0.850
110 34 4 8
Reference 1.13 (0.75–1.71) 1.87 (0.62–5.60) 1.18 (0.80–1.76) 0.295
432 326 43 369
Reference 1.18 (1.02–1.36) 0.88 (0.65–1.21) 1.14 (0.99–1.31) 0.341
148 112 4 116
Reference 1.19 (0.93–1.52) 0.83 (0.48–1.43) 1.13 (0.89–1.44) 0.660
338 245 41 286
Reference 1.07 (0.85–1.36) 0.92 (0.58–1.46) 1.05 (0.84–1.32) 0.887
11 5 0 5
Reference 0.66 (0.23–1.90) – 0.56 (0.19–1.60) 0.189
68 68 12 80
Reference 1.5 (1.05–2.14) 1.46 (0.75–2.82) 1.49 (1.06–2.10) 0.040
299 364 135 499
Reference 0.93 (0.80–1.08) 1.08 (0.88–1.32) 0.96 (0.84–1.11) 0.773
109 118 46 164
Reference 0.82 (0.63–1.06) 1.04 (0.74–1.47) 0.87 (0.68–1.11) 0.741
216 294 110 404
Reference 1.12 (0.87–1.44) 1.17 (0.84–1.63) 1.13 (0.90–1.44) 0.307
7 5 4 9
Reference 0.55 (0.17–1.73) 1.35 (0.40–4.62) 0.75 (0.28–2.00) 0.889
40 72 36 108
Ref 1.39 (0.93–2.09) 2.01 (1.24–3.26) 1.55 (1.06–2.27) 0.005
d
PCa = prostate cancer; SNP = single-nucleotide polymorphism; HR = hazard ratio; CI = confidence interval; OR = odds ratio. From Cox regression analysis with model adjustment for age at baseline. From logistic regression analysis with model adjustment for age at diagnosis. c With the risk score as a continuous variable. d Inverse association with total prostate cancer risk in the screening trial cohort, but not with metastatic prostate disease, was observed for other PDK1 SNPs: C2_POS173217127, C2_POS173170839, rs10930561, rs8179584, rs16860693, rs1530864, rs2357637, rs836628, rs2290563, rs12476089, rs12693006, and rs6433368. Of these, C2_POS173170839, C2_POS173217127, C2_POS173177578, rs10930561, rs1530864, and rs2357637 were associated with a lower risk of high-grade prostate cancer. e A similar risk association was observed for PGM2 rs2911874. a
b
high-grade or metastatic PCa relative to PSA and age alone (data not shown). Addition of the genetic risk score to the established risk factors PSA, age, tumour stage, and Gleason grade in the Cox regression model slightly improved the predictive value for PCa death (c-indices of 87.9 and 87.4 for models with and without the genetic risk score, respectively; model improvement p < 0.001; Fig. 2).
The association between genetic risk and PCa death was slightly weaker among metformin users than non-users of antidiabetic drugs, but the confidence intervals overlapped (Fig. 3C). Body mass index did not significantly affect the association between the genetic risk score and PCa risk in users or non-users of antidiabetic drugs (Supplementary Table 3).
3.7.
4.
Impact of metformin and other antidiabetic drugs
Neither metformin nor non-metformin antidiabetic drugs significantly modified the association between genetic risk score and either PCa risk overall or high-grade disease, although the genetic risk score was a significant risk predictor only in men who had not used antidiabetic drugs (Fig. 3A,B).
Discussion
The combined genetic risk score based on risk SNPs of glucose and energy metabolic genes predicted PCa risk, disease stage and grade, and prognosis in men under systematic PCa screening. Our results suggest a role of energy metabolism in the development and progression of PCa. The risk score predicted PCa death in both study
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 6
EUROPEAN UROLOGY XXX (2015) XXX–XXX
(A)
0.4 Age-adjusted HR 1.57; 95% Cl 1.37–1.81
Cumulative hazard for prostate cancer
p < 0.001
Genetic risk score above the median
0.3
Genetic risk score at or below the median
0.2
0.1
0.0 0.00
5.00
10.00
15.00
20.00
Time since FinPCST randomisation (yr)
Cumulative hazard for Gleason 7–10 prostate cancer
(B)
0.12
Age-adjusted HR 1.66; 95% Cl 1.30–2.11 Genetic risk score above the median
p < 0.001 0.10
0.08 Genetic risk score at or below the median
0.06
0.04
0.02
0.00 0.00
5.00
10.00
15.00
20.00
Time since FinPCST randomisation (yr) Fig. 1 – Prostate cancer risk among men stratified by a genetic risk score based on single-nucleotide polymorphisms of genes involved in glucose and energy metabolic pathways. Cohort of 3241 participants from the screening arm of the Finnish Prostate Cancer Screening Trial (FinPCST) between 1995 and 2009. The p value for trends were calculated by adding the genetic risk score as a continuous variable to the age-adjusted Cox regression model. (A) Overall risk of prostate cancer. (B) Risk of Gleason 7–10 prostate cancer. HR = hazard ratio; CI = confidence interval.
cohorts, but not tumour grade or stage in the hospital cohort. The difference can probably be attributed to differences in disease characteristics between the two cohorts: the role of genetic risk factors may be clearer in a screened population with a high proportion of screen-detected low-grade or earlystage cases. Systematic screening and the resulting stage migration of PCa cases at detection could probably modify the role of genetic risk factors. Because the risk prediction ability of the genetic risk score may vary with environmental
factors, our findings needs to be validated in additional study populations [13]. PDK and PGM1 SNPs exhibited consistent associations with both PCa risk and survival rates in both study cohorts. These genes in particular may play a role in PCa initiation and progression. PDK1 (pyruvate dehydrogenase kinase 1) inhibits the activity of PDH (pyruvate dehydrogenase), a mitochondrial gatekeeping enzyme linking cytosolic glycolysis to the
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 7
EUROPEAN UROLOGY XXX (2015) XXX–XXX
Table 4 – PCa progression and mortality among carriers of SNPs of genes involved in the energy metabolic pathway for 3241 men in the screening arm of the Finnish Prostate Cancer Screening Trial (screening trial cohort) and 1983 prostate cancer cases diagnosed outside screening protocols and treated at the Tampere University Hospital (the hospital cohort) Disease progression Cases (n)
HR (95% CI)
PCa mortality Deaths (n)
a
HR (95% CI)
PCa survival, PCa cases only Screening trial cases
a
Deaths (n) Genetic risk score Median or below Above the median p value for trend SNP PDK1 C2_POS173159592 Wild type One minor allele Two minor alleles All mutation carriers p value for trend PGM1 rs855314 Wild type One minor allele Two minor alleles All mutation carriers p value for trend
HR (95% CI)
Hospital cohort cases
a
Deaths (n)
HR (95% CI)
a
All cases combined
Deaths (n)
HR (95% CI)
a
244 239
Reference 1.00 (0.83–1.19) 0.97
16 28
Reference 1.45 (0.97–2.17) 0.015
16 28
Reference 1.71 (0.92–3.17) 0.014
121 163
Reference 1.40 (1.10–1.77) 0.001
137 191
Reference 1.51 (1.15–1.98) <0.001
303 163 17 180
Reference 0.95 (0.78–1.15) 0.90 (0.55–1.46) 0.95 (0.79–1.14) 0.52
23 17 4 21
Reference 1.44 (0.95–2.19) 1.48 (0.64–3.45) 1.45 (0.97–2.16) 0.09
23 17 4 21
Reference 0.92 (0.48–1.76) 1.52 (0.52–4.47) 0.99 (0.53–1.84) 0.74
191 84 9 93
Reference 0.82 (0.63–1.06) 0.78 (0.40–1.52) 0.82 (0.63–1.05) 0.12
214 101 13 114
Reference 0.80 (0.60–1.07) 0.85 (0.42–1.73) 0.81 (0.61–1.06) 0.84
366 109 8 117
Reference 0.98 (0.79–1.22) 0.93 (0.46–1.88) 0.98 (0.79–1.21) 0.81
32 11 1 12
Reference 1.45 (0.94–2.25) 1.49 (0.37–6.11) 1.46 (0.95–2.23) 0.09
32 11 1 12
Reference 1.11 (0.56–2.20) 1.72 (0.23–1.79) 1.14 (0.59–2.22) 0.63
216 58 10 68
Reference 1.02 (0.77–1.37) 1.95 (1.03–3.69) 1.10 (0.84–1.45) 0.23
248 69 11 80
Reference 1.07 (0.78–1.47) 2.10 (0.99–4.48) 1.14 (0.84–1.55) 0.20
PCa = prostate cancer; HR = hazard ratio; CI = confidence interval. From Cox regression analysis with model adjustment for age and PSA at diagnosis, and prostate cancer stage and grade.
a
0.05
tricarboxylic acid (TCA) cycle in mitochondria [14]. PDH is essential for oncogene-induced senescence, a cellular defence mechanism against malignant transformation; PDK1 inhibits this enzyme and is essential for cancer cell growth [14]. PDK1 also has a role in adaptation to a hypoxic cancer environment, and is directly activated by HIF (hypoxia-inducible factor) [1]. Inactivation of PDH via increased PDK1 activity reduces shuttling of pyruvate into
None All
0.03
Harrell’s c-index for Age, Gleason grade, M stage, and PSA: 87.4
0.02
Age, Gleason grade, M stage, PSA, and genec risk score: 87.9 Model improvement: p < 0.001
0.00
0.01
Net benefit
0.04
Age at dg, tumour Gleason grade, M stage and PSA at dg Age at dg, tumour Gleason grade, M stage, PSA at dg and the genetic risk score
0.0
0.2
0.4
0.6
0.8
Threshold probability Fig. 2 – Decision curve analysis for prostate cancer survival after diagnosis using the genetic risk score and known prognostic factors among prostate cancer cases from the combined study cohorts. dg = diagnosis; PSA = prostate-specific antigen.
mitochondria, facilitating its conversion to lactate in the cytosol (the Warburg effect). Because less pyruvate is shuttled into mitochondria, the amount of intermediates produced by the TCA cycle decreases. If further studies confirm that the PDK1 SNPs tested in our study affect enzyme activity, our results are consistent with this phenomenon occurring in PCa. Known cessation of secretion of citrate (a TCA intermediate) from the prostate epithelium during carcinogenesis supports a change in the role of the TCA cycle [15]. PGM1 SNPs were associated with increased PCa risk and poorer survival. This is the first enzyme in glycogenesis. A recent study showed that, like PDK1, PGM1 expression is increased by HIF under hypoxic conditions [16]. Glycogen storage is a survival mechanism of tumour cells for withstanding glucose and oxygen starvation; therefore, it likely has importance for metastatic progression. PDK1 C2_POS173159592 exhibited associations with PSA and PCa risk that are opposite to those observed for other PDK1 SNPs. This suggests that the minor allele for C2_POS173159592 is inherited with the major allele for the rest of the gene. We plan to further investigate this issue using haplotype analysis in future studies. Nevertheless, all PDK1 SNPs exhibited at least a suggestive association with disease outcome, pointing to a role of PDK1 in PCa progression. SNPs of several other genes encoding enzymes that regulate the conversion of glucose to pyruvate (HK2, GPI, and PCK1) and of pyruvate within mitochondria (PC, DLD, and SDHD) were associated with either PCa mortality or risk, but without consistency across cohorts.
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 8
EUROPEAN UROLOGY XXX (2015) XXX–XXX
(A)
2.5
HR (95%Cl)
2 1.5 1 0.5
Interacon with meormin use: p = 0.21
0 Non-users
HR (95% Cl)
(B)
Non-meormin an-DM drug users
Meormin users
2.10 (0.89–4.99)
1.29 (0.92–1.81)
1.63 (1.39–1.91)
Meormin amount in upper 50% 1.42 (0.87–2.33)
2.5
HR (95% Cl)
2 1.5 1 0.5 Interacon with meormin use: p = 0.094 0
HR (95%Cl)
(C)
Non-users
Non-meormin an-DM drug users
Meormin users
Meormin amount in upper 50%
1.85 (1.40–2.43)
1.22 (0.36–4.11)
1.10 (0.62–1.96)
1.14 (0.56–2.33)
3
Interacon with meormin use: p = 0.171
HR (95% Cl)
2.5 2 1.5 1
caused by direct antineoplastic medication effects rather than the underlying diabetes. The strength of our study lies in the ability to evaluate associations between SNPs and multiple PCa outcomes and to validate the findings with two distinct study cohorts. We were also able to evaluate the impact of PCa screening on the gene-risk association and to combine registry-based information on antidiabetic medication use with genotype data. The registry records medication purchases independently of PCa status. Therefore, the results are not affected by recall bias. Our study also has limitations. Diabetes was assessed in terms of antidiabetic drug usage; we did not have information on diabetes managed via diet. We also had no information on glucose control. This could have led to exposure misclassification and subsequent underestimation of the effects of antidiabetic drugs. Furthermore, we had only limited data on body mass index and cardiovascular disease, and no information on lifestyle risk factors that may have differed between diabetic and nondiabetic men, possibly affecting PCa risk or outcome. However, it is likely that such factors are not related to genotype, and thus are unlikely to bias our estimation of SNP associations. Our median follow-up after PCa diagnosis was 8–9 yr. Therefore, we were not able to reliably estimate very long-term effects of genetic risk factors on PCa survival. Our analyses of the effects of metformin on PCa were limited by the small sample size (only 133 medication users among PCa cases), so our confidence intervals were wide. Finally, our study is limited by its retrospective design, with the possibility of residual confounding by unmeasured variables.
0.5
5.
Conclusions
0
HR (95% Cl)
Non-users
Meormin users
2.40 (1.13–5.09)
0.15 (0.01–2.86)
Meormin use in upper 50% 1.00 (0.13–7.62)
Fig. 3 – Effect of antidiabetic drug use on the association between prostate cancer risk and survival and the genetic risk score calculated based on single-nucleotide polymorphisms of genes involved in the glucose and energy metabolic pathways. Cohort of 3241 participants in the screening arm of the Finnish Prostate Cancer Screening Trial between 1995 and 2009. Men with a genetic risk score above the median compared to those with a risk score below the median. (A) Overall prostate cancer risk. Hazard ratio (HR) from Cox regression with model adjustment for age. (B) Risk of high-grade (Gleason 7–10) prostate cancer. HR from Cox regression with model adjustment for age. (C) Risk of prostate cancer death. HR from Cox regression with model adjustment for age, PSA at diagnosis, tumour Gleason grade, and M stage. CI = confidence interval; DM = diabetes mellitus.
We demonstrated that genetic risk in terms of SNPs of genes involved in cellular glucose and energy metabolism is associated with PCa survival. These genes probably participate in metabolic rewiring of cancer cells to enhance disease progression. This points to the possibility of new avenues for prevention of PCa death. The clinical utility of these SNPs in PCa management should be investigated in validation studies in other populations. Author contributions: Teemu J. Murtola had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Murtola, Wahlfors, Schleutker, Auvinen. Acquisition of data: All authors. Analysis and interpretation of data: Murtola, Wahlfors, Schleutker,
Lower PCa mortality has been reported for metformin users [6]. In our analysis, the link between PCa survival and genetic risk was only slightly weaker among metformin users than among non-users and among users of other antidiabetic drugs. Thus, our results do not strongly support the previous results, although the associations did tend to change in the same direction. However, it remains unclear whether the associations found in epidemiologic studies are
Auvinen. Drafting of the manuscript: All authors. Critical revision of the manuscript for important intellectual content: All authors. Statistical analysis: Murtola, Wahlfors, Auvinen. Obtaining funding: Tammela, Auvinen. Administrative, technical, or material support: Taari, Tammela. Supervision: Schleutker, Auvinen. Other: None.
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026
EURURO-6132; No. of Pages 9 EUROPEAN UROLOGY XXX (2015) XXX–XXX
9
Financial disclosures: Teemu J. Murtola certifies that all conflicts of
[4] Abdollah F, Briganti A, Suardi N, et al. Does diabetes mellitus
interest, including specific financial interests and relationships and
increase the risk of high-grade prostate cancer in patients under-
affiliations relevant to the subject matter or materials discussed in the
going radical prostatectomy? Prostate Cancer Prostatic Dis 2011;
manuscript (eg, employment/affiliation, grants or funding, consultan-
14:74–8.
cies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Teemu J.
[5] Brown KA, Samarajeewa NU, Simpson ER. Endocrine-related cancers and the role of AMPK. Mol Cell Endocrinol 2013;366:170–9.
Murtola has received lecture fees from Janssen-Cilag and has received
[6] Spratt DE, Zhang C, Zumsteg ZS, Pei X, Zhang Z, Zelefsky MJ.
financial support for participation in scientific congresses from Ferring
Metformin and prostate cancer: reduced development of castration-
and Janssen-Cilag. Teuvo L.J. Tammela has received consultancy fees
resistant disease and prostate cancer mortality. Eur Urol 2013;
from Astellas, GlaxoSmithKline, Pfizer, Orion Pharma, and Amgen. Kimmo Taari has received consultancy fees from GlaxoSmithKline,
63:709–16. [7] Wang L, McDonnell SK, Hebbring SJ, et al. Polymorphisms in
Pfizer, Orion Pharma, and Amgen. The remaining authors have nothing to
mitochondrial genes and prostate cancer risk. Cancer Epidemiol
disclose.
Biomarkers Prev 2008;17:3558–66. [8] Tsilidis KK, Travis RC, Appleby PN, et al. Insulin-like growth factor
Funding/Support and role of the sponsor: This work was supported by the European Commission’s Seventh Framework Programme (HEALTH-F22009-223175), Cancer Research UK (C5047/A7357, C1287/A10118, C5047/A3354, C5047/A10692, C16913/A6135), and the National Institutes of Health (Cancer Post-Cancer GWAS initiative grant 1 U19 CA 148537-01). The sponsors played a role in data collection. Acknowledgments: We are grateful to Dr. Elizabeth Platz and Dr. William
pathway genes and blood concentrations, dietary protein and risk of prostate cancer in the NCI Breast and Prostate Cancer Cohort Consortium (BPC3). Int J Cancer 2013;133:495–504. [9] Eeles RA, Olama AA, Benlloch S, et al. Identification of 23 new prostate cancer susceptibility loci using the iCOGS custom genotyping array. Nat Genet 2013;45:385–91. [10] Kilpela¨inen TP, Tammela TL, Malila N, et al. Prostate cancer mor-
B. Isaacs, Johns Hopkins University, for helpful comments and advice
tality in the Finnish randomized screening trial. J Natl Cancer Inst
during preparation of the manuscript.
2013;105:719–25. [11] Martikainen J, Rajaniemi S. Drug reimbursement systems in EU member states, Iceland and Norway. Social Security and Health Report 54. Helsinki: Social Insurance Institution of Finland; 2002.
Appendix A. Supplementary data
http://www.kela.fi/in/internet/liite.nsf/ABID/030303101726PN/ $File/Drug_reimbursement.pdf [12] Pashayan N, Duffy SW, Neal DE, et al. Implications of polygenic risk-
Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j. eururo.2015.03.026.
[13] Vickers A. Prediction models in urology: are they any good, and
References
[14] Kaplon J, Zheng L, Meissl K, et al. A key role for mitochondrial
stratified screening for prostate cancer on overdiagnosis. Genet Med. In press. http://dx.doi.org/10.1038/gim.2014.192 how would we know anyway? Eur Urol 2010;57:571–3. gatekeeper pyruvate dehydrogenase in oncogene-induced senes-
[1] Warburg O. On the origin of cancer cells. Science 1956;123:309–14.
cence. Nature 2013;498:109–12.
[2] Lindberg J, Mills IG, Klevebring D, et al. The mitochondrial and
[15] Zadra G, Photopoulos C, Loda M. The fat side of prostate cancer.
autosomal mutation landscapes of prostate cancer. Eur Urol 2013;
Biochim Biophys Acta 2013;1831:1518–32. [16] Pelletier J, Bellot G, Gounon P, Lacas-Gervais S, Pouysse´gur J,
63:702–8. [3] Xu H, Jiang HW, Ding GX, Zhang H, Mao SH, Ding Q. Diabetes
Mazure NM. Glycogen synthesis is induced in hypoxia by the
mellitus and prostate cancer risk of different grade or stage: a
hypoxia-inducible factor and promotes cancer cell survival. Front
systematic review and meta-analysis. Diabetes Res Clin Pract
Oncol 2012;2:18.
2013;99:241–9.
Please cite this article in press as: Murtola TJ, et al. Polymorphisms of Genes Involved in Glucose and Energy Metabolic Pathways and Prostate Cancer: Interplay with Metformin. Eur Urol (2015), http://dx.doi.org/10.1016/j.eururo.2015.03.026