|
|
Pharmacogenet Genomics. Author manuscript; available in PMC 2009 August 1. Published in final edited form as: doi: 10.1097/FPC.0b013e3282fe3326. | PMCID: PMC2613759 NIHMSID: NIHMS83776 |
Association of genetic polymorphisms in CYP19A1 and blood levels of sex hormones among postmenopausal Chinese women Hui Cai,1 Xiao Ou Shu,1 Kathleen M. Egan,2 Qiuyin Cai,1 Ji-Rong Long,1 Yu-Tang Gao,3 and Wei Zheng1 1Department of Medicine, Vanderbilt Epidemiology Center, Vanderbilt University Medical Center and Vanderbilt-Ingram Cancer Center, Nashville, TN 2 Cancer Prevention and Control, Moffitt Cancer Center, University of South Florida, Tampa, FL 3Department of Epidemiology, Shanghai Cancer Institute, Shanghai, People's Republic of China Objective Circulating estrogen levels have been related to the risk of several female cancers. Blood levels of estrogen are controlled by estrogen synthesis enzymes. Genetic variation of estrogen genes thus may influence circulating estrogen levels. We investigated the associations of genetic polymorphisms in CYP19A1, a critical gene involved in estrogen synthesis, with plasma levels of sex hormones among postmenopausal Chinese women. Methods Included in this study were 345 postmenopausal community controls from a population-based case-control study conducted in Shanghai. Fasting blood samples from those women were measured for blood estradiol, estrone, estrone sulphate and testosterone. A total of 19 genetic polymorphisms in CYP19A1 were genotyped using ABI7900 or PCR-RFLP methods. Differences in plasma levels of hormones by genotype were examined using variance analysis. Result The geometric means of plasma levels of estradiol, estrone, estrone sulphate and testosterone were 10.1 pg/ml, 16.8 pg/ml, 969.0 pg/ml and 202.9 pg/ml, respectively, for this study population. We found that plasma levels of estrone were associated with rs28566535 (P=0.0180), rs730154 (P=0.0141) and rs936306 (P=0.0274) in block 2. In the same block, the haplotype CGCTA was related to level of estrone (P=0.0064). SNP rs1902584 in block 1 was associated with estradiol only in overweight postmenopausal women. No clear association with sex hormones was noted for the other genetic polymorphisms evaluated in the study. Conclusion SNPs in block 1 and block 2 of the CYP19A1 gene are related to plasma levels of estrogen among postmenopausal Chinese women and may therefore play an important role in the etiology of hormone-related cancers. Keywords: association, genetic polymorphism, CYP19A1, sex hormone, synthesis, BMI Endogenous sex-steroid hormones play an important role in the etiology of female cancers, including breast and endometrial cancer [ 1, 2]. Sex-steroid hormones are synthesized and eliminated through complex metabolic pathways. Polymorphisms in genes coding for key enzymes involved in these pathways could modulate endogenous hormone levels and through this mechanism, influence the risk of hormone-related cancers [ 2, 3]. In postmenopausal women, the most abundant circulating estrogens are estrone, and to a lesser extent, estradiol, both derived from the extragonadal aromatisation of androstenedione and testosterone [ 4- 6]. It is generally believed that genetic polymorphisms in hormone-related genes could explain a portion of breast cancer susceptibility through modulating the level of endogenous hormones. The effect of polymorphisms on circulating levels of hormones, however, has been evaluated only in few studies with no consistent evidence of substantial effect [ 2, 3, 7- 10]. The CYP19A1 gene codes for aromatase, which catalyzes the conversation of testosterone to estradiol and androstenedione to estrone [ 11]. In this study, we evaluated 19 polymorphisms in the CYP19A1 gene for associations with levels of several sex hormones in the blood of Chinese women who were involved as controls in a population-based, case-control study, the Shanghai Breast Cancer Study. Study subjects The current study was conducted in a subset of controls recruited as a part of the Shanghai Breast Cancer Study (SBCS). The study design and participant recruitment have been described in detail elsewhere [ 12]. Briefly, from August 1996 to March 1998 a population-based case-control study was conducted in Shanghai, China. 1724 healthy control women aged 25−64 years were enrolled in the study. They were randomly selected from the general population in Shanghai, using the resident registration information provided by the population-based Shanghai Resident Registry. An in-person interview was conducted by a trained interviewer for all eligible controls. Of those who were eligible for our study, 90.3% completed an in-person interview (refusals: n=166, 9.6%; death or prior cancer diagnosis: n=2, 0.1%). The in-person interview was conducted with the use of a structured questionnaire, which elicited information on demographic features, menstrual and reproductive history, use of sex steroid hormones, medical history, physical activity, dietary habits and family history of cancer. Of the 1556 controls, morning fasting blood samples were collected from 1310 controls (84.2%). These blood sample were processed to separate plasma within six hours of collection, and the plasma specimens were immediately stored at −70°C. Genetic marker selection and genotyping assays The genotyping assays were conducted in 2004 before the release of HapMap data. Therefore, the genetic markers of the CYP19A1 gene included in this study were selected according to a report from the Multiethnic Cohort Study [ 13]. Based on haplotype analyses of 74 densely spaced SNPs across the gene, 19 haplotype tagging SNPs (htSNPs) were identified in a Japanese population. Because of the similarity in linkage disequilibrium patterns between Chinese and Japanese populations [ 14], we used these 19 htSNPs in the present study. Genomic DNA was extracted from buffy coats using a Puregene DNA Purification Kit (Gentra Systems, Minneapolis, MN) following the manufacturer's protocol. SNP rs1004984 was genotyped using a Masscode assay at BioServe Biotechnologies, Ltd (Laurel, MD). The nonsynonymous SNP, rs700519, was genotyped using PCR-RFLP and confirmed by direct sequencing using BigDye Terminator Chemistry on an ABI 3700 (ABI, Applied Biosystems, Inc, Foster City, CA, USA). Genotyping of the other 17 SNPs was performed by running the 5’nuclease TaqMan allelic discrimination assay using an ABI 7900 (ABI). Details concerning assays, primers, probes, and procedures are available upon request. Laboratory staff were blind to the identity of the subjects. Quality control samples were included in the genotyping assays. Each 96-well plate of genomic DNA contained multiple quality controls, including one water, two samples of CEPH 1347−02, two known duplicates, and two blinded duplicates. The average agreement of the genotypes for these 19 markers determined for the blinded quality control samples was 98.7%. Sex hormone measurement Measurement of steroids in our study was performed in a reference laboratory at Diagnostic Systems Laboratories, Inc. (DSL, Webster, TX). The reference laboratory specializes in in vitro diagnostic testing of endocrine markers and is certified by the Clinical Laboratory Improvement Amendments (CLIA) and International Standard ISO 9002. Each sample was tested in duplicate. Commercial radioimmunoassays (RIA) from DSL were used for the measurement of steroids. The intra- and inter-assay precisions expressed as coefficient of variation (CV) were 3.2−5.3% and 8.1−9.3% for estradiol; 4.4−9.4% and 6.0−11.1% for estrone; 4.6−9.2% and 5.1−8.8% for estrone sulfate and 6.7−8.1% and 5.7−10.5% for testosterone, respectively. The assay's standard range and sensitivity (in brackets) were 20−6000 pg/ml (4.7 pg/ml) for estradiol; 15−2000 pg/ml (12 pg/ml) for estrone; 0.025−15 ng/ml (0.01 ng/ml) for estrone sulfate and 0.1−25 ng/ml (0.05 ng/ml) for testosterone. Statistical analysis The program PHASE, which is based on a Bayesian statistical model [ 15], was used to reconstruct haplotypes within each block of the CYP19A1 gene. The SNPs used for haplotype reconstruction were arranged according to their chromosomal locations in the order of rs2446405, rs2445765, rs2470144, rs1004984 and rs1902584 for block 1; rs28566535, rs12900137, rs730154, rs936306 and rs1902586 for block 2; rs749292, rs6493494 and rs1008805 for block 3; rs727479, rs2414096, rs700519, rs10046 and rs4646 for block 4. Because the measured values of the hormones were not normally distributed, all values of hormones were log-transformed in the statistical analysis. Variance analysis was used to evaluate the differences of levels of hormones among genotypes of CYP19A1 with adjustment for age, age at menarche, BMI and batch (because hormone fractions were measured in two laboratory batches). Adjusted means of sex hormones and their confidence intervals were calculated. All statistical analyses were performed using SAS version 9.1 (SAS Institute, Cary, NC) and all tests were based on two-tailed probability. There were 438 postmenopausal control women in the case-control study. Among them 93 women who had either used estrogen replacement therapy or smoked or consumed alcohol regularly or had non-natural menopause were excluded. The remaining 345 women were included in this study. The geometric means of plasma levels of estradiol, estrone, estrone sulfate and testosterone were 10.1 pg/ml, 16.8 pg/ml, 969.0 pg/ml and 202.9 pg/ml, respectively, for the study population. Table 1 presents characteristics of the study subjects. All of the women were between 39 and 64 years of age (57.1±4.8). Their average age at menopause was 48.3 years (48.3±3.9). Most were married (86.4%). The mean BMI and WHR were 24.2 and 0.82, respectively. Genotype and allele distributions for the 19 SNPs are summarized in Table 2. With the exception of rs2445765 and rs12907866, all other SNPs were consistent with Hardy-Weinberg equilibrium (HWE) distribution (P>0.05). | Table1Characteristics of 345 study participants |
| Table 2Information for the studied markers in the CYP19A1 gene in Chinese postmenopausal women |
Table 3 presents the relationship between levels of sex hormones (estradiol and estrone), estradiol:testosterone ratio and SNPs among four blocks in the CYP19A1 gene. Only genotypes in block 2 were associated with estrone levels. The mean estrone concentration was 15.7 pg/ml among women with the rs28566535 AA or rs730154 TT genotype, which was significantly lower (P=0.0180 and P=0.0141) than that of women with genotype AC/CC or TC/CC at these loci (means were 17.8 and 18.0 pg/ml, respectively). Also, the geometric means of estrone among women with rs936306 CC was 15.6 pg/ml, which was significantly lower (P=0.0274) than that of women with genotype CT/TT (geometric mean was 17.6 pg/ml). There were no significant associations between levels of estradiol, estradiol:testosterone ratio, estrone sulphate and testosterone and other SNPs (data not shown). | Table 3Geometric means and 95%CI of estrone and estradiol by SNPs1 |
Haplotype analyses were conducted to evaluate the combined effect of the SNPs located within each haplotype block. Because the genotype distributions of rs2445765 and rs12907866 deviated from HWE ( Table 2), they were eliminated from the haplotype reconstruction. As shown in Table 4, an association between haplotype and level of estrone was observed for the CGCTA haplotype (minor alleles for rs28566535, rs730154, rs936306, and rs1902586 in block 2) with statistical significance (P=0.0064); women with one or two copies of this haplotype had an 18% higher level of estrone. Estrone sulfate was 22% lower among women having one or two copies of another haplotype composed of minor alleles of 5 SNPs CCCTA. Haplotypes in the other three blocks were, in general, not related to blood levels of estrogen (data not shown). | Table 4CYP19A1 gene haplotypes in association with hormones1,2 |
As aromatase expression in adipose tissue occurs primarily via promoter I.4 in block 1, we explored relationships of SNPs in block 1 with estrogen levels in overweight women (BMI>=25kg/m2), who have higher levels of adiposity. We found that overweight women with the rs1902584 AT/TT genotype had 34% higher levels of estradiol, compared to women with the AA genotype (P=0.0312). We also found a significant interaction between rs1902584 and BMI (P=0.0201) after adjustment for age, age at menarche and batch. presents the ‘dose-response’ relationship between estradiol concentrations and BMI by rs1902584 genotype. Estradiol levels did not change materially with increasing BMI in women with the AA genotype. In contrast, estradiol levels increased markedly with increasing BMI in women with the AT/TT genotype. The peak in estradiol levels in this group was about 15 pg/ml, which occurred at a BMI of approximately 27 kg/m2. The association between estradiol and SNPs in block 1 was confirmed by the TCAT haplotype; overweight women with the TCAT haplotype had a higher level of estradiol (geometric mean=14.1 pg/ml) when compared to women with a non-TCAT haplotype (geometric mean=10.5, P=0.0411) (data not shown in the table). The association of estrone and CGCTA in block 2 remained (P=0.0358) in overweight women. We did not find relationships in the remaining SNPs with estradiol levels among overweight women nor did we observe relationships between SNPs in block 1 and estrogen levels in non-obese women. This is the first study to comprehensively evaluate the associations of 19 tagging SNPs in the CYP19A1 gene with levels of sex hormones among postmenopausal Chinese women. We found that rs1902584 located in haplotype block 1 was associated with estradiol levels among overweight postmenopausal women, whereas polymorphisms in haplotype block 2 were associated with estrone levels among all postmenopausal women. It is widely accepted that estrogens are involved in the development of breast cancer, and longer lifetime exposure to endogenous estrogen or to high levels of estrogen are known to increase breast cancer risk [ 16]. Among estrogen biosynthesis genes, CYP19A1 is the most important [ 17]. The CYP19A1 gene, mapped to chromosome 15q21.1, spans about 123 kb with translated exons II-X (about 30 kb) and a 93 kb regulatory region (exon I). The regulatory region contains at least 190 distinct promoters that regulate in a tissue- or signal pathway- specific manner [ 18]. Among these, promoters I.3, I.4 and II are expressed in adipose tissue and may also participate in the pathogenesis of malignant breast transformation, since different promoters in exon I were found to control aromatase expression in normal breast adipocytes (exon I.4) and during malignant breast transformation (exon I.3 and promoter II) [ 19]. Downstream of the transcriptional start site for promoter I.4, there is a specificity protein-1 (Sp1) binding site; upstream of the transcriptional start site of promoter I.4 there is a gamma interferon activating sequence (GAS) element, downstream of which is a glucocorticoid response element (GRE) [ 20]. All of these three elements are required for expression from promoter I.4. The SNPs located in haplotype block 1 are very close to promoter I.4. Among the five SNPs of haplotype block 1 in our study, rs1902584 has the shortest distance to the three response elements, 493 bp upstream of GAS, 646 bp upstream of GRE and 915 bp upstream of Sp1 [ 20]. Therefore this SNP may have an effect on regulating gene transcription and may affect the activity of aromatase in adipose tissue. We have previously reported a positive association between this genetic polymorphism (rs1902584) and obesity among Chinese women [ 21]. In postmenopausal women estrogen is produced primarily through the conversion of androgens to estrogens catalyzed by aromatase in the adipose tissue [ 22, 23], and the genetic effect of CYP19A1, thus, may be more evident among postmenopausal women with high BMI. Indeed, we found interactions between haplotype TCAT in block 1 and BMI. Overweight women with the TCAT haplotype had 34% higher levels of estradiol compared to women without the TCAT haplotype. We further observed that in this block only rs1902584 AA was related to low levels of estradiol among overweight postmenopausal women. This may indicate that rs1902584 AA is the main predictor of low estrogen levels in this population. We observed associations between each SNP in haplotype block 2 and levels of estrone and these associations were confirmed by haplotype analysis. Women with the minor allele at four of the five SNPs tended to have higher levels of estrone and estrone sulphate, compared to those with the major alleles. These associations may be related to the function of promoter I.7 in this block. Promoter I.7, recently cloned by Sebastian et al [ 24], is located approximately 36 kb upstream of the coding region. It is unique in that it is a GATA-2-regulated endothelial promoter of the human P450arom and may increase estrogen biosynthesis in the vascular endothelial cells of breast cancer [ 24]. The activity of this promoter may also be important for the intracrine and paracrine effects of estrogen on blood vessel physiology [ 18]. In the Multiethnic Cohort Study [ 13] a significant association was observed between our risk haplotype, CGCTA, and breast cancer with an odds ratio (95% CI) of 1.2 (1.1−1.4) in the entire study population. Interestingly, the association was strongest among the Japanese women (OR:1.4; 95% CI 1.1−1.8), a group genetically more similar to the Chinese women included in the present analysis. Our findings thus support the possibility that this promoter may have a potential role in breast cancer pathogenesis [ 25, 26]. In line with these results, we have previously reported that patients homozygous for the minor alleles in the five SNPs located in haplotype block 2 had a lower 5-year disease-free survival rate compared with those carrying the major allele [ 27]. In a study among 1747 postmenopausal women that examined two polymorphisms in the CYP19A1 gene, Dunning et al. found rs10046 to be a modest predictor of estradiol and estrone levels (about 6% increase per allele, P<0.01) [ 28]. In a more comprehensive evaluation of common variations spanning the entire CYP19A1 locus, Haiman et al. reported that the two-SNP haplotype rs749292-rs727479 (A-A) in CYP19A1 is significantly associated with about a 15% increase in estrogen levels in postmenopausal women (P<0.01) [ 29]. These SNPs, however, were not associated with estrogen levels in our study. The reasons for the inconsistencies in these studies remain unclear. It is possible that the association of these SNPs with estrogen levels may be mediated through their LD with true functional SNPs, and the degree of LD of these SNPs may differ in these study populations. Our study was based on an ethnically homogenous population among naturally postmenopausal women with no history of smoking, alcohol consumption or use of hormone replacement therapy. The results provide evidence that the variation of SNPs in haplotype block 1 and block 2 of the CYP19A1 gene may be functionally relevant to circulating hormone concentrations. Selection bias was minimized given the population-based study design and high response rate. A potential concern for the study is the relatively small sample size, which may have reduced our power to detect minor variations in hormone levels determined by CYP19A1 variants, especially in multivariate analysis and stratification. In summary, our study suggests that SNPs in haplotype block 2 of the CYP19A1 gene are associated with higher plasma levels of estrone in postmenopausal women, and rs1902584 may be associated with estradiol levels among overweight or obese postmenopausal women, contributing potentially to their greater breast cancer risk.
| Table 5Geometric means and 95%CI of sex hormones by SNPs in Block1 for women with BMI>=25.01 |
The authors thank Drs Qi Dai and Fan Jin and Ms Jia-Rong Cheng for their contributions in coordinating data and specimen collection in Shanghai, Ms Qing Wang and Ms Regina Courtney for technical assistance in genotyping assays, and Ms Bethanie Hull for technical assistance with the preparation of this manuscript. This study was supported by National Cancer Institute grants USPHS R01CA64277 and R01CA90899. 1. Henderson BE, Feigelson HS. Hormonal carcinogenesis. Carcinogenesis. 2000;21:427–33. [PubMed] 2. Mitrunen K, Hirvonen A. Molecular epidemiology of sporadic breast cancer – The role of polymorphic genes involved in oestrogen biosynthesis and metabolism. Mutat Res. 2003;544:9–41. [PubMed] 3. Huber JC, Schneeberger C, Tempfer CB. Genetic modeling of estrogen metabolism as a risk factor of hormone-dependent disorders. Maturitas. 2002;41(suppl1):S55–S64. [PubMed] 4. Neven P. The origin of postmenopausal oestrogens. European Journal of Cancer. 2002;38(Supplement 6):S29–S30. [PubMed] 5. Speroff L. Hormone biosynthsis, metabolism and mechanism of action. In: Speroff L, Glass RH, Kase NG, editors. Clinical Gynecologic Endocrinology and Infertility. Lippincott Williams & Wilkins; New York: 1999. pp. 31–105. 6. Simpson ER, Rubin G, Clyne C, Robertson K, O'Donnell L, Jones M, et al. The role of local estrogen biosynthesis in male and female. Trends Endocrinol Metab. 2000;11:184–8. [PubMed] 7. Ginsburg ES. Estrogen, alcohol and breast cancer risk. J. Steroid Biochem Mol Biol. 1999;69:299–306. [PubMed] 8. Hartge P. Genes, hormones, and pathways to breast cancer. N Engl J Med. 2003;348:2352–4. [PubMed] 9. Coughlin SS, Piper M. Genetic polymorphisms and risk of breast cancer. Cancer Epidemiol Biomarkers Prev. 1999;8:1023–32. [PubMed] 10. Kristensen VN, Borresen-Dale AL. Molecular epidemiology of breast cancer: genetic variation in steroid hormone metabolism. Mutat Res. 2000;462:323–33. [PubMed] 11. Mendelson CR, Means GD, Mahendroo ML, Corbin CJ, Steinkampf MP, Graham-Lorence S, et al. Use of molecular probes to study regulation of aromatase cytochrome P450. Biol Report. 1990;42:1–10. 12. Gao YT, Shu XO, Dai Q, Potter JD, Brinton LA, Wen WQ, et al. Association of menstrual and reproductive factors with breast cancer risk: results from the Shanghai Breast Cancer Study. Int. J. Cancer. 2000;87:295–300. [PubMed] 13. Haiman CA, Stram DO, Pike MC, Kolonel LN, Burtt NP, et al. A comprehensive haplotype analysis of CYP19 and breast cancer risk: the Multiethnic Cohort. Hum.Mol.Genet. 2003;12:2679–92. [PubMed] 14. Altshuler D, Brooks LD, Chakravarti A, Collins FS, Daly MJ, Donnelly P. A haplotype map of the human genome. Nature. 2005;437:1299–1320. [PubMed] 15. Stephens M, Smith NJ, Donnelly P. A new statistical method for haplotype reconstruction from population data. Am.J.Hum.Genet. 2001;68:978–89. [PubMed] 16. Key T, Appleby P, Barnes I, Reeves G. Endogenous Hormones and Breast Cancer Collaborative Group. Endogenous sex hormones and breast cancer in postmenopausal women: reanalysis of nine prospective studies. J Natl Cancer Inst. 2002;94:606–16. [PubMed] 17. Cheng ZN, Zhou HH. Contribution of genetic variations in estradiol biosynthesis and metabolism enzymes to osteoporosis. Acta Pharmacol Sin. 2000;21:587–90. [PubMed] 18. Bulun SE, Sebastian S, Takayama K, Suzuki T, Sasano H, Shozu M. The human CYP19 (aromatase P450) gene: update on physiologic roles and genomic organization of promoters. J Steroid Biochem Mol Biol. 2003;86:219–24. [PubMed] 19. Siegelmann-Danieli N, Buetow KH. Constitutional genetic variation at the human aromatase gene (Cyp19) and breast cancer risk. Br J Cancer. 1999;79:456–63. [PubMed] 20. Zhao Y, Mendelson CR, Simpson ER. Characterization of the sequences of the human CYP19 (aromatase) gene that mediate regulation by glucocorticoids in adipose stromal cells and fetal hepatocytes. Mol Endocrinol. 1995;9:340–9. [PubMed] 21. Long JR, Shu XO, Cai Q, Wen W, Kataoka N, Gao YT, et al. CYP19A1 genetic polymorphisms may be associated with obesity-related phenotypes in Chinese women. Int J Obes. 2006;8:1–6. 22. Mohamed-Ali V, Goodrick S, Rawesh A, Katz DR, Miles JM, Yudkin JS, et al. Subcutaneous adipose tissue releases interleukin-6, but not tumor necrosis factor-alpha, in vivo. J Clin Endocrinol Metab. 1997;82:4196–200. [PubMed] 23. Purohit A, Newman SP, Reed MJ. The role of cytokines in regulating estrogen synthesis: implications for the etiology of breast cancer. Breast Cancer Res. 2000;4:65–9. [PubMed] 24. Sebastian S, Takayama K, Shozu M, Bulun SE. Cloning and characterization of a novel endothelial promoter of the human CYP19 (aromatase P450) gene that is up-regulated in breast cancer tissue. Mol.Endocrinol. 2002;16:2243–54. [PubMed] 25. Zeitoun K, Takayama K, Michael MD, Bulun SE. Stimulation of aromatase P450 promoter (II) activity in endometriosis and its inhibition in endometrium are regulated by competitive binding of steroidogenic factor-1 and chicken ovalbumin upstream promoter transcription factor to the same cis-acting element. Mol Endocrinol. 1999;3:239–53. [PubMed] 26. Agarwal VR, Bulun SE, Leitch M, Rohrich R, Simpson ER. Use of alternative promoters to express the aromatase cytochrome P450 (CYP19) gene in breast adipose tissues of cancer-free and breast cancer patients. J Clin Endocrinol Metab. 1996;81:3843–9. [PubMed] 27. Long JR, Kataoka N, Shu XO, Wen WQ, Gao YT, Cai QY, et al. Genetic polymorphisms of the CYP19A1 gene and breast cancer survival. Cancer Epidemiol Biomarkers Prev. 2006;15:2115–22. [PubMed] 28. Dunning AM, Dowsett M, Healey CS, Tee L, Luben RN, Folkerd E, et al. Polymorphisms associated with circulating sex hormone levels in postmenopausal women. J Natl Cancer Inst. 2004;96(12):936–45. [PubMed] 29. Haiman CA, Hsu C, de Bakker P, Frasco M, Sheng X, Van Den Berg D, et al. Comprehensive association testing of common genetic variation in DNA repair pathway genes in relationship with breast cancer risk in multiple populations. Hum Mol Genet. 2007 Dec 3; [Epub ahead of print].
|
The publisher's final edited version of this article is available at 
