Genetic variation in candidate obesity genes ADRB2, ADRB3, GHRL, HSD11B1, IRS1, IRS2, and SHC1 and risk for breast cancer in the Cancer Prevention Study II

Obesity is a leading risk factor for postmenopausal breast cancer. High body mass index (BMI) and adult weight gain have consistently been associated with increased risk for breast cancer [1‐3]. This increased risk is thought to result primarily from circulating estrogens derived from extraglandular aromatization of plasma androstenedione to estrone in adipose tissue [4, 5]. However, adipose tissue is also an active endocrine and metabolic organ that can have far-reaching effects on the physiology of other tissues [2]. Therefore, proteins that are secreted by adipose tissue or are involved in regulating body mass may play a role in breast tumor development through mechanisms independent from estrogen production.

To examine whether genes that encode adiposity-related proteins influence the risk for breast cancer, we conducted a nested case control study among predominantly white, postmenopausal women from the American Cancer Society Cancer Prevention Study II Nutrition Cohort. Seven candidate genes were chosen for inclusion in this study: ADRB2, ADRB3, GHRL, HSD11B1, IRS1, IRS2, and SHC1. Although many more genes are related to obesity and energy balance, we chose candidate genes that had previously been associated with at least one measure of obesity and with risk for breast cancer [6‐10] or other cancer [11, 12], or had demonstrated activity in breast cancer cell lines [13‐15].

A total of 648 cases and 659 controls were included in the analysis. Characteristics of the study population are shown in Table 1. The study population was 99% white, and the mean age at breast cancer diagnosis was 68.9 years. Cases were slightly more likely than controls to have a BMI of 30.0 kg/m² or greater (16.4% versus 15.6%), but this difference was not statistically significant (P = 0.16). Cases and controls reported similar weight change since age 18 years (27.9 lb [12.7 kg] versus 27.1 lb [12.3 kg]; P = 0.58). It is well established (in this cohort [23] and others [1, 24]) that an interaction exists between obesity and PMH use, in which the deleterious effect of obesity is observed only among women not using PMH. The apparent lack of association in this analysis can be accounted for by this relationship. Sixty per cent of the study population use PMH; among those women not using PMH, obesity is associated with increased risk for breast cancer.

Figure 1 shows the log-additive P values for the association between breast cancer and all 38 SNPs analyzed. Table 2 shows the association between breast cancer and tagging SNPs in HSD11B1. In multivariate adjusted models, SNPs rs11807619 and rs932335 were both associated with breast cancer. For rs11807619, the T allele was associated with a 40% increased risk for breast cancer (OR = 1.41, 95% CI = 1.10 to 1.80 among women with any T allele versus G/G women; P = 0.006). For rs932335, the adjusted ORs were 1.56 (95% CI = 1.22 to 2.00) for G/C and 1.83 (95% CI = 1.01 to 3.33) for C/C compared with G/G (P for trend = 0.0002). Further examination of these SNPs using Haploview indicated that they were highly correlated (D' = 0.99 and r ² = 0.74; Figure 2). Haplotype analysis of these two SNPs showed that rs11807619 was no longer an independent predictor of risk. The two haplotypes that carry the C allele of rs932335 were both associated with breast cancer.

Results for IRS2 SNPs are shown in Table 3. Three of the 11 tagging SNPs for IRS2 were associated with breast cancer (rs4773082, P = 0.007; rs2289046, P = 0.016; and rs754204, P = 0.03). The C allele of rs4773082 was associated with increased risk for breast cancer (OR = 1.41, 95% CI = 1.10 to 1.81 for T/C and C/C versus T/T). For rs2289046 the G allele was associated with a reduced risk for breast cancer; the adjusted ORs were 0.79 (95% CI = 0.61 to 1.01) for A/G and 0.52 (95% CI = 0.32 to 0.84) for G/G compared with A/A. The T allele of rs754204 was associated with increased risk for breast cancer (OR = 1.32, 95% CI = 1.03 to 1.70 for T/C and T/T versus C/C). These SNPs span a region of 8.1 kb, with D' ranging from 1.0 to 0.85 and r ² ranging from 0.71 to 0.42 (Figure 2). Thus, we created a haplotype of these three SNPs. The TGC haplotype is associated with decreased risk (P = 0.0018), and the magnitude of the effect is similar to that of single SNP results for rs2289046.

No other SNPs in any of the other five genes were associated with breast cancer (Figure 1 and Additional file 1), and neither was there any evidence of effect modification by BMI, PMH use, family history of breast cancer, or history of breast cysts for any gene or SNP in this analysis (data not shown). None of our results for associations with breast cancer change if BMI is used in the model in place of weight change; adding BMI at age 18 years also does not change any associations with breast cancer. It should be noted that only one of the tested SNPs (rs932335 in HSD11B1) would be statistically significant after adjusting for multiple testing using the Bonferroni method.

Table 4 shows mean BMI and adult weight change by genotype for the IRS2 SNPs separately for cases and controls. We did not find any associations among our control women. Among the cases, the G allele for IRS2 rs2289046, which was protective for breast cancer, also was associated with less weight gain in adulthood compared to the A allele (P = 0.016). The T allele of rs754204, which was associated with increased risk for breast cancer, was associated with greater adult weight gain compared with the C allele (P = 0.012). The A allele of rs12584136 was associated with greater weight gain (P = 0.026), but was not associated with breast cancer. None of these SNPs would be statistically significant after adjusting for multiple testing using the Bonferroni method. None of the SNPs were associated with BMI. None of the SNPs in any of the other genes examined, including HSD11B1, were associated with adult weight change or BMI (data not shown), and neither were any of the SNPs associated with self-reported BMI at age 18 years (data not shown).

Our results suggest that common variation in HSD11B1 and IRS2 may be associated with breast cancer among postmenopausal women. These associations appear to be independent of adiposity. The HSD11B1 rs932335 minor allele (allele frequency = 18% among controls) was associated with a nearly twofold increased risk for breast cancer (P = 0.0002). This SNP is highly correlated with rs11807619, and in haplotype analysis only the two common haplotypes that carried the rs932335 C allele were associated with risk for breast cancer. Three SNPs in IRS2 were associated with breast cancer and were also found to be in LD (rs4773082, rs2289046, and rs754204). When these SNPs were modeled as a haplotype, the haplotype that included the G allele of rs2289046 (allele frequency = 34% among controls) was the only common haplotype associated with breast cancer (OR = 0.76, 95% CI = 0.63 to 0.92 for one copy of TGC versus CAT; P = 0.0050). Thus, these tagging SNPs may mark regions of the genome that harbor risk alleles for breast cancer.

HSD11B1 is located at 1q32-41 and encodes the enzyme 11β-hydroxysteroid dehydrogenase type 1, which converts the inactive steroid cortisone into the active glucocorticoid cortisol, principally in the liver and adipose tissue [25]. To our knowledge, HSD11B1 has not been examined previously in any candidate gene study. However, 11 SNPs in HSD11B1 were included in a recent genome-wide scan of breast cancer [26, 27] and none were associated with breast cancer, including SNPs that were highly correlated with ours. Two SNPs, rs4393158 and rs2235543, included in the scan, were also genotyped in our study with similar (negative) results.

Polymorphisms in HSD11B1 have been associated with type 2 diabetes and hypertension [25]. Recently, HSD11B1 was identified from a genome-wide linkage analysis as a possible candidate gene for weight change [28]. However, others [29] have reported no association with HSD11B1 and body composition. HSD11B1 variation has also been associated with polycystic ovarian syndrome [25], and ovarian steroids have been shown to regulate 11β-hydroxysteroid dehydrogenase expression in experimental models [13].

IRS2, a member of the insulin receptor substrate family, is expressed in nearly all cells and tissues. IRS2 regulates body weight control and glucose homeostasis [11]. Although the primary role played by IRS2 is in insulin signaling, IRS2 can also act as a substrate for insulin-like growth factor receptor and thus it acts in insulin-like growth factor signal transduction. Evidence also suggests that IRS proteins may play a role in estrogen signaling [8]. Previous studies of IRS2 variation and measures of adiposity have been inconsistent. Lautier and coworkers [30] suggested that IRS2 haplotypes were associated with severe obesity (BMI ≥40 kg/m²) among French individuals, but a later report by Sweeney and colleagues [31] did not confirm this association among US white and Hispanic women. Genome-wide linkage studies of obesity and BMI have not identified IRS2 as an important locus; however, the 13q region near IRS2 showed some evidence of linkage to BMI in a recent meta-analysis [32]. We observed modest associations between three IRS2 SNPs and adult weight change among our cases. Because these associations are not seen among the controls, we suspect that these results are due to chance. Furthermore, we do not think that the relationship between IRS2 and weight change explains the association between IRS2 SNPs and breast cancer. As shown in Table 3, the associations between these SNPs and breast cancer change very little, if at all, when adult weight change is included in the multivariate models. This suggests that the associations we observed with breast cancer are not mediated through increasing adiposity.

At least two previous studies have examined whether IRS2 SNPs are associated with breast cancer, but neither found an association [8, 10]. The first study [10] sequenced coding regions to identify variants of interest, and then examined two IRS2 SNPs (rs4773092 and rs1805097) in a familial breast cancer study. Slattery and coworkers [8] used a large case-control study, including both white and Hispanic women, to examine the association between breast cancer and rs1805097 (G1057D), a SNP that had previously been associated with both colon cancer [11] and obesity [30]. We did not find an association between rs4773092 and postmenopausal breast cancer, and were unable to genotype rs1805097 because of assay failure. Data available from the Cancer Genetic Markers of Susceptibility initiative [27] indicates that one SNP in IRS2 is marginally associated with breast cancer (rs12584136, P = 0.051). This SNP was not associated with risk in our data. None of the other IRS2 SNPs genotyped in Cancer Genetic Markers of Susceptibility were associated with breast cancer, including those that were highly correlated with SNPs in our study.

It should be noted that we might have missed an association with breast cancer and the 27 singleton SNPs that we excluded from this study. Seven of these were in HSD11B1, six each in ADRB2 and IRS2, and four each in GHRL and IRS1. Singleton SNPs, by definition, are not correlated with any other SNPs that have been genotyped in the region. We chose to exclude singleton SNPs that were not near an exon, because we reasoned that they were likely to be less informative than SNPs that were correlated with other nearby SNPs, or SNPs that were located in or near an exon.

The strengths of this study include the population-based design and the use of tagging SNPs to conduct an effective survey of the variation in these genes. The primary limitation of this study is its relatively small sample size. Although these data suggest that HSD11B1 and IRS2 may be associated with risk for breast cancer, it is also possible that these findings are due to chance. If these findings are confirmed in subsequent studies, then additional genotyping in these regions is warranted.

We conducted a nested case-control study among 648 cases and 659 controls to determine whether germline variation in seven genes that encode adipose-related proteins was associated with breast cancer. Our results suggest that common variation in HSD11B1 and IRS2 may be associated with breast cancer among postmenopausal white women. The HSD11B1 rs932335 minor allele (allele frequency = 18% among controls) was associated with a nearly twofold increased risk for breast cancer (P = 0.0002). Three SNPs in IRS2 were also associated with breast cancer (rs4773082, rs2289046, and rs754204). When these SNPs were modeled as a haplotype, the haplotype that included the G allele of rs2289046 was the only common haplotype associated with breast cancer (P = 0.0018). IRS2 rs2289046 and rs754204 were also associated with adult weight gain. No associations with breast cancer or with adiposity were found in the other genes included in this study: ADRB2, ADRB3, GHRL, IRS1, and SHC1.