Genetic Polymorphisms in the IGFBP3 Gene: Association with Breast Cancer Risk and Blood IGFBP-3 Protein Levels among Chinese Women

The insulin-like growth factor (IGF) family includes the polypeptide ligands IGF-I and IGF-II, the IGF receptors, and six binding IGF proteins (i.e., IGFBP-1 to IGFBP-6; refs. 1-3). A large number of in vitro studies have shown that IGFs are strong mitogens for a variety of cancer cells including many breast cancer cell lines (1-3). Besides mitogenic stimulation, IGFs also inhibit apoptosis (1-3). The combination of these mitogenic and antiapoptotic effects has a profound impact on tumor growth (1-3). IGF-I and IGF-II are present in the circulation where the majority of them (>90%) are bound to IGFBP-3 and become less available to the target tissues (1-3). On the other hand, the binding of IGF-I/IGF-II to IGFBP-3 prevents these growth factors from degradation and, in some situations, may enhance the action of IGFs (1-3). IGFBP-3 also directly binds to its putative cell surface receptor and induces apoptosis independent of IGF-I/IGF-II (1-4). It has been reported that IGFBP-3 can mediate the growth inhibitory action of transforming growth factor-β1 and induce apoptosis through the tumor suppressor gene p53 in breast cancer cells (2, 3).

Although IGF-I levels in the blood have been fairly consistently shown to be positively associated with premenopausal breast cancer risk in previous epidemiologic studies (1, 5-8), the reports on IGFBP-3 have been conflicting (1, 5-13). Both positive and inverse associations of breast cancer with this molecule have been reported. The levels of IGFBP-3 can be affected by both lifestyle and genetic factors (14-16). Recently, a polymorphism (A → C transversion) in the promoter region of the IGFBP3 gene was reported. It was found that the variant genotype was associated with a significantly decreased level of IGFBP-3 in both in vitro and in vivo studies (15, 16). Several other common polymorphisms have also been reported in the IGFBP3 gene (17-19), including a G → C transversion in exon 1 (Gly → Ala substitution at codon 32; refs. 17, 18), a C → G transversion located only 17 nucleotides upstream of exon 2 (17, 18), and an A → C transversion at nucleotide 7,580 in intron 3 (19). Recently, we identified two additional polymorphisms in intron 3 of the IGFBP3 gene and evaluated, in this study, all these polymorphisms in relation to breast cancer risk.

Included in this study were subjects recruited during 1996 to 1998 in the Shanghai Breast Cancer Study. Detailed study methods have been published elsewhere (20). Briefly, this study included 1,459 incident breast cancer cases diagnosed at an age between 25 and 64 years and 1,556 age frequency-matched community controls. Cancer cases were identified through the population-based Shanghai Cancer Registry, which records virtually all cancer cases diagnosed among residents of urban Shanghai. A total of 1,602 eligible breast cancer cases were identified during the study period, of which 1,459 (91.1%) cases completed in-person interviews. Cancer diagnoses for all patients were reviewed and confirmed by two senior pathologists. Controls were randomly selected from the general population in Shanghai using the Shanghai Resident Registry, a population registry containing demographic information for all residents of urban Shanghai, and were frequency matched on age (5-year intervals) to the expected age distribution of the case subjects in a 1:1 ratio. The inclusion criteria for controls were identical to those of the cases with the exception of a breast cancer diagnosis. Of the 1,724 eligible women, 1,556 (90.3%) completed in-person interviews. The major reason for nonparticipation was refusal, accounting for 6.8% (n = 109) for cases and 9.6% (n = 166) of controls identified for the study.

A structured questionnaire was used to elicit detailed information on demographic factors, menstrual and reproductive history, hormone use, dietary habits, prior disease history, physical activity, tobacco and alcohol use, weight, and family history of cancer. All participants were measured for their current weight and circumference of waist and hips. Blood samples (10 mL from each woman) were obtained from 1,193 (82%) cases and 1,310 (84%) controls who completed the in-person interviews. These samples were processed on the same day, typically within 6 hours of sample collection, and were stored at −70°C until relevant bioassays.

The laboratory work was conducted during 2002 at the Molecular Epidemiology Laboratory of the Vanderbilt-Ingram Cancer Center (Nashville, TN). Genomic DNA was extracted from buffy coats with the use of the Puregene DNA Purification Kit (Gentra Systems, Minneapolis, MN) following the manufacturer's protocol. DNA concentration was measured using PicoGreen dsDNA Quantitation Kit (Molecular Probes, Eugene, OR). Genotyping was performed using PCR-based RFLP assays. The RFLP method failed to reveal the polymorphism (A → C) at nucleotide 7,580 reported by Sun et al. (19). To confirm this finding, we sequenced 10 study samples [from nucleotides 7,343 to 7,953 based on the Genbank accession no. M35878 sequence or from nucleotides 5,438 to 6,048 relative to transcription start site (21)]. The primers used for PCR amplification were 5′-CTCCGAGTCACTGGCATTTC-3′ and 5′-ACCAGCCCTTGTAGAACCTC-3′ (19). The PCR reactions were performed in a 20 μL mixture containing 5 ng template DNA, 0.5 unit Hotstar Taq DNA polymerase (Qiagen, Valencia, CA), 1× Qiagen PCR buffer, 1.5 mmol/L MgCl₂, 0.2 mmol/L each of deoxynucleotide triphosphates, and 0.5 μmol/L each primer. After denaturation at 95°C for 15 minutes, the PCR was performed in 35 cycles of 94°C for 45 seconds, 55°C for 45 seconds, and 72°C for 45 seconds. At the end, the reactions were extended for 7 minutes at 72°C. The PCR products were purified with a Qiagen PCR purification kit and directly sequenced with the Big Dye Terminator sequencing chemistry in ABI PRISM 3700 DNA analyzer (Applied Biosystems, Foster City, CA) for both DNA strands of the 10 study samples. Results from direct DNA sequencing confirm that the A7580C polymorphism does not exist in our study population. Two novel polymorphisms, however, were found when compared with the original IGFBP3 gene sequence in Genbank: an insertion of adenine at nucleotide 5,606 (5606InsA) and a transition from C → T at nucleotide 5,827 (C5827T) relative to transcription start site (21). These two polymorphisms, along with the three previously reported polymorphisms (A−202C, G227C, and C3804G) that were confirmed in our study population, were included in this study.

Table 1 shows PCR primers, restriction enzymes, and length of the resulting fragments in each genotype. The PCR conditions for genotyping were similar to that described above for DNA sequencing, except for a different annealing temperature and the addition of Q-solution (Qiagen) for A−202C and G227C that are in the GC-rich region. Ten microliters of each PCR product were digested with 5 units of the corresponding restriction enzyme (NEB, Beverly, MA) following the manufacturer's instructions in a final reaction volume of 20 μL. The restriction fragments were separated by gel electrophoresis on a 1.5% agarose gel containing ethidium bromide; the gel was visualized and pictured on a UV transilluminator. Because of insufficient genomic DNA or laboratory failure, genotyping data were unavailable for 6.6% of study samples.

Plasma concentrations of IGFBP-3 were measured in 390 control women who were included in an ancillary study conducted as part of the Shanghai Breast Cancer Study (6). A commercially available ELISA kit (DSL Inc., Webster, TX) was used in the assay as has been used in the majority of previous epidemiologic studies (1, 5-8). The coefficients of variation for intraassay and interassay precisions were 0.5% to 1.9% and 1.8% to 3.9%, respectively.

χ² statistics were used to evaluate case-control differences in the distribution of IGFBP3 genotypes. Odds ratios (OR) and 95% confidence intervals (CI) were used to measure the strength of the association between IGFBP3 polymorphisms and breast cancer risk after adjusting for age and other potential confounding factors in logistic regression models (22). Stratified analyses based on age (<45 and ≥45 years) were performed to examine whether the association was stronger among younger women. Tests for Hardy-Weinberg equilibrium were conducted by comparing observed and expected frequencies of genotypes using the χ² test. The linkage disequilibrium level D′ was computed for every pair of polymorphisms using the software GOLD (23). Assuming Hardy-Weinberg equilibrium, haplotype frequencies for case and control groups were estimated (24), and the ORs for common haplotypes were calculated in comparison with the haplotype carrying only the most frequent alleles in all five loci evaluated in this study. ANOVA was used to compare plasma IGFBP-3 levels by genotypes. P < 0.05 (two-sided probability) was interpreted as statistically significant.

Results on the demographic variables and known breast cancer risk factors in the study were published previously (20). Breast cancer patients and controls were comparable in age and education level. With the exception of a family history of breast cancer, statistically significant associations were observed for all major risk factors for breast cancer, including early age at menarche, late age at menopause, late age at first live birth, physical inactivity, a prior history of breast fibroadenoma, high body mass index, and high waist-to-hip ratio. More cases than controls had a family history of breast cancer, although the differences were not statistically significant, perhaps due to a few subjects having a positive family history of breast cancer.

The associations between IGFBP3 genotypes and breast cancer risk are presented in Table 2. No subjects were homozygous for the G allele at nucleotide 3,804, and an elevated OR of borderline significance was observed among women heterozygous for the G allele. For each of the other four polymorphisms, elevated risks of breast cancer were observed among women who were homozygous for the variant allele, and three of the four age-adjusted ORs were statistically significant. The ORs were essentially unchanged after adjusting for the nongenetic breast cancer risk factors that are listed in the footnote of Table 2. Only a small percentage of women had used hormone replacement therapy (2.5%) or had a family history of breast cancer (2.4%). Adjusting for these variables did not appreciably affect the risk estimate. Additional analyses were performed to evaluate the association of breast cancer risk with having more than one risk genotype. Compared with women who did not have any risk genotype, those with one to two and three to five risk genotypes had elevated risks of 1.3 and 1.7, respectively, and the trend for increases in risk was statistically significant (P < 0.01). When stratified by age, the positive association appeared to be stronger among younger women (OR 1.7, 95% CI 1.1-2.5) than older women (OR 1.2, 95% CI 0.8-1.7).

The degree of linkage disequilibrium was evaluated for every pair of polymorphism. The D′ values ranged from 0.88 to 1.00, indicating that the five polymorphisms under investigation are in close linkage disequilibrium. Because the interaction of multiple polymorphisms within a haplotype could potentially affect biological phenotypes, the frequencies of extended haplotypes were estimated, and their associations with breast cancer risk were evaluated (Table 3). Four common haplotypes (frequency of >1%) were found to account for 96.5% of chromosomes in the study population. When compared with the haplotype containing only the most frequent allele in all five polymorphic sites, a 40% elevated risk of breast cancer was found to be associated with the haplotype containing all variant alleles. The positive association was found primarily among younger women, with OR (95% CI) of 2.3 (1.3-3.9). No elevated risk of breast cancer, however, was found to be associated with other haplotypes containing one to four variant alleles.

To assess potential joint effects of IGFBP3 polymorphisms and factors related to endogenous estrogen exposure, subjects were classified based on the joint distribution of IGFBP3 polymorphisms and estrogen-related factors, including waist-to-hip ratio, body mass index, age at first live birth, age at menarche, age at menopause, and regular physical activity. The risks of breast cancer associated with IGFBP3 polymorphisms were elevated in virtually all strata defined by estrogen-related factors (data not shown in table). None of the tests based on multiplicative models, however, were statistically significant.

Figure 1 shows plasma IGFBP-3 protein levels among controls by IGFBP3 genotypes. Again, no subjects were homozygous for the variant G allele in C3804G. The age-adjusted plasma IGFBP-3 level was lower in heterozygotes than homozygotes of the C allele (wild-type). For each of the other four polymorphisms, women homozygous for the wild-type allele had the highest mean of blood IGFBP-3, and the level declined in a stepwise manner with the presence of one and two copies of the variant alleles (P for trend < 0.05). There were statistically significant differences between wild-type and variant homozygous genotypes in all four polymorphisms but none between heterozygotes and homozygotes. When all polymorphisms were considered, the mean plasma IGFBP-3 protein was the highest in subjects homozygous for the wild-type allele in all five loci (n = 235, x̄ ± SD = 4,153 ± 1,658) followed by heterozygotes in all but C3804G (n = 105, x̄ ± SD = 3,774 ± 1,195) and those heterozygous for the variant allele in C3804G or homozygous for the variant allele in any one of the other four sites (n = 32, x̄ ± SD = 3,417 ± 1,022). The trend test for this association was highly significant (P = 0.002; data not shown in Fig. 1). The level of blood IGFBP-3 protein appeared to reduce further with an increased number of risk genotypes (mean = 3,465 for one to two risk genotype groups and mean = 3,326 for three to five risk genotype groups). The sample size, however, was small, and the difference between these two groups was not statistically significant.

Unlike most Mendelian disorders that are caused by the mutation of a single gene, breast cancer is a complex disease, resulting from the insult of multiple genetic and environmental factors (25-27). Germ line mutation of the BRCA1 and BRCA2 genes accounts for <10% of breast cancer cases in the general population (26, 27). Over the past 5 years, a great deal of research has been focused on identifying genetic factors that are responsible for breast cancer, particularly the genes that are involved in estrogen and carcinogen metabolism (26). Because of their high frequencies in the general population and their potential interactions with environmental factors, these genes may be responsible for a high proportion of breast cancer cases (26, 27). In this large, population-based, case-control study, we found significant associations of genetic polymorphisms in the IGFBP3 gene with the risk of breast cancer and a correlation between IGFBP3 genotype and phenotype. These findings suggest that IGFBP3 polymorphisms may be an important genetic factor for breast cancer susceptibility.

Several epidemiologic studies have been conducted to evaluate the association of blood IGFBP-3 levels with breast cancer risk along with other IGF-related molecules. The results, however, have been conflicting. Several case-control studies reported an inverse association (9, 10), whereas others reported a positive association (6, 11, 13). Results from prospective studies have also been inconsistent. Both positive (7) and inverse (5, 7) associations have been reported. Two prospective cohort studies showed no association of IGFBP-3 levels with breast cancer risk (8, 12). We recently reported that blood IGFBP-3 level was positively associated with the risk of breast cancer in a subset of participants from the Shanghai Breast Cancer Study (6). This result appears contradictory to the findings from the current study, in which the risk of breast cancer was associated with the genotypes related to reduced blood IGFBP-3 levels. The conflicting results from epidemiologic studies on blood IGFBP-3 levels are perhaps not unexpected, given the dual roles of IGFBP-3 protein in regulating the actions of IGF-I. The net effect of IGFBP-3 on IGF-I is heavily determined by the activity of IGFBP protease, particularly at the target tissue. Blood level of IGFBP-3 may not reflect the level and action of this protein in the target tissues, and most epidemiologic studies have no access to normal target tissue samples to evaluate the association of this protein with cancer risk. Therefore, genetic polymorphisms in the IGFBP3 gene could be a better indicator of the level of IGFBP-3 protein in the target tissues than that in the circulation. Our findings for a reduced blood IGFBP-3 level and an elevated risk of breast cancer among subjects carrying the variant alleles of this gene are consistent with the evidence from most in vitro experiments and some epidemiologic studies. Because of the limitations of using blood IGFBP-3 levels as the surrogate measure of the level of this protein in the target tissue, caution should be exercised in interpreting findings based on blood IGFBP-3 measurement.

It was recently reported that a transversion from A to C at nucleotide −202 in the promoter region of the IGFBP3 gene significantly decreased the expression of this gene in an in vitro study (16). This polymorphism was also found to be associated with a reduced level of blood IGFBP-3 protein in three human studies (15, 16, 28). In agreement with these reports, we observed in our study a gene-dose association between blood IGFBP-3 levels and the number of variant C alleles in this polymorphism. Furthermore, we have shown in this study that women homozygous for the C allele were at an elevated risk of breast cancer, whereas heterozygosity for the C allele was unrelated to risk. The reduction in the levels of circulating IGFBP-3 protein, however, was less evident in women heterozygous for the C allele than those homozygous for this allele.

Similar positive associations were observed in our study for the other four polymorphisms, although there are currently no in vitro data available to support the functionality of these polymorphisms. The G → C transversion at nucleotide 227 results in a change of Gly → Ala in codon 32 (exon 1). The other three polymorphisms were identified in introns, which could affect the splicing site on mRNA after transcription. In particular, the polymorphism (C → G) at nucleotide 3,804 is located only 17 nucleotides upstream of exon 2, within the branch site that is important in RNA splicing (25). If more than one of the polymorphisms evaluated in the study were functionally important, we would expect that the risk of breast cancer might be increased with the number of the risk genotypes. In agreement with this notion, we did observe an increase in breast cancer risk with the number of risk genotypes. In the analysis of extended haplotypes of this gene, we found that the risk of breast cancer was only associated with the haplotype that carries the variant allele in all five polymorphic sites, suggesting that more than one of the polymorphisms may be functionally significant.

The frequencies of variant alleles in our study differ from those reported previously in other populations. For the A−202C polymorphism, the frequencies for the C allele were 0.40 and 0.53 in Caucasian men and women, respectively (15, 16). In our study, however, the frequency of this allele was only 0.23. In a small study of 23 Caucasians, Eggermann et al. found the variant allele frequencies were 0.59 and 0.22 in G227C and C3804G, respectively (13), again much higher than Chinese women whose frequencies were 0.21 and 0.03, respectively, for these two polymorphisms. These data are consistent with the lower incidence of breast cancer observed among Chinese women when compared with their American counterparts (29). Very recently, Schernhammer et al. (28) reported results from a case-control study evaluating the association between A-202 polymorphism and breast cancer risk. Although no overall association was found, the AA genotype was associated with ∼30% reduced risk of breast cancer among premenopausal women <50 years old. However, the sample size for this group of women (70 cases and 70 controls) was small, and the OR was not statistically significant. Nevertheless, the result from this subgroup of women is consistent with the role of IGFBP-3 in breast tumorigenesis and in agreement with the findings from our study, which was conducted in a relatively younger population with a mean age of 47.14 years in controls and 47.65 years in cases.

Noteworthy strengths of this study are its large sample size, population-based design, and a very high response rate, which minimizes potential selection bias. The consistency of the findings for both IGFBP3 phenotype and haplotype with breast cancer risk, as well as in stratified analyses, strongly indicates that the associations identified in this study are unlikely due to a type I error, a major concern in many association studies of genetic polymorphisms with a small sample size. Over 98% of study subjects belong to a single ethnic group (Han Chinese); thus, possible confounding effects due to population stratification are not a major concern in this study. Extensive survey data were collected in this study, and potential confounding factors were adjusted for in the analysis.

In summary, in this large, carefully conducted, population-based, case-control study, we have shown that IGFBP3 polymorphisms were associated with the level of circulating IGFBP-3 protein and the risk of breast cancer. These findings are consistent with the cancer inhibitory effect of IGFBP-3 demonstrated in most in vitro experiments and some epidemiologic studies showing an inverse association of this protein with breast cancer risk. Our findings are new and may have significant public health implications for identifying high-risk women for the prevention of breast cancer, the most common malignancy in many parts of the world.

We thank Drs. Qi Dai, Fan Jin, and Jia-Rong Cheng for contributions in coordinating data and specimen collection in Shanghai, Bethanie Hull for technical assistance in the preparation of this article, and all study participants and research staff of the Shanghai Breast Cancer Study for support.