Circulating levels of insulin-like growth factor-I (IGF-I) and insulin-like growth factor-binding protein 3 (IGFBP-3) vary considerably between normal individuals. Recent epidemiological studies have provided evidence that these levels are predictive of risk of several common cancers. To evaluate possible sources of variation of the levels of circulating IGF-I and IGFBP-3 in females, we studied specific candidate genetic and nongenetic factors in 311 nulliparous, premenopausal Caucasian women, 17–35 years of age. Women who used oral contraceptives (OC) had reduced levels of IGF-I (269 versus 301 ng/ml; P = 0.001 adjusted for age) and increased levels of IGFBP-3 (4213 versus 4009 ng/ml; P = 0.002, adjusted for age) compared with nonusers. The ratio of IGF-I:IGFBP-3 was associated with the dose of estrogen contained in the OC (Ptrend = 0.006, adjusted for age). We identified a novel single bp polymorphism in the promoter region of the gene encoding IGFBP-3. This polymorphism was related to the level of IGFBP-3 in the circulation. Mean IGFBP-3 levels were 4390, 4130, and 3840 ng/ml for the AA, AC, and CC genotypes, respectively (Ptrend = 0.006, adjusted for age and OC use). We observed no effect of a recently described polymorphism in the promoter region of the gene encoding IGF-I on the plasma IGF-I level, but there was evidence for a modifying effect of this locus on the influence of OC on the IGF-I level. Our results support the view that circulating IGF-I levels and IGFBP-3 levels are complex traits and are influenced by a number of interacting genetic and nongenetic factors.
The IGF3 system plays key roles in regulating cell proliferation and apoptosis, both at whole-organism and cellular levels (1). IGFs have characteristics of both classic endocrine hormones and tissue growth factors. Circulating IGF-I and IGFBP-3 levels are largely dependent on hepatic production and are tightly regulated. In addition, autocrine and paracrine production of IGF-I, IGFBP-3, and other molecules of the IGF system within IGF-responsive tissues are of major physiological importance (1, 2).
There is considerable interindividual variability in circulating levels of IGF-I and IGFBP-3 and possibly parallel interindividual variability in the tissue expression of the genes encoding these proteins. Circulating concentrations of these two proteins are highly correlated because IGFBP-3 carries at least 90% of IGF-I present in the circulation (3, 4, 5). However, there are complex equilibria in the circulation whereby IGF-I as well as IGF-II may be present in complexes with IGFBP-3, with other IGF-binding proteins, or in a free state.
Recent epidemiological studies suggest that circulating IGF-I levels are positively correlated with risk of several common cancers and that this risk is attenuated by higher IGFBP-3 levels (e.g., Refs. 6, 7, 8, 9). A hypothesis to explain this observation is that the circulating IGF-I:IGFBP-3 ratio represents a surrogate for tissue IGF bioactivity, which is itself a surrogate for (or even a determinant of) cell renewal rate in epithelial cell populations (10). There is evidence that if factors such as mutation rate and DNA repair are held constant, then accumulation of somatic cell genetic damage leading to carcinogenesis will be influenced by kinetics of cell renewal, including factors such as cell turnover rate and probability of survival of partially transformed cells (11, 12).
Twin studies (13, 14) have shown that about 50% of the interindividual variability in circulating IGF-I and IGFBP-3 levels is genetically determined, but specific loci involved have not been described. We undertook studies to evaluate interindividual variability in circulating IGF-I and IGFBP-3 levels in relation to specific genetic and nongenetic factors.
Materials and Methods
Five hundred and seventeen women between 17 and 35 years of age were recruited from the University of Toronto, from the Bay Center for Birth Control, and from the Toronto community through flyers and e-mail. The protocol was approved by the Ethics Committee of the University of Toronto. The present study deals with the results on 311 white women. These women had never been pregnant, had no past diagnosis of cancer, and did not have diabetes. The remaining 206 women were excluded for the following reasons; 172 were of other ethnic backgrounds (non-Caucasian), Asian (n = 77), Indian-Pakistani (n = 28), black (n = 15), or other or mixed ethnicity (n = 52). Four women had been pregnant, three women had diabetes, one woman had undergone hysterectomy, and one woman was currently using Norplant. Twenty-five women were excluded because of missing information, and six women had missing IGF-I and IGFBP-3 values. We were unable to reach one woman who had left all of the OC questions blank and 18 women who had not answered the diabetes question.
The mean age of the 311 women included in the study was 25.4 years; 71% had graduated with a degree or diploma from college, 87% were born in Canada, 14% kept a vegetarian diet, 15% were current smokers, and 90% sometimes used alcoholic beverages. The mean age of the 206 women excluded from the study was 25.4 years; 61% had graduated with a degree or diploma, 52% were born in Canada, 10% kept a vegetarian diet, 19% were current smokers, and 73% sometimes used alcoholic beverages. BMI was missing for 16 women who were included in the study. The study protocol included a morning visit to the study center on a random day of the menstrual cycle. All of the women completed a self-administered questionnaire, detailing information on age, ethnic group, menstrual history, current diet (vegetarian or not), exercise, and smoking. A brief medical history included the current use of OCs and other medications and a family history of cancer. Samples of various packages of OC brands were displayed at the center, and the women were asked to identify their current brand. All of the contraceptive formulations contained 17α ethinyl estradiol as estrogen, but the type of progesterone varied. Each type of contraceptive was coded as to the amount of estrogen and the type of progesterone it contained. The average daily ethinyl estradiol dose was calculated for women who used triphasic pills. Seven different types of progesterone were recorded. Women were asked to indicate the date of the first menstrual flow of the previous cycle and were required to report by telephone when the onset of menses began for the next cycle. The date of blood draw was then classified according to the calculated day of the menstrual cycle. Menstrual cycle phase was missing for 39 women, 12 women did not call back with the date of their next period, and 27 women had irregular periods. Menstrual cycle phase was categorized into seven categories relative to the onset of bleeding and the onset of the next menstrual period. Blood samples were taken for analysis of plasma levels of IGF-I and IGFBP-3 and for DNA extraction.
IGF-I and IGFBP-3 Plasma Levels
Levels of IGF-I and IGFBP-3 in plasma were assayed using ELISA methodology as described previously (7).
Genomic DNA was extracted from 20 ml of peripheral blood using Puregene DNA extraction kits. The previously described variable number tandem repeat polymorphism in the IGF1 gene (15) was evaluated using the following PCR primers: sense 5′ GCT AGC CAG CTG GTG TTA TT 3′ and antisense 5′ ACC ACT CTG GGA GAA GGG TA 3′. PCR reactions were performed using 25 ng of template DNA for each reaction, 0.4 mm of each primer, 200 mm of each deoxynucleotide triphosphate, 1.0 mm MgCl2, 2% DMSO, 1.25 units of Taq polymerase (Life Technologies, Inc.), and the manufacturer’s standard buffers in a final reaction volume of 25 μl. To evaluate the variable number tandem repeat in the IGF1 gene, the amplified PCR products, labeled with a Cy5 fluorescent label, were subject to acrylamide gel electrophoresis. The gels were dried and scanned on a Storm 860 Phosphorimaging System (Molecular Dynamics, Sunnyvale, CA) to visualize and score the repeat alleles. The number of repeat elements was determined by measuring the size of the PCR products compared with that predicted from the controls.
We identified novel polymorphisms in the IGFBP-3 promoter regions by directed sequencing of 40 genomic DNA samples from a multiethnic Montreal population. A single bp polymorphism at position −202 relative to the CAP site was selected for further study because of preliminary evidence that genotype at this locus was related to circulating IGFBP-3 level. We designed an RFLP assay for this polymorphism. A 50 to 250-ng aliquot of genomic DNA was mixed with PCR buffer, supplemented by 1.6 μm of primers, 1 mm MgCl2, 0.1 mm of each deoxynucleotide triphosphate, 2% DMSO, and 2 units of Taq DNA polymerase (Life Technologies, Inc., Burlington, ON). Primer sense and antisense were 5′-CCA CGA GGT ACA CAC GAA TG-3′ and 5′-AGC CGC AGT GCT CGC ATC TGG-3′, respectively. The cycling parameters consisted of an initial incubation of 10 min at 94°C followed by 35 cycles of 30 s at 96°C, 30 s at 64°C, and 1 min at 72°C. The reaction was terminated after a final extension time of 5 min at 72°C. PCR product (20 μl) was digested with 5 units of Alw21I (MBI Fermentas, Flamborough, ON) from 3 to 14 h at 37°C. Digestion products were visualized on a 2% agarose gel stained with ethidium bromide.
Spearman’s correlation coefficient was used to estimate the correlation between levels of IGF-I, IGFBP-3, and age. The effects of other variables were measured using multivariate linear regression models, adjusting for age. Contraceptive preparations were classified according to the dose of estrogen (continuous) and the type of progestin (categorical). The unstandardized regression coefficient was labeled as “β” and presented when a multivariate model was used.
All of the women were between 17 and 35 years of age. The mean plasma IGF-I level and the mean plasma IGFBP-3 level declined continuously with age (IGF-I, rs = −0.40, P = 10−6; IGFBP-3, rs = −0.27, P = 10−6; Fig. 1). The IGF-I:IGFBP-3 ratio was also negatively associated with age (rs = −0.30; P = 10−7). The correlation between the IGF-I level and the IGFBP-3 level was 0.561 (P = 10−7).
OC use was an important determinant of the IGF-I level. One hundred and fifty five women (50% of total) were current users of OCs. Women who used OCs had significantly lower IGF-I levels (269 ng/ml versus 301 ng/ml; β = −32.3; P = 0.001), significantly higher IGFBP-3 levels (4209 ng/ml versus 4013 ng/ml; β = 203; P = 0.006), and significantly lower IGF-I:IGFBP-3 ratios (0.064 versus 0.075; β = −1.1 × 10−2; P = 5 × 10−10), compared with nonusers. There were no significant differences between OC users and nonusers in terms of age, age at menarche, height, weight, or BMI. The strength and statistical significance of the association between OC use and IGF-I and IGFBP-3 levels remained essentially the same after adjustment for age.
A clear dose-response relationship was present between the average daily dose of ethinyl estradiol and the plasma IGF-I level (β = −3.52; P = 0.02, adjusted for age; Fig. 2). In contrast, IGFBP-3 levels were not significantly influenced by the estrogen dose. The ethinyl estradiol content was also associated with the IGF-I:IGFBP-3 ratio (β = −8 × 10−4; P = 0.006, adjusted for age). There was no association observed between the IGF-I or IGFBP-3 levels and the type of progesterone used.
We also evaluated the plasma hormone levels in relation to the day of the menstrual cycle. Among those women who were not taking OCs and who experienced natural menses, the IGF-I:IGFBP-3 ratio was lowest in the early follicular phase, then rose steadily during the follicular and luteal phase, peaking just before menses (Fig. 3). Among OC users, it is common to take the OC for 21 days, followed by 7 days of no pill (or placebo). The IGF-I level declined when the OC was taken and then rose when the placebo was introduced (Fig. 3). The IGF-I:IGFBP-3 ratio showed a similar pattern. Statistical modeling indicated that the observed IGF-I data better fit a parabolic pattern than a straight line (P = 0.002). Among OC users, IGFBP-3 levels were highest during the first 4 days of the menstrual cycle, when placebo pills are consumed. Among nonusers, IGFBP-3 levels showed no consistent pattern over the menstrual cycle.
Age at Menarche.
Age at menarche was not significantly associated with IGF-I levels, IGFBP-3 levels, or the IGF-I:IGFBP-3 ratio in the overall study group. However, among OC users, age at menarche was positively associated with the IGF-I plasma levels (β = 10.5; P = 0.02, adjusted for age and estrogen dose) and with the IGF-I:IGFBP-3 ratio (β = 2 × 10−3; P = 0.02 adjusted for age and estrogen dose). There was no significant association between age at menarche and the IGFBP-3 level.
Height, Weight, and BMI.
In the overall study group, there was no significant association between birth weight, height, current weight, or BMI and the plasma level of IGF-I or IGFBP-3. However, in the subgroup of women on OCs, weight was negatively associated with the IGF-I:IGFBP-3 ratio, although the significance level was marginal (β = 2.4 × 10−4; P = 0.035, adjusted for age and estrogen dose). In this subgroup, BMI was negatively associated with the IGF-I plasma level and with the IGF-I:IGFBP-3 ratio (β = 3.7; P = 0.04 and β = 9.1 × 10−4; P = 0.007, respectively, adjusted for age and estrogen dose; Fig. 4).
Smoking, Physical Exercise, and Diet.
Neither current nor past smoking, the number of h spent in intense physical exercise/week, a vegetarian diet, nor the amount of coffee consumed were found to be significant predictors of the plasma levels of IGF-I or IGFBP-3 or of the IGF-I:IGFBP-3 ratio. Thirty-one percent of the women had not eaten breakfast at the time of the blood draw. No significant difference in IGF-I or IGFBP-3 plasma levels was seen between women who had eaten breakfast and women who were fasting.
Women who consumed four or more alcoholic drinks/week had significantly lower IGFBP-3 plasma levels than women who drank less (3953 ng/ml versus 4173 ng/ml; β = −208; P = 0.009, after adjustment for age and OC use). The suppressive effect of alcohol consumption on IGFBP-3 level was particularly evident in OC users (Fig. 5). IGF-I levels were lower among women who consumed four or more drinks/week, compared with women who drank less (273 ng/ml versus 290 ng/ml), but this difference was not significant (β = −8.3; P = 0.39).
Promoter Polymorphism of IGFBP3.
The −202 polymorphism was a statistically significant predictor of IGFBP-3 plasma levels in young women (Fig. 6). Women with the AA genotype had the highest levels of plasma IGFBP-3, women with AC were intermediate, and women with CC had the lowest IGFBP-3 levels (β = −298; test for trend P = 6 × 10−10, adjusted for age and OC use). The odds ratio for falling below the median value for IGFBP-3 in the population of women with no variant allele was 2.0 for carriers of one C allele (P = 0.02) and 5.4 for carriers of two C alleles (P = 6 × 10−6). Fig. 7 and Fig. 8 show possible interactions between the IGFPB3 genotype, the consumption of alcohol, the use of OCs, and plasma hormone levels. Fig. 9 shows the percentage of the IGFBP-3 genotypes by BMI quartile.
Polymorphic Variation of IGF1.
A recent report by Rosen et al. (15) suggested that the number of copies of the common 19-repeat allele in the promoter region of the IGF1 gene is related to the serum IGF-I level. In our population, plasma levels of IGF-I, IGFBP-3, and the IGF-I:IGFBP-3 ratio were similar among women with zero, one, or two copies of the 19-repeat allele of the IGF1 gene. However, among the subgroup of women who used OCs, women with at least one copy of the 19-repeat allele had significantly lower IGF-I levels than women without the 19-repeat allele (264 ng/ml versus 315 ng/ml; β = −46.3; P = 0.025, adjusted for age and estrogen dose; Fig. 10). The interaction between the presence of the 19-repeat allele of IGF1 and OC use on the IGF-I level was statistically significant (β = −66.8; P = 0.04). Fig. 11 shows IGF1 genotype by BMI quartile.
Sources of IGF-I and IGFBP-3 Variation.
In our population, the circulating level of IGF-I was influenced by age, alcohol use, and OC use. In univariate models, 18% of the variability in the IGF-I level was because of age, 4% was because of OC use, and 2% was because of alcohol consumption. In a multivariate model, these three factors explained 21% of the total IGF-I plasma variation. Overall, the plasma IGFBP-3 levels were significantly influenced by age, by the IGFBP3 genotype, by alcohol use, and by current OC use. In univariate models, the percentage of variability of IGFBP-3 levels was 9% because of the IGFBP3 polymorphism, 8% because of age, 3% because of alcohol consumption, and 2% because of current OC use. In a multivariate model, 24% of the IGFBP-3 plasma variation was explainable by these four factors.
Prior studies concerning circulating IGF-I and IGFBP-3 levels have emphasized issues related to the pathological extremes provided by acromegaly and growth hormone deficiency. Relatively little is known about the determinants of or the clinical significance of the considerable interindividual variability in IGF-I and IGFBP-3 levels within normal populations. However, it is clear that these levels are complex traits determined by interacting genetic and nongenetic factors.
We observed a strong negative correlation between the circulating IGF-I level and age. These results are consistent with previous reports (3, 4, 5) and provide evidence that the collection, storage, and assay of our samples were appropriate.
Our findings concerning the influence of OCs confirm and extend prior reports (16) and provide the first evidence for an effect of OCs on IGFBP-3 level. However, caution is required in interpreting the underlying physiology. In menopausal women, transdermal estrogen replacement raises IGF-I levels whereas oral therapy suppresses them (17, 18). Thus, it is possible that the suppressive effects of oral estrogens on serum IGF-I represent a pharmacological effect resulting from hepatic exposure to superphysiological estrogen levels delivered via the portal circulation. It is reasonable to hypothesize that, in general, autocrine and paracrine IGF-I expression in extrahepatic tissues may be regulated in parallel with hepatic expression and circulating level. This assumption may not be valid, however, in the setting of oral estrogen therapy.
The observed variation of IGF-I and IGFBP-3 levels with the menstrual cycle was more marked in the subgroup of women taking OCs. This is not unexpected, because these women have less variability in hormone levels during the menstrual cycle. However, our study is cross-sectional, and it will be necessary to obtain serial measurements in subjects throughout the menstrual cycle to obtain precise estimates of the influence of the menstrual cycle on hormone levels. It also will be of interest to determine whether there are subsets of women in whom IGF-I and IGFBP-3 levels vary more or less than the general population with menstrual cycle, inasmuch as prior studies (19, 20) did not detect IGF-I variation with the menstrual cycle. This is an important issue because in clinical practice and in most research studies there is no attempt to control for time of menstrual cycle or exogenous estrogens in interpreting IGF-I levels. We recommend that future epidemiological studies control for these factors in interpreting IGF-I and IGFBP-3 levels.
The −202 IGFBP3 polymorphism represents a novel genetic factor that has a significant influence on IGFBP-3 plasma levels in young women. In vitro functional studies of promoter activity undertaken with the polymorphic variants showed that the C allele had 52% of the promoter activity of the A allele, in keeping with the clinical observations reported here.4 It will be important to determine whether allele frequency at this locus varies with race, because racial variations of IGFBP-3 levels have been reported (21, 22). This may have relevance to cancer risk, either in a direct way or through a modifying effect involving classic environmental or genetic risk factors. For example, there has been speculation that the higher rate of prostate cancer in African-Americans is related at least in part to their lower IGFBP-3 levels (21, 22). Our study also provides preliminary evidence that genotype at this locus may modify nongenetic factors related to IGFBP-3 level. For example, although in general OC use tends to raise IGFBP-3 level, individuals with an AA genotype already have levels higher than the population as a whole, and these are not further raised by OCs.
Our findings with respect to the IGF1 polymorphism reported by Rosen et al. (15) require follow-up because they are based on small subsets. However, the data provide evidence that plasma IGF-I level is paradoxically increased with OC use in the minority of women who lack a 19-repeat allele. If this unusual response to OCs in a genetically defined subset of women is confirmed, it will be important to determine whether there are any clinical correlates. Although most population studies (e.g., Ref. 23) do not demonstrate a substantial increase in breast cancer risk associated with OC use, the possibility that a genetically defined subgroup of women is at increased risk if exposed to OC has not been ruled out.
We hope to extend the present study to identify additional genetic and lifestyle factors that influence plasma levels of IGF-I, IGFBP-3, and other related molecules. The extent to which variability in IGF-I levels and IGFBP-3 levels within the normal range will have clinical relevance, in terms of growth, aging, and cancer risk, is a topic deserving of further study.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Supported by the National Cancer Institute of Canada, the Canadian Breast Cancer Research Initiative, and MFR (the Swedish Medical Research Council).
The abbreviations used are: IGF, insulin-like growth factor; IGFBP-3, insulin-like growth factor-binding protein-3; BMI, body mass index; OC, oral contraceptive.
C. Deal, J. Ma, F. Wilkin, J. Paquette, F. Rozen, T. Hudson, M. Stampfer, M. Pollak. J. Clin. Endocrinol. and Metabolism, in press, 2001.
We would also like to thank Dr. Phillipa Holowaty, Elizabeth Hoodfar, Kelly Metcalf, Jalil Hakimi, Caitlin Springate, Danielle Hanna, Minnie Ho, Patricia de los Rios, and Amy Finch for their contributions to this research.