Figure 1. The linkage disequilibrium (LD) pattern for single-nucleotide polymorphisms in the IGF1 gene for the Chinese participants (cases and controls) of our study (A) and the Han Chinese (B) and whites (C) of the HapMap Project. The LD measures are indicated as the correlation coefficient (r2). The shades of gray indicate the magnitudes of the r2 values, and black is equal to 100% or 1.00, which is not indicated in the diagram to avoid cluttering. Note that, on the basis of the definition in Gabriel et al,38 no LD block can be constructed in this region for the Chinese participants of our study or for the whites of the HapMap Project. CEU indicates Utah residents with Northern and Western European ancestry from the Centre d'Etude du Polymorphisme Humain collection; CHB, Han Chinese in Beijing, China; kb, kilobase.
Figure 2. The linkage disequilibrium (LD) pattern for single-nucleotide polymorphisms in the IGFBP3 (A) and IGFBP4 (B) genes for case participants and control participants. The LD measures are indicated as the correlation coefficient (r2). The shades of gray indicate the magnitudes of the r2 value. Note that, on the basis of the definition in Gabriel et al,38 no LD block can be constructed in these 2 regions.
Mak JYY, Yap MKH, Fung WY, Ng PW, Yip SP. Association of IGF1 Gene Haplotypes With High Myopia in Chinese Adults. Arch Ophthalmol. 2012;130(2):209-216. doi:10.1001/archophthalmol.2011.365
Author Affiliations: Department of Health Technology and Informatics (Mss Mak and Fung, Mr Ng, and Dr Yip) and Centre for Myopia Research, School of Optometry (Dr Yap and Mr Ng), The Hong Kong Polytechnic University, Hong Kong Special Administrative Region.
Objective To investigate the association of high myopia with common single-nucleotide polymorphisms (SNPs) in the IGF1, IGFBP3, and IGFBP4 genes in a Chinese population.
Methods For our case-control study, we recruited 600 unrelated participants: 300 case participants with high myopia (−8.00 diopters or less) and 300 emmetropic controls (within ±1.00 diopter). Twenty-one tag SNPs were selected from these candidate genes and were genotyped. Genotype data were analyzed by logistic regression. Multiple comparisons were corrected by running 15 000 permutations of case-control status.
Results Although we did not find any significant association for IGF1 SNPs using single-marker analysis, we identified many windows with haplotype frequencies significantly different between case participants and control participants using a variable-sized sliding-window strategy. In particular, the most significant association was shown by a 3-SNP window consisting of rs12423791, rs7956547, and rs5742632 (omnibus test: asymptotic Pasym = 3.70 × 10−9 and empirical Pemp = 6.67 × 10−5). There were 3 high-risk haplotypes: CAC (Pasym = 2.35 × 10−6; odds ratio [OR], 5.06), GAT (Pasym = 3.56 × 10−4; OR, 3.18), and GGC (Pasym = 2.25 × 10−3; OR, 25.10). There was 1 protective haplotype: GAC (Pasym = 8.43 × 10−4; OR, 0.63). On the other hand, our single-marker and haplotype analyses did not show any significant association for IGFBP3 and IGFBP4.
Conclusions IGF1 haplotypes are associated with genetic susceptibility to high myopia in Chinese adults, whereas IGFBP3 and IGFBP4 are unlikely to be important in the genetic predisposition to high myopia.
Clinical Relevance IGF1 is associated with high myopia, and identifying the causal variants is important for understanding the underlying mechanisms.
Myopia is a complex disease caused by multiple genetic and environmental factors and, possibly, by the interaction of these factors.1,2 It is the principal cause of vision impairment and may affect approximately 2.5 billion people worldwide by 2020.3 The prevalence is particularly high in Asian urban populations such as in Singapore, Hong Kong, and Taiwan.4- 6 High myopia is usually defined as a refractive error of −6.00 diopters (D) or less, and it is associated with an increased risk of blinding disorders like cataract, glaucoma, and retinal detachment.7
An elongated eyeball with a thinner fibrous sclera accounts for the majority of cases of high myopia.8,9 Proteoglycans are important components in the scleral extracellular matrix, and hence alteration in their synthesis may affect scleral remodeling and contribute to the development of myopia.10 The retina is the source of ocular growth–regulating signals,11 whereas the retinal pigment epithelium (RPE) is intimately involved in conveying the retinal growth signals to the choroid and the sclera.12 Therefore, molecular signals or growth factors that affect the physiology of the RPE might contribute to the development of myopia.
Insulin-like growth factor 1 (IGF1), insulin-like growth factor binding protein 3 (IGFBP3), and insulin-like growth factor binding protein 4 (IGFBP4) are members of the human growth hormone–insulin-like growth factor pathway that plays a key role in growth and metabolism. Insulin-like growth factors are structurally and functionally related to insulin, and their biological functions are regulated by insulin-like growth factor binding proteins. Individuals with a high-carbohydrate intake have tended to be more myopic, and a high glycemic load carbohydrate diet might induce permanent changes in the development and progression of refractive errors.13,14 The IGF1 gene is located on chromosome 12q23.2 and is thus within the MYP3 interval that is mapped for autosomal dominant high-grade myopia.15 Intriguingly, intravitreal injection of IGF1 in chicks increased ocular growth and elongated the axial length.16 In addition, IGF1 can also regulate scleral proteoglycan production,10 perhaps by upregulating gene transcription, translation, or activation of sulfotransferases and, in turn, increasing the synthesis of sulfated proteoglycans (the most abundant type of proteoglycan in the sclera).17 Therefore, it might influence scleral remodeling and myopia development. IGFBP3 prevents IGF1 from binding to its receptor and hence is antagonistic to IGF1. The combination of a reduced IGFBP3 level and an accompanying elevated free IGF1 level in scleral tissue has been proposed to enhance scleral tissue growth and lead to myopia development.13,18 In addition, the RPE expresses receptors for IGF1 and secretes both IGF1 and IGFBP3.19 In turn, IGF1 stimulates gene expression in RPE cells,20 and thus the autocrine/paracrine function of IGF1 and its associated binding proteins may play a role in the RPE physiology and contribute to myopia genesis. Moreover, endogenously expressed IGFBP3 may also regulate retinal endothelial cell behavior.21 On the other hand, proteolysis of IGFBP4 is important in the regulation of IGF1 action,22 and both are also expressed in the human sclera.23 The IGFBP3 gene is located on chromosome 17q12-q21.1, very close to the MYP5 interval for the autosomal dominant high myopia locus.24
Moreover, the association between IGF1 single-nucleotide polymorphisms (SNPs) and high myopia was recently reported in an international cohort of 265 white multiplex families with 1391 participants.25 This study identified 3 SNPs (rs6214, rs10860860, and rs2946834) associated with the myopia phenotype. Apart from the IGF1 gene itself, the MYP3 locus where IGF1 resides has been studied intensively. Its linkage with high myopia or refractive error has been replicated repeatedly.15,26- 29 Many candidate genes within the MYP3 interval have also been examined for an association with high myopia, with both positive and negative results.30,31 In our study, we used a case-control study design32 to examine the relationship between high myopia and the SNPs of these 3 candidate genes (IGF1, IGFBP3, and IGFBP4) in the Chinese population of Hong Kong.
Participants were unrelated Han Chinese adults (18-45 years of age) recruited through the use of posters placed in the Optometry Clinic in the Hong Kong Polytechnic University and throughout the university campus that promoted the study, through the use of visual screening activities outside the campus, and through referrals of myopic individuals from local optometrists. The recruitment criteria were a spherical equivalent of −8.00 D or less for both eyes for high myopes (case participants) and a spherical equivalent within ±1.00 D for both eyes for emmetropes (control participants). In total, 300 case participants and 300 control participants were recruited. Participants were excluded if they showed symptoms of obvious ocular diseases (other than myopia-associated retinal changes), had known inherited disorders (eg, Stickler syndrome or Marfan syndrome) with myopia as one of the typical clinical features, or had a past history of ocular trauma. Our study was approved by the Human Subjects Ethics Subcommittee of the Hong Kong Polytechnic University and adhered to the tenets of the Declaration of Helsinki. All participants gave their written informed consent. Details of the eye examination, blood collection, and DNA extraction from blood leukocytes have been reported elsewhere.33
We used the Tagger software34 to select tag SNPs from the Chinese data of the International HapMap Project (release 24, phase II, Nov08 [ http://hapmap.ncbi.nlm.nih.gov/]) with the following criteria: pairwise tagging algorithm, r2 ≥ 0.8, and a minor allele frequency of at least 0.10 for each candidate gene and its flanking regions (3 kilobases [kb] upstream and 3 kb downstream). In total, 21 tag SNPs were selected: 10 from IGF1, 5 from IGFBP3, and 6 from IGFBP4. The SNPs were genotyped by restriction fragment length polymorphism analysis. Two exceptions were rs12423791 and rs6539035 in the IGF1 gene, which were genotyped by unlabeled probe melting analysis.35 Details of primer sequences and reaction conditions are provided in the eTable and eMethods.
We analyzed ocular data with Stata version 8.2 (StataCorp) and genotype data with PLINK version 1.07 (http://pngu.mgh.harvard.edu/~purcell/plink/) and Haploview version 4.2 (http://www.broadinstitute.org/scientific-community/science /programs/medical-and-population-genetics/haploview /haploview).36,37 With the exact test implemented within PLINK, we tested genotypes in case participants and control participants separately for Hardy-Weinberg equilibrium. Haploview was used to calculate the correlation coefficient (r2) as a measure of linkage disequilibrium (LD) between pairs of SNPs and to construct an LD block based on the definition by Gabriel et al.38 Association analysis of single markers and haplotypes was performed with PLINK using logistic regression. In particular, we adopted a sliding-window strategy to exhaustively analyze all possible haplotypes consisting of a variable number of consecutive SNPs. We performed a single case-control omnibus test for each sliding window to jointly evaluate the significance of the haplotype effects for this sliding window; this test has (h − 1) degrees of freedom, where h is the number of haplotypes for the window of interest. For a given window size within a gene under consideration, the test was performed for all possible windows of the same size, shifting one SNP at a time toward the 3′ end of the gene. Individual tests were assessed by asymptotic P values (Pasym). “Multiple hypotheses” testing was corrected by running 15 000 permutations to generate empirical P values (Pemp). In each permutation, the case-control status of the participants was permuted without altering the genotypes across all single markers and all haplotypes. Odds ratios (ORs) were also calculated for single markers, with the major allele as the reference, and for specific haplotypes, with the group of all other haplotypes as the reference, as appropriate. Note that PLINK gives 95% CIs for ORs for single markers but not for haplotypes. For sliding windows showing significant association with high myopia, conditional logistic regression analysis implemented within PLINK was used to identify the SNPs contributing independent effects to the association and to identify a subset of SNPs that could explain the association for haplotypes composed of a larger set of SNPs.
The ocular data of the 300 case participants with high myopia and the 300 emmetropic control participants have been reported elsewhere33 and are briefly summarized herein. The mean (SD) spherical equivalent and axial length were −10.53 (2.48) D and 27.76 (1.13) mm, respectively, for the case participants and 0.03 (0.43) D and 23.85 (0.83) mm, respectively, for the control participants. These are the ocular data for the right eyes. The case participants were, on average, older than the control participants (27.6 vs 24.6 years; P < .001, determined by use of the t test). There were fewer men in the case group than in the control group (27.7% vs 43.7%; P < .001, determined by use of the χ2 test).
For the sake of easy reference and discussion, the SNPs were also designated as IGF1.S1 to IGF1.S10 for IGF1, as IBP3.S1 to IBP3.S5 for IGFBP3, and as IBP4.S1 to IBP4.S6 for IGFBP4 in the sequential order from the 5′ end to the 3′ end of the sense strand of the respective genes (Table 1). The genotypes were in Hardy-Weinberg equilibrium (P >> .05) for all SNPs in the case group and in the control group separately. The only exception was rs12579077 (IGF1.S1) in case participants (P = .01). This SNP was included in association analysis because a departure from Hardy-Weinberg equilibrium in case participants can be a sign of marker-disease association.39 The extent of LD among SNPs was, in general, quite weak for SNPs in the respective gene locus, and there was no LD block constructed on the basis of the definition by Gabriel et al38 (Figures 1A and 2).
Allelic association analysis showed that 2 SNPs were associated with high myopia: rs2270628 (IBP3.S5; Pasym = .004; OR, 1.49) and rs535058 (IBP4.S3; Pasym = .03; OR, 1.33) (Table 1). However, both did not survive correction for multiple comparisons (the Pemp values being .08 for IBP3.S5 and .53 for IBP4.S3). In summary, we did not find a significant association with high myopia for all 21 SNPs under study using single-marker analysis (Table 1).
For haplotype analysis, there were 91 sliding windows in total for the 3 genes under study: 55 for IGF1 at 12q23.2, 15 for IGFBP3 at 7p13-p12, and 21 for IGFBP4 at 17q12-q21.1 (Table 2). For IGFBP3 and IGFBP4, no haplotypes of any window size were found to be associated with high myopia. However, there were 26 sliding windows showing significant differences (Pemp < .05) in IGF1 haplotype frequencies between case participants and control participants with at least 1 significant window among sliding windows of a given size with at least 2 SNPs (Table 2). The importance of IGF1.S4 (rs12423791) was obvious because all 26 “positive” windows contained this SNP. The 3-SNP window IGF1.S4..IGF1.S6 (ie, rs12423791, rs7956547, and rs5742632; Table 2) gave the most significant result: Pasym = 3.70 × 10−9 and Pemp = 6.67 × 10−5 for 15 000 permutations. Our conditional logistic regression analysis indicated that each of these 3 SNPs contributed independent effects to the significant association between this 3-SNP window and high myopia: P = 1.47 × 10−13 for IGF1.S4, P = 5.60 × 10−7 for IGF1.S5, and P = 9.09 × 10−13 for IGF1.S6 (Table 3). Of the 26 positive sliding windows, 20 windows carried these 3 SNPs (Table 2). Among the 20 windows carrying IGF1.S4..IGF1.S6, these 3 SNPs could fully explain the significant association between the haplotypes and high myopia in 15 windows (Table 3). For windows with 4 SNPs or more, this was demonstrated by P >> .05 when we performed the logistic regression, controlling for (ie, conditional on) these 3 SNPs (IGF1.S4, IGF1.S5, and IGF1.S6). This means that, once these 3 SNPs were included in the models, the effects of other SNPs in the windows became insignificant.
There were 3 high-risk IGF1.S4..IGF1.S6 (ie, rs12423791, rs7956547, and rs5742632) haplotypes: CAC (112; OR, 5.06), GAT (211; OR, 3.18), and GGC (222; OR, 25.10), for which alleles 1 and 2 stand for the major and the minor alleles, respectively (Table 4). For the first 2 haplotypes (CAC and GAT), the frequencies were about 8% to 10% in the case group and 2% to 3% in the control group. For the GGC haplotype, the frequencies were about 2.5% in the case group and less than 0.5% in the control group. There was only 1 protective haplotype: GAC (212; OR, 0.63), which was found in about 21% of case participants and 30% of control participants.
These results were based on analysis that did not take sex and age as covariates into account, even though there were significant differences in the proportion of male participants and in the mean age between case participants and control participants. When sex and age were included as covariates in the analysis, the overall conclusion remained the same, with slight variation in the actual ORs and P values (data not shown).
There is both direct evidence and indirect evidence suggesting that IGF1, IGFBP3, and IGFBP4 may be involved in myopia development.13- 24 We used a case-control study to explore the role of the common SNPs of these 3 candidate genes in the genetic susceptibility to high myopia in a Chinese population. We analyzed the relationship between the case-control status and the genotype data with logistic regression. Our initial single-marker analysis did not find any significant differences in allele frequencies between the 300 case participants with high myopia (−8.00 D or less) and the 300 emmetropic control participants for all 21 tag SNPs selected from these 3 candidate genes (Table 1).
We adopted a variable-sized sliding-window strategy to evaluate the haplotypic effects thoroughly. This strategy has been proven to be more powerful than single-marker analysis and LD-block–based haplotype analysis, particularly in genomic regions with low LD as in these 3 gene loci (Figure 1).40 The SNPs of the IGFBP3 and IGFBP4 genes still did not exhibit any haplotypic effects with this comprehensive approach (Table 2). However, 26 of 55 possible sliding windows in the IGF1 gene showed significant differences in haplotype frequencies between case participants and control participants (Table 2). The 3-SNP window rs12423791-rs7956547-rs5742632 (ie, IGF1.S4..IGF1.S6) gave the most significant result, and the 3 constituent SNPs each contributed independent effects to the haplotypic association (Table 3). Our conditional logistic regression further showed that these 3 SNPs could fully account for the significant association for the majority (75%) of positive windows harboring these SNPs. There were 3 high-risk haplotypes and 1 protective haplotype defined by these 3 SNPs (Table 4). These 3 intronic SNPs do not seem to have any biological functions as predicted using the Web-based tool FuncPred (http://snpinfo.niehs.nih.gov/snpfunc.htm). Taken together, this indicates that the causal variants driving the significant results are in strong LD with these haplotypes or are carried by chromosomes defined by these associated haplotypes.
As was mentioned in the introduction, an association between IGF1 SNPs and high myopia was also recently reported in an international cohort of 265 white multiplex families with 1391 participants.25 This study defined myopia with spherical refraction or spherical equivalent. Myopia was defined as −0.50 D or less and high myopia as −5.00 D or less. Tag SNPs were selected with the criteria of r2 >> 0.67 and a minor allele frequency of greater than 0.05 from the HapMap data on whites. Of the 13 SNPs examined, 3 showed significant association after correction for multiple testing: rs6214 and rs10860860, with high myopia and any myopia status, and rs2946834, with any myopia status only. Intriguingly, rs6214 (IGF1.S9) was also investigated in our study but did not show any significant association with high myopia (Table 1). Also, rs6214 did not contribute any independent effect to any of the 26 positive windows (data not shown). This SNP is found at the 3′ untranslated region of the IGF1 gene and is in very weak LD with most neighboring SNPs. It has similar minor allele frequencies (~45%) for both Chinese and whites in the HapMap database. Nonreplication of this SNP in our study remained to be explained, although sample size, case definition, and variation of fine-scale LD pattern in different populations may contribute to this discrepancy. For the 10 tag SNPs under study, the overall LD patterns for the Chinese participants of our study and the Han Chinese of the HapMap Project are very similar, although the exact r2 values vary between these 2 groups of Chinese participants. The LD patterns vary to a greater extent between Chinese participants and white participants (Figure 1), and the variation is more obvious when the LD patterns are examined for all SNPs genotyped in the HapMap Project (eFigure).
What is more interesting is the association of the other 2 SNPs (rs2946834 and rs10860860) with myopia status (high myopia and/or any myopia) in whites.25 These 2 SNPs are in the 3′ flanking region of the IGF1 gene: rs2946834 is about 5.8 kb downstream of rs6214 (IGF1.S9), and rs10860860 is about 12.7 kb downstream of rs6214. According to HapMap Chinese data (release 24, phase II), rs2946834 is in strong LD with rs5742632 (IGF1.S6; r2 = 0.773), and rs10860860 is also in strong LD with rs7956547 (IGF1.S5; r2 = 0.835). Note that the intronic IGF1.S6 is about 69 kb upstream of rs2946834, and the intronic IGF1.S5 is about 78 kb upstream of rs10860860. We also note that IGF1.S5 and IGF1.S6 each contribute an independent effect to the most significant association between haplotypes (IGF1.S4..IGF1.S6) and high myopia in the present study (Table 3). In other words, the LD pattern among these SNPs implies that our significant haplotype association with high myopia in a Chinese population is compatible with the significant single-marker association with myopia status in whites. This parallel finding is a further argument that the causal variants driving the association are not the significant SNPs or haplotypes themselves, but some untyped variants in strong LD with these markers. Because the study25 of white multiplex families pinpoints the positive signal to the 3′ flanking region of the IGF1 gene, we explore the IGF1 3′ flanking region with the Web-based program TargetScan version 5.1 (http://www.targetscan.org/).41 TargetScan predicts many binding sites for microRNAs in this region. MicroRNAs are small noncoding RNA molecules that usually bind to the 3′ end of messenger RNA molecules and regulate the expression of these transcripts.42 Interestingly, some SNPs in the 3′ untranslated region of IGF1 are in strong LD with rs12423791 (IGF1.S4) or rs7956547 (IGF1.S5), and are predicted (using FuncPred) to affect the binding of microRNAs. Therefore, these potentially functional SNPs are worthy of further investigation.
In conclusion, we examined the relationship between high myopia and common polymorphisms in the IGF1, IGFBP3, and IGFBP4 genes. Using single-marker analysis and haplotype analysis, we did not find any significant association of high myopia with IGFBP3 or IGFBP4. However, with haplotype analysis, we found that the IGF1 3-SNP window rs12423791-rs7956547-rs5742632 (ie, IGF1.S4..IGF1.S6) showed the most significant association with high myopia. Therefore, IGF1 polymorphisms are likely to play an important role in the genetic predisposition to high myopia.
Correspondence: Shea Ping Yip, PhD, Department of Health Technology and Informatics, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong Special Administrative Region (firstname.lastname@example.org).
Submitted for Publication: June 21, 2011; final revision received August 26, 2011; accepted August 31, 2011.
Author Contributions: Dr Yip 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.
Financial Disclosure: None reported.
Funding/Support: This study was supported by The Hong Kong Polytechnic University (grants RP2G, J-BB7P, 87MS, and 87LV).
Role of the Sponsor: The funding body did not participate in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.
Additional Contributions: We thank all the participants who were in our myopia genetics study.