TGFβ-Induced Factor: A Candidate Gene for High Myopia

Abstract

purpose. To investigate the coding exons of transforming growth factor (TGF)-β–induced factor (TGIF) for mutations in Chinese patients with high myopia.

methods. Seventy-one individuals with high myopia of −6.00 D or less and 105 control subjects were screened by DNA sequencing for sequence alterations. Univariate analysis and logistic regression were performed to identify single-nucleotide polymorphisms (SNPs) and their interactions in TGIF that may be associated with myopia.

results. Six SNPs showed a significant difference (P < 0.05) between patient and control subject in univariate analysis. Four of them cause codon changes: G223R, G231S, P241T, and A262G. Among all the SNPs that entered multivariate analysis, only 657(T→G) showed statistical significance in the logistic regression model (odds ratio 0.133; 95% confidence interval 0.037–0.488; P = 0.002).

conclusions. TGIF is a probable candidate gene for high myopia. Further studies are needed to identify the underlying mechanism.

Myopia is the most common eye disorder worldwide. In the United States, the prevalence of myopia among blacks or whites is approximately 25%, ¹ whereas in Asian populations such as Chinese or Indians, the prevalence may exceed 65%. ² A pilot study showed the prevalence of myopia among Hong Kong Chinese adults was approximately 40%. ³ The early onset and fast progression of myopia among Chinese children is also well documented. ^{4 5 6} There is thus evidence to suggest a higher prevalence of myopia in Chinese than white populations. 

Myopia is a complex disease involving multiple interacting genetic and environmental factors. Environmental factors, such as educational level, occupation, and individual income, have been associated with the prevalence of myopia. ⁷ Other personal factors, such as reading habits and use of computers, may also affect the progression to high myopia. ^{8 9} Although no candidate gene for myopia has been identified, there are several lines of evidence to support a genetic basis. For example, children of parents with myopia are more likely to have myopia than children of those without myopia. ^{10 11 12 13} Evidence of significant linkage was found on 12q21-23 and 18p11.31 based on family linkage studies of autosomal dominant high myopia. ^{14 15} We also found a maximum LOD score of 3.3 between markers D18S476 and D18S62 on 18p11 in six families of high myopes. ¹⁶ The transforming growth factor (TGF)-β–induced factor (TGIF), which was mapped to 18p11.3, ¹⁷ is one of the identified genes within the reported region of 18p. 

TGIF is a DNA-binding homeodomain protein that belongs to the TALE family homeobox. It is a transcription repressor that acts in various ways. ¹⁸ It has been demonstrated to play a role in TGF-β–regulated transcription ¹⁹ and control of retinoid-responsive transcription. ²⁰ Mutation in TGIF has been found to be associated with holoprosencephaly, a prevalent congenital disorder of brain and craniofacial malformation. ²¹ It is a structural defect of the developing human forebrain and midface affecting 1 in 250 conceptuses and approximately 1 in 16,000 live births. ²² In our genomic search for the myopia locus in two Chinese families with high myopia, we found a possible link with a region in 18p11. ¹⁶ Consistent association of this locus with high myopia has been reported in an Italian population. ²³ In this study, we screened the whole TGIF gene for single-nucleotide polymorphisms (SNPs) that may be associated with myopia. 

Methods

Subjects

Seventy-one unrelated Chinese subjects with high myopia of −6.00 D or less in both eyes were recruited at the Hong Kong Eye Hospital, Hong Kong. Subjects with myopia had a mean age of 39.52 ± 16.09 years, and a male-to-female ratio of 1.0:1.8. One hundred five unrelated control subjects were recruited from patients who attended the hospital for conditions other than high myopia. They and their family members did not have eye diseases except senile cataract, floaters, and itchy eyes. The control subjects had no or mild myopia, with refractive errors of −2.00 D or more in either eye. They had a mean age of 66.98 ± 9.72 years and a male-to-female ratio of 1.0:2.0. All patients and control subjects involved in this study were similar in social background and were from the local ethnic Han Chinese population, with no ethnic subdivision. All study subjects were given a complete ocular examination, and venous blood was collected and stored at −20° for less than 2 months before extraction of DNA. The study protocol was approved by the Ethics Committee for Human Research, the Chinese University of Hong Kong, and adhered to the tenets of the Declaration of Helsinki. Informed consent was obtained from the study subjects after an explanation of the nature and possible consequences of the study. 

Mutation Analysis

Genomic DNA was extracted from 200 μL of whole blood with a kit (Qiamp; Qiagen, Hiden, Germany). Sequence alterations in all coding exons were detected by PCR, followed by confirmation-sensitive gel electrophoresis (CSGE) ²⁴ and direct sequencing. ²⁵ The coding sequence was amplified by primer pairs reported previously. ²⁶ Sequencing was performed with an automated DNA sequencer (Big Dye Terminator; ABI 377-XL; Applera Corp., Foster City, CA). 

Statistical Analysis

Either the χ² tests or the Fisher exact test was used to compare the allele frequencies of sequence alterations in patients and control subjects. SNPs with P < 0.20 were selected for the logistic regression analysis. In contrast to univariate analysis, which considers only one SNP at a time, logistic regression analysis is a multivariate analysis method that deals with all selected SNPs as a whole in the same model. All variables were considered to be categorical in this analysis. The dependent variable was disease status (patient, 1; control subject, 0), and the independent variables were SNPs (homozygote, 2; heterozygote, 1; wild type, 0). The logistic regression model was optimized by using a backward approach. The significance of interactions between SNPs was estimated by adding corresponding interaction items in the model. An optimal model was established when all the independent variables (SNPs) were significant (P < 0.05) in the model. Allelic frequencies of all detected SNPs were also assessed for Hardy-Weinberg equilibrium. Statistical analyses were performed on computer (SPSS software, ver. 10.1; SSPS Science, Chicago, IL). 

Results

Twenty-five SNPs were identified exclusively within exon 3 (555–1130 bp) of TGIF (mRNA, GenBank accession number: NM_003244 http://www.ncbi.nlm.nih.gov/Genbank; provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD) (Table 1) . All SNPs existed in both patients and control subjects. The prevalence of homozygous and heterozygous changes was higher than normal in all except one of the SNPs, 488(C→T). Six SNPs were significant (P < 0.05) in univariate analysis, and four of them caused codon changes. The sequence changes were 667(G→A), 691(G→A), 721(C→A), and 785(C→G), and the corresponding codon changes were G223R, G231S, P241T, and A262G. Ten SNPs with P < 0.20 between patients with myopia and control subjects were input to logistic regression analysis. Among them only 657(T→G) showed a significant result in the regression model (odds ratio 0.133; 95% confidence interval 0.037–0.488; P = 0.002). 

Discussion

In the present study, an elderly control group was selected on the basis that they were eye-disease free with an especially low likelihood for the development of high myopia. There is a possibility that the SNPs that are significantly associated with myopia alter survival in the long term and create a segregation in SNP frequencies among patient and control groups. However, the deterioration of eyesight essentially only causes inconvenience in daily life that can be easily alleviated by using glasses. It is not likely that myopia affects survival in a developed country in the long term. 

Twenty-five SNPs were identified in TGIF, and they are likely to be population specific, because they were commonly present in both patients and control subjects. We adopted the TGIF sequence from GenBank as reference. A sequence different from the reference was regarded as a sequence alteration. This may not be the most appropriate way to define normal and changed alleles, because polymorphism may be population specific. This is likely in our patients, in that 96% of the SNPs—that is, all but one—had a higher prevalence in changed alleles than in reference alleles. 

Some of the SNPs found in our study were not at Hardy-Weinberg equilibrium (Table 1) . It may be that those SNPs are fresh in our study population and had not yet reached Hardy-Weinberg equilibrium. However, this would not necessarily produce population stratification and thus would affect the validity of our results. First, most of the SNPs (four of six) that showed association with myopia were at Hardy-Weinberg equilibrium in the control group: 573(G→A), 657(T→G), 667(G→A), and 721(C→A). The other two myopia-associated SNPs, 691(G→A) and 785(C→G), were not at Hardy-Weinberg equilibrium in both the control group and in the patients (Table 1) . Thus, the effects of Hardy-Weinberg disequilibrium on patients and control subjects should be canceled out and, accordingly, should not produce population stratification. The second reason is that most of the SNPs (five of seven) that showed Hardy-Weinberg disequilibrium both in patients and control subjects were not associated with myopia. They were 726(C→T), 741(C→T), 791(T→C), 792(G→A), and 804(A→G). Meanwhile, there were more SNPs (18/25 among control subjects and 16/25 among patients) that were at Hardy-Weinberg equilibrium than those that were not. For validation of our findings we repeated the genotyping practical analysis several times, and consistent results were obtained. Our sequences were mostly clean at baseline. Most important, the mutated peaks in the heterozygous conditions were much higher than the normal peaks, which clearly indicated the presence of heterozygous mutation. Therefore, occurrence of genotyping errors in this study was kept to a minimum. 

In univariate analysis, we classified each SNP into three categories (genotype 1/1, 1/2, and 2/2, respective to normal, heterozygous, and homozygous alterations), instead of two categories that merely indicate whether the sample carries sequence change. This classification should address the disease association more precisely, because it considers heterozygous and homozygous as different categories. Our data showed significant results after making such a classification. 

Four SNPs: 667(G→A), 691(G→A), 721(C→A), and 785(C→G), with corresponding codon changes G223R, G231S, P241T, and A262G, showed a significant difference between patient and control (P < 0.05). The first three sequence changes is likely to be a protective factor, because control samples tended to have a higher prevalence of this homozygous alterations than did patient samples—although for 785(C→G) it may also be protective, because the heterozygous prevalence in the control samples was relatively higher. 

The traditional uninformative polymorphism has been shown to carry important information about the responsiveness and expression levels of receptors when they are arranged in phase-known haplotype format. ^{27 28} Although haplotype analysis is informative, its drawback is that it is extremely laborious and is still not routinely practicable. ²⁹ We attempted to analyze multiple SNPs together by logistic regression, which may reveal more information than can be delivered by univariate analysis. The logistic regression is applied to analyze the association of various SNPs and disease pathogenesis as a whole. The model suggested the risk for high myopia was reduced 86.7% in subjects with the 657GG genotype. Although 657(T→G) does not account for codon change, it may be that silent SNPs have some effect on the structural folding, stability, and degradation of the mRNAs and in turn may affect the efficiency of protein translation. ³⁰ These also affect the final protein function. 

The five mutations mentioned earlier were present in the carboxyl terminus of TGIF. This domain represses transcription in a histone deacetylase (HDAC)–dependent manner. ¹⁸ These mutations may alter TGIF’s binding efficiency to HDAC and thus affect the transcription regulation. TGIF has been proposed to have influence on the developmental program by altering Nodal/TGF-β signaling. Its interaction with repressor proteins through other DNA-binding proteins or direct DNA binding to cognate sites contributes to regulation of the TGFβ transcription pathway. ³¹ It interacts with SMAD2 and, specifically, with midline brain structures. ²⁴ It is possible that mutations in TGIF also alter its function and hence the phase of eye development that involves growth of the sclera. In conclusion, our study demonstrated TGIF to be a possible candidate gene for high myopia. We are in the process of a large-scale screening for TGIF mutations in persons with high myopia. Other environmental factors, such as reading habits, use of computers, and education background that may contribute to the progress of myopia will also be thoroughly investigated. 

 Supported by the T. K. Lam Foundation and a block grant from the Chinese University of Hong Kong.

 Submitted for publication January 18, 2002; revised May 20 and August 15, 2002; accepted September 20, 2002.

 Commercial relationships policy: N.

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Table 1.

 

 

Genotype Frequencies and Probabilities of TGIF Sequence Alterations in Patients with Myopia and Control Subjects

Table 1.

 

Genotype Frequencies and Probabilities of TGIF Sequence Alterations in Patients with Myopia and Control Subjects

Sequence Alterations	Genotype Frequency (%)															P (χ2 Test/Fisher Exact Test)
		Patients (n = 71)							Control Subjects (n = 105)
		1/1	1/2	2/2	HWDχ2	1/1	1/2	2/2	HWDχ2
420(A→G)	45	48	7	0.53	51	43	6	0.38	0.70
488(C→T)	62	32	6	0.097	70	29	1	0.74	0.15, †
573(G→A)	0	61	39	9.53*	1	38	61	2.60	0.005, † , ‡
576(C→T)	0	3	97	0.014	0	4	96	0.039	1.00
607(A→G)	0	4	96	0.031	0	4	96	0.039	1.00
612(G→A)	1	24	75	0.041	0	14	86	0.58	0.088, †
618(A→C)	0	15	85	0.46	1	15	84	0.037	1.00
624(C→T)	0	18	82	0.67	0	12	88	0.43	0.28
644(A→G)	0	6	94	0.062	0	9	91	0.20	0.46
656(C→T)	0	14	86	0.38	0	8	92	0.16	0.16, †
657(T→G)	0	17	83	0.55	0	3	97	0.02	0.001, † , ‡
667(G→A)	1	59	40	6.72*	1	41	58	3.37	0.027, † , ‡
678(G→C)	0	0	100	0.00	0	0	100	0.00	1.00
691(G→A)	1	71	28	11.9*	1	50	49	6.69*	0.009, † , ‡
721(C→A)	1	14	85	0.21	0	5	95	0.06	0.027, † , ‡
722(C→T)	0	4	96	0.031	0	3	97	0.02	0.69
726(C→T)	0	73	27	15.67*	0	67	33	18.00*	0.35
741(C→T)	1	99	0	43.99*	0	99	1	66.64*	0.65
763(G→A)	0	14	86	0.38	0	8	92	0.16	0.16, †
782(C→T)	0	7	93	0.094	0	10	90	0.31	0.44
785(C→G)	27	70	3	10.32*	9	88	3	36.14*	0.007, † , ‡
791(T→C)	1	95	4	35.11*	0	96	4	57.94*	0.67
792(G→A)	0	86	14	25.45*	0	80	20	30.00*	0.31
802(G→C)	0	8	92	0.14	0	15	85	0.66	0.18
804(A→G)	3	84	13	20.89*	2	88	10	37.13*	0.75