Family-Based Association Analysis of Hepatocyte
Growth Factor (HGF) Gene Polymorphisms
in High Myopia
Wei Han,1,2Maurice K. H. Yap,2Jing Wang,1and Shea Ping Yip3
PURPOSE. To investigate the association of high myopia with
polymorphisms in the hepatocyte growth factor (HGF) gene, a
potential candidate for myopia development.
METHODS. Single nucleotide polymorphisms (SNPs) were
screened and identified in the HGF gene region with denatur-
ing high-performance liquid chromatography, and their linkage
disequilibrium pattern was established in a Han Chinese pop-
ulation (n ? 150). Tag SNPs were selected and genotyped
using restriction digestion and fluorescence polarization assays
for 128 nuclear families with 133 severely myopic (mean
spherical equivalent [MSE] ? ?10.0 D) offspring. A family-
based association study was performed using FBAT and GenAs-
soc (Cambridge University, Cambridge, UK).
RESULTS. Of three tag SNPs (HGF5-5b, HGFe9, and HGFe10b)
selected for association study, HGF5-5b, located in the up-
stream region, was found to be associated with high myopia
considered as a quantitative trait (MSE) in additive, dominant,
and recessive models (P ? 0.0157, 0.0108, and 0.0108, respec-
tively). The genotype relative risk was 2.19 for the genotype
C/T, and 2.14 for T/T with reference to C/C of HGF5-5b.
Significantly reduced transmission was demonstrated for the
haplotypes C-A-C (HGF5-5b, HGFe9, and HGFe10b; P ?
0.0031) and C-A (HGF5-5b and HGFe9; P ? 0.0015) in the
recessive model, whereas significantly increased transmission
was found for haplotype T-C (HGF5-5b and HGFe10b; P ?
0.0040) under the dominant model. Preferential transmission
of haplotypes remained significant even after correction for
multiple comparisons. Analysis gave similar results, with myo-
pia considered to be a qualitative trait.
CONCLUSIONS. HGF is a potential locus associated with high
myopia in the Han Chinese population. This is the first study
reporting the association of an HGF gene polymorphism with
high myopia. (Invest Ophthalmol Vis Sci. 2006;47:2291–2299)
populations, and has become a public health problem in mod-
ern society.1,2High myopia, typically in excess of ?6 D,3could
result in severe ocular morbidity, visual impairment, and even
blindness (Chiang LM, et al. IOVS 1993;34:ARVO Abstract
Myopia is a complex trait in which multiple genes, multiple
environmental factors, and their interactions have been all
implicated.5–8The much higher prevalence of myopia in Asian
populations than in white and African populations suggests a
higher genetic susceptibility to myopia in Asian populations.1
Twin studies also have shown high concordance of myopia
traits in monozygotic twins.9,10Although whether low to mod-
erate myopia, also known as school myopia, is genetic is still
controversial,11high myopia seems to have an obvious genetic
background.12Because the first myopia locus was mapped to
Xq28,13many more myopia loci have been found, including
the two latest loci at 2q and 4q.14,15However, the common
nonsyndromic high myopia might be multifactorial or com-
plex.16Genetic association study is currently regarded as the
most powerful approach to mapping the genes underlying
such complex traits.17The transmission disequilibrium test
(TDT)18and its modifications are based on families instead of
unrelated cases and controls and are effective for detecting the
association of traits and disease-susceptibility genes with mod-
est impact on disease.19They are also robust in population
stratification, which may lead to a false-positive association in
conventional case–control studies.
Hepatocyte growth factor (HGF) is an important multifunc-
tional cytokine for cellular scattering and proliferation.20HGF
and its receptor are broadly expressed in the eye and play a
critical role in many ocular physiological and pathologic pro-
cesses.21–23Linkage analysis of mouse eye size showed that the
HGF gene may be a strong positional candidate responsible for
myopia.24Matrix metalloproteinases (MMPs) and tissue inhib-
itors of metalloproteinases (TIMPs) are critical in sclera remold-
ing, which is the characteristic change in axial myopia and is
particularly remarkable in high myopia.25–28HGF was found to
be closely related to the biological activities of MMPs and
TIMPs.22,29–31Moreover, HGF can also induce the expression
of egr-1/ZENK,32,33which was recently found to be associated
with myopia.34Therefore, we hypothesize that the HGF gene
may be a potential candidate susceptibility gene for human
The HGF gene maps to chromosome 7, region q21.1, spans
approximately 70-kb and has 18 exons. In this study, we
identified the single nucleotide polymorphisms (SNPs) within
and around the HGF coding region and established the pattern
of linkage disequilibrium (LD) among the identified SNPs in a
Han Chinese population. SNP markers for association analysis
were then selected on the basis of the LD pattern. Using the
approach of family-based association study, we investigated the
yopia, a very common ocular refractive condition, has a
high prevalence all over the world, particularly in Asian
Hospital, Medical College, Zhejiang University, Hangzhou, China.; the
2School of Optometry, The Hong Kong Polytechnic University, Hung
Hom, Kowloon, Hong Kong SAR, China.; the3Department of Health
Technology and Informatics, The Hong Kong Polytechnic University,
Hung Hom, Kowloon, Hong Kong SAR, China.
Supported by grants from The Hong Kong Polytechnic University
(PolyU) and Zhejiang University (ZU): A-PC13 and G-T538 (internal
competitive grants, PolyU), A362 (Area for Strategic Development for
Myopia Research, PolyU) and 2004A029 (Department of Health Grant,
Zhejiang Province). The Wave DNA Fragment Analysis System was
purchased with a PolyU Large Equipment Grant (G.53.27.9027), and
the Victor3V Multilabel Reader with another Large Equipment Grant
Submitted for publication October 13, 2005; revised February 9,
2006; accepted April 24, 2006.
Disclosure: W. Han, None; M.K.H. Yap, None; J. Wang, None;
S.P. Yip, None
The publication costs of this article were defrayed in part by page
charge payment. This article must therefore be marked “advertise-
ment” in accordance with 18 U.S.C. §1734 solely to indicate this fact.
Corresponding author: Wei Han, Department of Ophthalmology,
The First Affiliated Hospital, Medical College, Zhejiang University,
Hangzhou, China; email@example.com.
1Department of Ophthalmology, The First Affiliated
Investigative Ophthalmology & Visual Science, June 2006, Vol. 47, No. 6
Copyright © Association for Research in Vision and Ophthalmology
genetic association between high myopia and SNP markers of
the HGF gene in a group of Han Chinese nuclear families with
highly myopic sibs.
MATERIALS AND METHODS
Blood samples used in the SNP screening and genotyping stage were
collected as described by Han et al.35In contrast, high-myopia families
were recruited from the Department of Ophthalmology, the First
Affiliated Hospital in Hangzhou, China, with written informed consent
being given. Blood samples were also collected from all recruited
family members. All study subjects were ethnic Han Chinese from
Southern China. The study was approved by the Human Subject Ethics
Subcommittees of the Hong Kong Polytechnic University and Zhejiang
University and adhered to the tenets of the Declaration of Helsinki.
Each nuclear family consisted of two parents and at least one
affected sib with high myopia. Refractive error was determined with
cycloplegic refraction for affected sibs and only noncycloplegic autore-
fraction was used in the parents. For myopic sibs, the entry criterion
was a spherical equivalent of ?10.0 D or worse for both eyes, where
spherical equivalent was calculated as sphere diopters plus half-cylin-
der diopters. Mean spherical equivalent (MSE) of the two eyes for each
sib was used for analysis. All subjects received keratometric measure-
ment with an autorefractor (Humphrey, Carl Zeiss Meditec, Inc., Dub-
lin, CA). Every affected sib also received the measurement of intraoc-
ular pressure with a noncontact tonometer (Reichert Ophthalmic
Instruments, Depew, NY) and corneal curvature (Obscan II; Orbtek,
Bausch & Lomb, Tampa, FL). A-ultrasonography (Scanner A2500;
Sonomed, Lake Success, NY) was used to measure the ocular axial
length (AXL), anterior chamber depth (ACD), and lens thickness (LT).
The earliest time when myopia was first diagnosed was recorded and
used as the surrogate for the onset age of myopia. Myopic sibs were
excluded from the study if they had a premature birth history, early-age
refractive media opacity, known genetic diseases (such as Stickler or
Marfan syndrome) with myopia as one of the presenting features, a
history of ocular trauma, or increased intraocular pressure (?20 mm
Hg). Myopic sibs were also excluded if they had AXL ?26.0 mm or an
average corneal power (CP) of two meridians ?47.0 D in either eye.
DNA from university students and healthy blood donors was extracted
as described previously.35For blood samples collected from high-
myopia families, DNA was extracted (NucleoSpin Blood L kit; Mach-
erey-Nagel, Du ¨ren, Germany), according to the manufacturer’s instruc-
Polymerase Chain Reaction
Forty-four primer pairs were designed with the software Oligo (ver.
6.57; Molecular Biology Insights, Cascade, WA) to amplify the 18 HGF
exons and their immediate flanking regions (within 100 bp), and
noncoding sequences approximately 3.0 kb upstream of the start
codon and 5.0 kb downstream of the stop codon. Primer sequences are
available on request. Touchdown PCR was used to amplify the pooled
or individual DNA samples, as described previously.35In the screening
stage, the same DNA pooling strategy for identifying SNPs was applied
again.35For genotyping purposes, DNA samples were amplified indi-
SNP Identification and Genotyping for Healthy
In the screening stage, the WAVE DNA Fragment Analysis System
(Transgenomic, Omaha, NE) was used to analyze the PCR products.
The details of denaturing high performance liquid chromatography
(DHPLC) analysis for SNP identification were described in our previous
report.35For each identified SNP, 150 Chinese samples were geno-
typed to establish the allele frequencies using the same DHPLC-based
approach.35The software ElDorado (http://www.genomatix.de/ Geno-
matix GmbH, Munich, Germany) was used to search the potential
transcription factor binding sites and the promoter region of the HGF
SNP Genotyping for Myopia Families
Three SNPs (HGF5-5b, HGFe9, and HGFe10b) were selected for asso-
ciation study on the basis of their LD pattern. Because of the logistic
arrangement in the use of instruments in our laboratory, we switched
to different platforms for genotyping DNA samples from myopia fam-
ilies. The SNP HGF5-5b, located within the recognition sequence of
BglII, was genotyped with restriction analysis. A 10-?L reaction mix
consisting of 5 ?L PCR product, 1? NEBuffer 3, and 5 U BglII (New
England Biolabs, Beverly, MA) was prepared and incubated at 37°C for
16 hours. The restriction products were analyzed by agarose gel elec-
The other two SNPs HGFe9 and HGFe10b were genotyped an SNP
detection kit (AcycloPrime-FP; PerkinElmer, Boston, MA) according to
the manufacturer’s instructions. The method was a modification of
template-directed dye-terminator incorporation with fluorescence po-
larization detection.36The sequences of the SNP primers were 5?-
GTTATCGCTATTCTGAGTCCAAAA-3? for HGFe9 and 5?-AAGTCCAAT-
GAATATCAAGGC-3? for HGFe10b. The optimal number of thermal
cycles of the AcycloPrime-FP protocol was 25 cycles for HGFe9 and 35
(Victor3V Multilabel Reader; PerkinElmer), and genotypes were called
automatically using the provided allele-calling software.
LD Analysis of Common SNPs
The genotype data of SNPs with the minor allele frequency (MAF)
?0.10 were input to the software Haploview37(version 3.11; http://
www.broad.mit.edu/mpg/haploview/ provided in the public domain
by the Massachusetts Institute of Technology, Cambridge, MA) which
then performed the Hardy-Weinberg equilibrium test, allelic associa-
tion, and haplotype block analysis. Haploview calculated Lewontin’s
original and standardized LD parameters (D and D?), r2, and the 95%
confidence intervals (CI) of D? (using a bootstrap algorithm). Haplo-
view defined SNP pairs to be in “strong LD” if the upper 95% confi-
dence limit on D? was ?0.98, and the lower confidence limit was
?0.7. The SNP pairs were described as showing “strong evidence of
historical recombination” if the upper 95% confidence limit was ?0.9.
Other SNP pairs were categorized as “uninformative.” A haplotype
block was defined if the outermost pair of SNPs was in strong LD, and
within the block region the number of pairs in strong LD was at least
19 times greater than those in weak LD.38
Statistical Analysis of Ocular Data
The software package SPSS (ver. 11.0; SPSS Inc., Chicago, IL) was used
to test the partial correlation between MSE of the affected sibs and
other ocular components (AXL, ACD, LT, and CP).
Genetic Association Study
Genetic association study was performed using the Family-Based Asso-
ciation Test software package (FBAT, ver. 1.5.5; http://www.biostat.
harvard.edu/?fbat/default.html/ provided in the public domain by
Harvard Medical School, Boston, MA) which is a generalized approach
derived from original TDT method.18FBAT compares the genotype
distribution observed in offspring with its expected distribution.39This
approach allows the user to dictate a genetic model for association
analysis.40Association tests for single loci and haplotypes under addi-
tive, dominant, and recessive models were performed using FBAT with
high myopia being treated as a quantitative or a qualitative trait.41
When high myopia was analyzed as a quantitative trait, MSE was taken
as the measured trait. When high myopia was analyzed as a qualitative
trait, the phenotypes were simplified to a dichotomous trait: affected
2292 Han et al.
IOVS, June 2006, Vol. 47, No. 6
with high myopia (MSE ? ?10.0 D) or unaffected. The parental
affection status, also included in the input pedigree files, did not affect
the FBAT analysis because the high myopia pedigrees in this study
were all nuclear families consisting of two parents and their children
only. The null hypothesis was no linkage and no association, and the
alternative hypothesis was that there was both linkage and association.
The multiple-comparison problems for the alleles of a given marker
were solved by the global statistic for each SNP tested under any given
genetic model. For bi-allelic markers, dominant and recessive models
give reciprocal results and thus are equivalent to one test for the
purpose of accounting for multiple testing. Therefore, for the global
statistics, there were three SNPs, each tested under two different
genetic models (additive and dominant/recessive) and thus six tests of
global association. The more powerful false-discovery rate (FDR)42was
used to control for multiple-hypothesis testing, instead of the conven-
tional Bonferroni adjustment. The FDR is the expected proportion of
the true null hypotheses rejected out of the total number of null
hypotheses rejected. Multiple-comparison procedures controlling FDR
can be regarded as post hoc maximizing procedures, and are more
powerful than the commonly used multiple-comparison procedures
based on the family-wise error rate. After adjustment for multiple
comparisons and with an FDR level of 0.05, the cutoff for significant
global association was 0.0167.
Similarly, the multiple-comparison issues for the haplotypes of two
or three SNPs were solved by the global statistic under any given
genetic model. For the global statistics, there were four groups of
haplotypes each tested under three different genetic models and thus
12 tests of global association. After adjustment for multiple compari-
sons with an FDR of 0.05, the cutoff for significant global association
A matched case–control dataset was generated with each affected
(myopic) sib matched to three possible pseudocontrol subjects created
from the untransmitted parental allele.43,44Conditional logistic regres-
sion was used to analyze this case–pseudocontrol dataset, to calculate
the effect size of the marker genotype on the disease risk as the
genotype relative risk (GRR) and the corresponding 95% CIs. Analysis
was performed with the GenAssoc package (http://www-gene.cimr.
cam.ac.uk/clayton/software/ provided in the public domain by the
Cambridge Institute for Medical Research, University of Cambridge,
Cambridge, UK) and executed within the software (STATA, ver. 8.2;
Stata Corp., College Station, TX).
LD Pattern of the SNPs Identified
Of the total 18 SNPs identified in 11 fragments (Table 1), nine
SNPs were common with MAF ? 0.10 and were selected as the
markers for LD analysis. Two of these nine SNPs were novel.
The Han Chinese population under study was in Hardy-Wein-
berg equilibrium for all common SNPs. According to the defi-
nition of LD blocks proposed by Gabriel et al.,38one block
covering HGFe10a, HGFe10b, HGF3-12a, HGF3-12b, HGF3-
16b, and HFG3-17b was defined in the HGF gene region (Fig.
1). Meanwhile, HGFe8 was also in strong LD with this block of
SNPs. The SNP HGFe9 showed no LD with all other SNPs. The
LD between HGF5-5b and all other SNPs was “uninformative.”
Based on these LD data, three SNPs in HGF gene region—
namely, HGF5-5b, HGFe9, and HGFe10b—were selected as
markers for further association study.
Clinical Myopia Data Analysis
The detailed information of the myopic siblings is listed in
Table 2. The MSE of high myopia offspring is ?12.08 ? 3.37 D
with single eye’s spherical diopter ranging from ?9.5 to ?22.5
D and cylinder diopter from 0 to ?4.25 D. Of the total 128
families (256 parents and 133 highly myopic offspring) studied,
TABLE 1. Details of the SNPs Identified in the HGF Gene in a Han Chinese Population
ID No. (NCBI)
Flanking Sequence (5?33?) and Alleles
?1680delA‡ HGF5-5a5? Flanking region
3? Flanking region
3? Flanking region
NovelAAGGATTAGC[A/–] ATAGAAACGG 0.0033—
n ? 150. MAF, minor allele frequency; HWE, Hardy-Weinberg equilibrium; UTR, untranslated region.
* The SNPs are named in accordance with the recommended nomenclature systems,45whereas alternative designations (based on the names
of the fragments) are also given for the sake of easy discussion in this article. GenBank No. NC_000007 was used as the reference sequence. SNPs
shown in bold were included in LD analysis.
† The 5? and 3? flanking sequences, each 10 bases long, are shown before and after the SNP respectively, with the alleles of the SNP in square
brackets. The major allele refers to the more frequent allele and the minor allele the less frequent allele.
‡These five SNPs were identified only during the stage of genotyping population samples.
IOVS, June 2006, Vol. 47, No. 6
HGF Polymorphisms in High Myopia2293
52 (40.6%) had one myopic parent with both eyes’ spherical
equivalent ? ?0.75 D and 26 (20.3%) had two myopic parents,
whereas 37 (28.9%) had one highly myopic parent with both
eyes’ spherical equivalent ? ?6.0 D, and 4 (3.1%) had two
highly myopic parents. No significant sex association for the
prevalence of high myopia was observed in the myopic sib
group (P ? 0.05). For all affected sibs, corneal curvature
examination (Obscan; Bausch & Lomb) showed no corneal
shape anomaly, such as keratoconus and posterior ectasia.
Partial correlation analysis showed significant correlation be-
tween MSE and ocular refractive components of AXL, ACD, and
CP, with the exception of LT. In agreement with previous
studies,46,47AXL had the strongest correlation (r ? ?0.70; P ?
0.001) to refractive error.
Genetic Association Study
With high myopia as a quantitative trait measured as MSE, no
significant association of the SNPs HGFe9 and HGFe10b with
high myopia was found under all the genetic models tested
(Table 3). In contrast, SNP HGF5-5b showed a significant asso-
ciation under additive and dominant/recessive models with the
quantitative trait MSE (Table 3). The major and minor alleles of
HGF5-5b showed opposite preferential transmission under the
additive model. The minor allele (2 or T) showed significantly
increased transmission under the dominant model (z ? 3.008,
P ? 0.0026), whereas the major allele (1 or C) exhibited
significantly reduced transmission under the recessive model
(z ? ?3.008, P ? 0.0026). It is interesting to note the recip-
the second line (in parentheses) the 95% CI of D?; and the bottom line r2. The cells in black denote complete LD according to the confidence
intervals of D?, the cells in gray the uninformative LD, and the other cells strong historical recombination, as defined by Gabriel et al.38The
diagonal scale shows the sites of each SNP around the HGF gene region.
Pair-wise LD measures of D?, the 95% CI, and r2for common SNPs of the HGF locus. For each cell, the number on the top line is D?;
2294 Han et al.
IOVS, June 2006, Vol. 47, No. 6
rocal relationship when the marker is bi-allelic. The global
statistics (P ? 0.0157 or 0.0108) were significant under the
three models tested. They were still statistically significant,
even on correction for multiple comparisons (n ? 6) based on
For the sake of easy discussion, each haplotype is indicated
by three digits representing the alleles of the SNPs HGF5-5b,
HGFe9, and HGFe10b (in that order), and a zero is inserted
when a particular SNP is not involved. No association of any
haplotypes involving only HGFe9 and HGFe10b (e.g., 0-1-1 and
0-2-1) with the measured trait MSE was found under all three
genetic models tested with FBAT (Table 4). However, signifi-
cant association was demonstrated for haplotypes involving
HGF5-5b (Table 4). When individual haplotypes were consid-
ered, significantly increased transmission (z ? 2 and P ? 0.05)
was found for haplotypes carrying allele 2 of HGF5-5b (2-1-1,
2-1-0, and 2-0-1) under additive and dominant models. When a
group of haplotypes involving two or three SNPs were consid-
ered as a whole, a similar significant association was main-
tained except for haplotypes involving only HGF5-5b and
HGFe9 under the dominant model (P ? 0.0883). The haplo-
type 2-0-1 was still transmitted in significant excess (z ? 3.308,
global P ? 0.0040) to the myopia offspring when the global
statistic was adjusted for multiple comparisons based on FDR.
Significantly reduced transmission (Z ? ?2 and P ? 0.05)
was observed for haplotypes carrying allele 1 of HGF5-5b
(1-1-1, 1-1-0, and 1-0-1) under the recessive model, and haplo-
type 1-1-0 under the additive model, no matter whether the
haplotypes were considered individually or globally with other
haplotypes within the same group (Table 4). With correction
for multiple comparisons for global statistics, reduced trans-
mission of haplotypes 1-1-1 and 1-1-0 to the myopic offspring
was still significant (z ? ?3.139, global P ? 0.0031; and z ?
?3.445, global P ? 0.0015, respectively).
With high myopia considered as a dichotomous qualitative
trait, similar results were also obtained except that most prob-
abilities were slightly larger than when high myopia was
treated as a quantitative trait (MSE). Detailed results are avail-
able online in Supplementary Tables S1, S2 at http://www.iovs
.org/cgi/content/full/47/6/2291/DC1. In particular, the three
critical global probabilities for haplotype analysis were 0.0061
for three-locus analysis and 0.0025 for two-locus (HGF5-5b and
HGFe9) analysis under the recessive model, and 0.0076 for
two-locus (HGF5-5b and HGFe10b) analysis under the domi-
nant model. In comparison, the respective probabilities were
0.0031, 0.0015, and 0.0040 when analysis was performed with
the measured trait MSE (Table 4). Both sets of probabilities
remained significant after correction for multiple testing based
on FDR. However, with high myopia considered as a qualita-
tive trait, the initial significant results for single marker analysis
did not survive correction for multiple testing.
TABLE 2. Clinical and Demographic Information
of High-Myopia Siblings*
Age at entry ? SD (y)
Sex ratio (male/female)†
Onset age of myopia SD (y)
Families with no myopic
Families with one myopic
Families with one high-
Families with two myopic
Families with two high
MSE ? SD (D)
Range of spherical
Range of cylinder diopter
AXL ? SD (mm)
CP ? SD (D)
ACD ? SD (mm)
LT ? SD (mm)
22.32 ? 11.18
70/63 (p ? 0.05)
7.54 ? 2.95
?12.08 ? 3.37
?9.50 to ?22.50
0 to ?4.25
28.07 ? 1.64 (r ? ?0.70, p ? 0.001)
42.79 ? 1.90 (r ? ?0.26, p ? 0.001)
3.69 ? 0.29 (r ? ?0.29, p ? 0.001)
3.65 ? 0.27 (r ? ?0.05, p ? 0.385)
n ? 133.
* The data in parentheses show the results of partial correlation
test between MSE and ocular components of AXL, CP, ACD and LT.
AXL showed the most significant correlation to MSE.
† The sex difference of myopia prevalence was tested among the
‡ The parents with both eyes having a spherical equivalent of
?0.75 D or worse are defined as myopic and those with ?6.0 D or
worse are defined as highly myopic.
TABLE 3. Single Locus Association Tests for the HGF Gene by FBAT Analysis with the Measured Trait MSE*
Frequency in parents
FBAT: additive model
FBAT: dominant model
FBAT: recessive model
98 98 7373
df ? 1, ?2? 5.831, P ? 0.0157†
df ? 1, ?2? 0.623, P ? 0.4299
df ? 1, ?2? 2.915, P ? 0.0877
df ? 2, ?2? 1.187, P ? 0.5524
df ? 2, ?2? 2.941, P ? 0.2298
df ? 2, ?2? 9.051, P ? 0.0108†
df ? 2, ?2? 2.941, P ? 0.2298
df ? 2, ?2? 9.051, P ? 0.0108†
df ? 2, ?2? 1.187, P ? 0.5524
* FBAT analysis was performed, with high myopia being considered as a quantitative trait measured as MSE. Alleles 1 and 2 are the major and
minor alleles, respectively. n, the number of informative families in which there is at least one heterozygous parent. The z scores are shown in bold
if outside the range of ? 2.000. The probabilities are shown in bold if significant at the 0.05 level and marked by symbol † if significant (?0.0167)
after correction for multiple comparisons (n ? 6) based on FDR. Note the reciprocal relationship between the dominant and the recessive models
for bi-allelic markers.
IOVS, June 2006, Vol. 47, No. 6
HGF Polymorphisms in High Myopia 2295
TABLE 4. Haplotype Association Tests for the HGF Gene by FBAT Analysis with the Measured Trait MSE*
df ? 5,
P ? 0.0328
df ? 5,
P ? 0.0381
df ? 2,
P ? 0.0031†
df ? 3,
P ? 0.0542
df ? 4,
P ? 0.0883
df ? 2,
P ? 0.0015†
df ? 3,
P ? 0.0209
df ? 3,
df ? 2,
P ? 0.0192
df ? 2,
P ? 0.2147
df ? 3,
P ? 0.3788
df ? 2,
P ? 0.5460
* FBAT analysis was performed with high myopia being considered as a quantitative trait measured as MSE. HF, represents haplotype frequency; n, the number of informative families in which
there is at least one heterozygous parent. Alleles 1 and 2 are the major and minor alleles, respectively. For the sake of easy discussion, a zero (0) is inserted when one of the three SNPs is not included
in the haplotype concerned. The degree of freedom indicates the number of haplotypes tested in the global statistic for a given set of haplotypes involving two or three SNPs. Note that haplotypes
with frequency ?0.15 or with the number of informative families ?10 are not shown. The Z scores are shown in bold if outside the range of ? 2.000. The probabilities are shown in bold if significant
at the 0.05 level, and marked by † if significant (?0.0125) after correction for multiple comparisons based on FDR.
2296Han et al.
IOVS, June 2006, Vol. 47, No. 6
Also treating high myopia as a qualitative trait, GenAssoc
generated a case–pseudocontrol dataset. Analysis of HGF5-5b
with GenAssoc gave a GRR of 2.19 (95% CI ? 1.29–3.72, P ?
0.004) for genotype 1/2 and 2.14 (95% CI ? 1.02–4.49, P ?
0.043) for 2/2 with reference to 1/1. This is consistent with the
increased transmission of allele 2 (T) under the dominant
Haplotype pattern in the HGF locus was partitioned according
to the CI of D? in this study. On the basis of the LD map (Fig.
1), three SNPs—HGF5-5b, HGFe9, and HGFe10b—were se-
lected as the markers for subsequent association test. SNP
HGFe9 was categorized as a “hole” inserting in the haplotype
block, a type of departure from strict haplotype criteria,48and
should be tested individually in association study. Simply se-
lecting markers for a genetic association test may substantially
decrease the efficacy of the TDT test, as TDT crucially depends
on the LD between the markers and disease loci.49The “hole”
HGFe9, as categorized by the LD pattern, could be missed if
selecting markers for analysis was arbitrary. It is also desirable
to select markers based on homogeneous ethnic population as
ethnic population heterogeneity for high myopia could con-
found the results of genetic analysis.50We also calculated the
LD for these three marker SNPs in the parents of high myopic
nuclear families. The SNP pair of HGF5-5b and HGFe10b
showed a higher LD level but still remained “uninformative,”
whereas SNP HGFe9 still behaved like a “hole” (data not
shown). The results were in agreement with those (Fig. 1) for
random population. The similar LD pattern in the parental
population suggests the homogeneity of the populations in our
study and the usefulness of establishing LD patterns in ad-
vance. A recent Japanese study also showed a similar LD pat-
tern in the HGF locus.51Moreover, online data of the Human
Haplotype Map Project also showed that LD patterns in the
HGF locus for the Chinese, Japanese and European populations
(http://www.hapmap.org) are largely similar to the LD pat-
terns reported herein. No coding SNPs in the HGF gene were
identified in our random Chinese population, and the SNPs
identified in the 5? upstream region were not involved in the
promoter region and regulatory sites documented in ElDorado.
We speculate that other potential common polymorphisms of
interest may give rise to the variation of gene regulation, but
not amino acid substitution in the Chinese population.
Genetic Association Analysis
The etiology of myopia is still not well understood. Finding the
susceptibility genes will lead to a better understanding of the
mechanisms underlying myopia and hence finding effective
ways to control or treat myopia. Up to now, mapping genes
predisposing to myopia is still a great challenge. This work is
the first effort to assess the relationship between the HGF gene
and human high myopia.
According to the results of FBAT with the quantitative trait
MSE, no significant association was observed for HGFe9 and
HGFe10b for either allele or haplotype (Tables 3, 4). In con-
trast, the 5? upstream side seems to be an interesting region of
the HGF locus, as the marker HGF5-5b is always implicated in
the significant results for either allele or haplotype. FBAT
analysis showed that HGF5-5b was associated with high myo-
pia under the three genetic models tested, even after correc-
tion for multiple testing (Table 3). The GRR estimates indicated
that the genotypes 1/2 (C/T) and 2/2 (T/T) were risk factors for
high myopia with reference to 1/1 (C/C): 2.19 (P ? 0.004) and
2.14 (P ? 0.043), respectively. This is compatible with FBAT
results for HGF5-5b under the dominant–recessive models.
Such analysis highlights the complexity of myopia inheritance,
as has been suggested by previous studies.16,52The fact that 50
families had no myopic parent in our study also reflects the
complexity of inheritance modes of high myopia. In line with
the results of single-marker association, tests for haplotypes
also showed similar patterns. Haplotypes 1-1-1 or 1-1-0 showed
significant association under the recessive model, whereas
haplotype 2-0-1 demonstrated significant association in the
dominant model (Table 4). The association was also significant
after correction for multiple comparisons.
In general, the same conclusion was reached no matter
whether high myopia was considered as a quantitative trait
(MSE) or a qualitative trait (affected or unaffected). But, our
results suggest that analysis based on the measure trait MSE is
slightly more powerful than that based on the dichotomous
trait. Because AXL correlated highly with MSE (r ? ?0.70;
Table 2), we expect that analysis based on AXL should give
similar results. This is in fact the case: the results paralleled
those when myopia was considered a qualitative trait, except
that the probabilities were in general slightly larger (data not
It is tempting to speculate that other SNPs of interest may
exist in proximity to HGF5-5b and on its implicated haplo-
types. HGF is an inducible cytokine and the promoter and
some regulatory factors of the HGF gene have been character-
ized.53,54SNP HGF5-5b is not located in these experimentally
verified regulatory sites. Other nearby sequence variations are
worthy of further investigation to identify the functional SNPs
that may play a role in the development of myopia. Recently,
two case?control studies in Japanese populations reported
that the polymorphisms in intron 8 (43839A3T; see Table 1)
and intron 13 (not reported in our study) of the HGF locus may
be associated with vascular diseases including hypertension
and atherosclerosis.51,55Our results suggest that the HGF5-5b
and its adjacent polymorphisms may be the interesting loci for
high myopia. The finding that the HGF locus is associated with
different diseases may not be too surprising, considering the
multifunctional roles of HGF.20–23,29–33
Less frequent SNPs like HGFe14b and HGFe18-1a (Table 1)
may also contribute to a common trait like myopia. Our pre-
liminary data showed that these two SNPs were also not com-
mon in our high myopia population (MAF ? 0.03 for HGFe14b
and 0.06 for HGFe18-1a) and were not associated with high
myopia (data not shown).
Myopia is a delicate change in refractive error with ocular
components of AXL, ACD, CP, and LT all being involved.56The
profile of genetic and/or environmental components for myo-
pic subjects is highly variable. These could confound the asso-
ciation study of myopia. Strict entry criteria and careful phe-
notyping of myopic subjects are crucially important for the
success of myopia association study. The price to pay here is
the corresponding increased difficulty in subject recruitment.
With the TDT-based approach in our study, the spurious asso-
ciation caused by population stratification in conventional
case–control study was effectively eliminated. Additional ef-
forts were made to enhance the efficacy of FBAT in our study.
First, only typical high myopia (? ?10.0D) was included, and
the onset age of myopia for all affected subjects was less than
12 years. Such criteria may enhance the contribution of the
genetic component to the myopia trait in the subjects stud-
ied.57Second, the ocular components ACD and CP were fac-
tored out, and AXL was taken into account purposely in myo-
pic subject recruitment to diminish the complexity of genetic
background.56This is because typical high myopia is mainly
due to the elongation of ocular AXL, and inheritance has a
significant impact on AXL but not on ACD or CP.12,58The most
significant correlation between AXL and refractive error (Table
IOVS, June 2006, Vol. 47, No. 6
HGF Polymorphisms in High Myopia2297
2) demonstrated that the posterior axial myopia was studied
and the confounding due to other ocular components was
largely ruled out. Third, establishing the LD pattern in the same
ethnic population beforehand in this study also enhanced the
validity of association test. Nevertheless, the high complexity
of genetic background for our high myopic population still
exists, and each potential candidate gene may only have a mild
effect on myopia onset and severity. A larger sample size and
replication with independent sample sets are always instruc-
tive in drawing a clearer conclusion regarding the relation
between the HGF gene and high myopia.
As a part of our ongoing joint effort to identify myopia
susceptibility genes, this study established the LD pattern in
the HGF locus and investigated the association between three
tag SNPs and high myopia in a group of Chinese families with
high myopia. Our study followed a logical approach to associ-
ation test on the basis of characterizing the LD patterns, since
the selection of markers based on LD pattern in candidate
genes in advance was a critical part of the association study.59
Our results at this stage showed the evidence of association of
the HGF locus with early-onset high myopia in a Han Chinese
population. According to the results of this exploratory work,
we consider that the 5? region of the HGF gene may contain
potential polymorphisms affecting myopia susceptibility in the
Han Chinese population. Study with a larger sample size and
denser SNP markers or in other populations is needed to
further elucidate the relationship between the HGF gene and
high myopia and to identify functional SNPs that play a role in
1. Wong TY, Foster PJ, Hee J, et al. Prevalence and risk factors for
refractive errors in adult Chinese in Singapore. Invest Ophthalmol
Vis Sci. 2000;41:2486–2494.
2. Zhao JL, Pan XJ, Sui RF, Munoz S, Sperduto RD, Ellwein LB.
Refractive error study in children: results from Shunyi District,
China. Am J Ophthalmol. 2000;129:427–435.
3. Curtin BJ. The Myopias: Basic Science and Clinical Management.
Philadelphia; Harper & Row; 1985.
4. Seet B, Wong TY, Tan DTH, et al. Myopia in Singapore: taking a
public health approach. Br J Ophthalmol. 2001;85:521–526.
5. Yap M, Wu M, Liu ZM, Lee FL, Wang SH. Role of heredity in the
genesis of myopia. Ophthalmic Physiol Opt. 1993;13:316–319.
6. Mutti DO, Zadnik K, Adams AJ. Myopia: the nature versus nurture
debate goes on. Invest Ophthalmol Vis Sci. 1996;37:952–957.
7. Saw SM, Chua WH, Wu HM, Yap E, Chia KS, Stone RA. Myopia:
gene-environment interaction. Ann Acad Med Singapore. 2000;
8. Lyhne N, Sjølie AK, Kyvik KO, Green A. The importance of genes
and environment for ocular refraction and its determiners: a pop-
ulation based study among 20–45 year old twins. Br J Ophthal-
9. Hammond CJ, Snieder H, Gilbert CE, Spector TD. Genes and
environment in refractive error: the twin eye study. Invest Oph-
thalmol Vis Sci. 2001;42:1232–1236.
10. Teikari JM, O’Donnell J, Kaprio J, Koskenvuo M. Impact of hered-
ity in myopia. Hum Hered. 1991;41:151–156.
11. Morgan I, Rose K. How genetic is school myopia? Prog Retin Eye
12. Liang CL, Yen E, Su JY, et al. Impact of family history of high
myopia on level and onset of myopia. Invest Ophthalmol Vis Sci.
13. Schwartz M, Haim M, Skarsholm D. X-linked myopia: Bornholm
eye disease—linkage to DNA markers on the distal part of Xq. Clin
14. Paluru PC, Nallasamy S, Devoto M, Rappaport EF, Young TL.
Identification of a novel locus on 2q for autosomal dominant
high-grade myopia. Invest Ophthalmol Vis Sci. 2005;46:2300–
15. Zhang Q, Guo X, Xiao X, Jia X, Li S, Hejtmancik JF. A new locus for
autosomal dominant high myopia maps to 4q22–q27 between
D4S1578 and D4S1612. Mol Vis. 2005;11:554–560.
16. Farbrother JE, Kirov G, Owen MJ, Guggenheim JA. Family aggre-
gation of high myopia: estimation of the sibling recurrence risk
ratio. Invest Ophthalmol Vis Sci. 2004;45:2873–2878.
17. Risch N, Merikangas K. The future of genetic studies of complex
human diseases. Science. 1996;273:1516–1517.
18. Spielman RS, McGinnis RE, Ewens WJ. Transmission test for link-
age disequilibrium: the insulin gene region and insulin-dependent
diabetes mellitus (IDDM). Am J Hum Genet. 1993;3:506–516.
19. Ewens WJ, Spielman RS. The transmission/disequilibrium test:
History, subdivision, and admixture. Am J Hum Genet. 1995;57:
20. Tamagnone L, Comoglio PM. Control of invasive growth by hepa-
tocyte growth factor (HGF) and related scatter factors. Cytokine
Growth Factor Rev. 1997;8:129–142.
21. Sun W, Funakoshi H, Nakamura T. Differential expression of he-
patocyte growth factor and its receptor, c-Met in the rat retina
during development. Brain Res. 1999;851:46–53.
22. Daniel JT, Limb GA, Saarialbo-Kere U, Murphy G, Khaw PT. Human
corneal epithelial cells require MMP-1 for HGF-mediated migration
on collagen I. Invest Ophthalmol Vis Sci. 2003;44:1048–1055.
23. Grierson I, Heathcote L, Hiscott P, Hogg P, Briggs M, Hagan S.
Hepatocyte growth actor/scatter factor in the eye. Prog Retin Eye
24. Zhou G, Williams RW. Eye1 and Eye2: gene loci that modulate eye
size, lens weight, and retinal area in the mouse. Invest Ophthalmol
Vis Sci. 1999;40:817–882.
25. Jones BE, Thompson EW, Hodos W, Waldbillig RJ, Chader GJ.
Scleral matrix metalloproteinases, serine proteinase activity and
hydrational capacity are increased in myopia induced by retinal
image degradation. Exp Eye Res. 1996;63:369–381.
26. Siegwart JT, Norton TT. The time course of changes in mRNA
levels in tree shrew sclera during induced myopia and recovery.
Invest Ophthalmol Vis Sci. 2002;43:2067–2075.
27. Gentle A, Liu Y, Martin JE, Conti GL, McBrien NA. Collagen gene
expression and the altered accumulation of scleral collagen during
the development of high myopia. J Biol Chem. 2003;278:16587–
28. McBrien NA, Gentle A. Role of the sclera in the development and
pathological complications of myopia. Prog Retin Eye Res. 2003;
29. Dunsmore SE, Rubin JS, Kovacs SO, Chedid M, Parks WC, Welgus
HG. Mechanisms of hepatocyte growth factor stimulation of kera-
tinocyte metalloproteinase production. J Biol Chem. 1996;271:
30. Hamasuna R, Kataoka H, Moriyama T, Itoh H, Seiki M, Koono M.
Regulation of matrix metalloproteinase-2 (MMP-2) by hepatocyte
growth factor/scatter factor (HGF/SF) in human glioma cells:
HGF/SF enhances MMP-2 expression and activation accompanying
up-regulation of membrane type-1 MMP. Int J Cancer. 1999;82:
31. Gong RJ, Rifai A, Tolbert EM, Centracchio JN, Dworkin LD. Hepa-
tocyte growth factor modulates matrix metalloproteinases and
plasminogen activator/plasmin proteolytic pathways in progres-
sive renal interstitial fibrosis. J Am Soc Nephrol. 2003;14:3047–
32. Gaggioli C, Deckert M, Robert G, et al. HGF induces fibronectin
matrix synthesis in melanoma cells through MAP kinase-dependent
signaling pathway and induction of Egr-1. Oncogene. 2005;24:
33. Gerritsen ME, Tomlinson JE, Zlot C, Ziman M, Hwang S. Using
gene expression profiling to identify the molecular basis of the
synergistic actions of hepatocyte growth factor and vascular en-
dothelial growth factor in human endothelial cells. Br J Pharma-
34. Zhong X, Ge J, Smith III EL, Stell WK. Image defocus modulates
activity of bipolar and amacrine cells in macaque retina. Invest
Ophthalmol Vis Sci. 2004;45:2065–2074.
35. Han W, Yip SP, Wang J, Yap MKH. Using denaturing HPLC for SNP
discovery and genotyping, and establishing the linkage disequilib-
2298 Han et al.
IOVS, June 2006, Vol. 47, No. 6
rium pattern for the all- trans-retinol dehydrogenase (RDH8) gene. Download full-text
J Hum Genet. 2004;49:16–23.
36. Chen X, Levine L, Kwok PY. Fluorescence polarization in homo-
geneous nucleic acid analysis. Genome Res. 1999;9:492–498.
37. Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and visu-
alization of LD and haplotype maps. Bioinformatics. 2005;21:263–
38. Gabriel SB, Schaffner SF, Nguyen H, et al. The structure of haplo-
type blocks in the human genome. Science. 2002;296:2225–2229.
39. Rabinowitz D, Laird N. A unified approach to adjusting association
tests for population admixture with arbitrary pedigree structure
and arbitrary missing marker information. Hum Hered. 2000;50:
40. Laird NM, Horvath S, Xu X. Implementing a unified approach to
family-based tests of association. Genet Epidemiol. 2000;19:S36–
41. Horvath S, Xu X, Lake SL, Silverman EK, Weiss ST, Laird NM.
Family based tests for association haplotypes with general trait
data: application to asthma genetics. Genet Epidemiol. 2004;26:
42. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a
practical and powerful approach to multiple testing. J R Statist Soc
43. Cordell HJ, Clayton DG. A unified stepwise regression procedure
for evaluating the relative effects of polymorphisms within a gene
using case/control or family data: application to HLA in type 1
diabetes. Am J Hum Genet. 2002;70:124–141.
44. Cordell HJ, Barratt BJ, Clayton DG. Case/pseudocontrol analysis in
genetic association studies: a unified framework for detection of
genotype and haplotype associations, gene-gene and gene-environ-
ment interactions, and parent-of-origin effects. Genet Epidemiol.
45. den Dunnen JT, Antonarakis SE. Nomenclature for the description
of human sequence variations. Hum Genet. 2001;109:121–124.
46. Garner LF, Meng CK, Grosvenor TP, Mohidin N. Ocular dimen-
sions and refractive power in Malay and Melanesian children.
Ophthalmic Physiol Opt. 1990;10:234–238.
47. Goss DA, Van Veen HG, Rainey BB, Feng B. Ocular components
measured by keratometry, phakometry, and ultrasonography in
emmetropic and myopic optometry students. Optom Vis Sci. 1997;
48. Wall JD, Pritchard JK. Assessing the performance of the haplotype
block model of linkage disequilibrium. Am J Hum Genet. 2003;
49. Xiong M, Guo SW. The power of linkage detection by the trans-
mission/disequilibrium tests. Hum Hered. 1998;48:295–312.
50. Edwards MH. Effect of parental myopia on the development of
myopia in Hong Kong Chinese. Ophthalmic Physiol Opt. 1998;18:
51. Takiuchi S, Mannami T, Miyata T, et al. Identification of 21 single
nucleotide polymorphisms in human hepatocyte growth factor
gene and association with blood pressure and carotid atheroscle-
rosis in the Japanese population. Atherosclerosis. 2004;173:301–
52. Edwards MH, Lewis WH. Autosomal recessive inheritance of my-
opia in Hong Kong Chinese infants. Ophthalmic Physiol Opt.
53. Jiang J, Gao B, Zarnegar R. The concerted regulatory functions of
the transcription factors nuclear factor-1 and upstream stimulatory
factor on a composite element in the promoter of the hepatocyte
growth factor gene. Oncogene. 2000;19:2786–2790.
54. Zarnegar R. Regulation of HGF and HGFR gene expression. EXS.
55. Motone M, Katsuya T, Ishikawa K, et al. Association between
hepatocyte growth factor gene polymorphism and essential hyper-
tension. Hypertens Res. 2004;27:247–251.
56. Sorsby A, Leary GA, Richard MJ. Correlation ammetropia and
component ammetropia. Vis Res. 1962;2:309–313.
57. Rebbeck TR. Inherited genetic predisposition in breast cancer: a
population-based perspective. Cancer. 1999;86:2493–2501.
58. Saw SM, Carkeet A, Chia KS, Stone RA, Tan TDH. Component
dependent risk factors for ocular parameters in Singapore Chinese
children. Ophthalmology. 2002;109:2065–2071.
59. Cardon LR, Abecasis GR. Using haplotype blocks to map human
complex trait loci. Trend Genet. 2003;19:135–140.
IOVS, June 2006, Vol. 47, No. 6
HGF Polymorphisms in High Myopia 2299