High Temperature Requirement Factor A1 (HTRA1) Gene
Regulates Angiogenesis through Transforming Growth
Factor-? Family Member Growth Differentiation Factor 6*
Li Zhang‡§1, Siok Lam Lim§1, Hongjun Du§, Ming Zhang‡, Igor Kozak§, Gregory Hannum¶, Xiaolei Wang?,
Hong Ouyang§, Guy Hughes§, Ling Zhao§, Xuemei Zhu§, Clara Lee§, Zhiguang Su‡, Xinrong Zhou§, Robert Shaw§,
Dongho Geum§, Xinran Wei‡§, Jin Zhu‡§, Trey Ideker¶, Chio Oka**, Ningli Wang?, Zhenglin Yang‡‡2,
Peter X. Shaw‡§3, and Kang Zhang‡§4
andSichuanProvincialPeople’sHospital,Chengdu,Sichuan 610072, China
Background: Genetic variants of high temperature requirement factor A1 (HTRA1) associate with AMD risk.
Results: Growth differentiation factor 6 (GDF6) gene polymorphism significantly associated with AMD. HTRA1 knock-out
mice display reduced blood vessel in retina and up-regulation of GDF6.
Conclusion: HTRA1 regulates angiogenesis via TGF-? signaling by GDF6, a novel disease gene.
Significance: This novel pathway of HTRA1 in regulation of vascularization is critical for understanding AMD pathogenesis.
Genome-wide association study (GWAS) has identified
genetic variants in the promoter region of the high temperature
requirement factor A1 (HTRA1) gene associated with age-re-
lated macular degeneration (AMD). As a secreted serine prote-
ase, HTRA1 has been reported to interact with members of the
signaling pathways. Growth differentiation factor 6 (GDF6), a
and eye development. Mutations in GDF6 have been associated
mia and anophthalmia. In this report, we identified a single
nucleotide polymorphism (SNP) rs6982567 A/G near the GDF6
10?8). We demonstrated that the GDF6 AMD risk allele
(rs6982567 A) is associated with decreased expression of the
GDF6 and increased expression of HTRA1. Similarly, the
HTRA1 AMD risk allele (rs10490924 T) is associated with
decreased GDF6 and increased HTRA1 expression. We
observed decreased vascular development in the retina and sig-
nificant up-regulation of GDF6 gene in the RPE layer, retinal
and brain tissues in HTRA1 knock-out (htra1?/?) mice as com-
enhanced SMAD signaling in htra1?/?mice. Our data suggests
a critical role of HTRA1 in the regulation of angiogenesis via
Age-related macular degeneration (AMD)5is the leading
cause of blindness among elderly patients in developed coun-
tries (1). Advanced AMD can be categorized as geographic
atrophy or choroidal neovascularization (CNV). Geographic
atrophy, also called dry AMD, is characterized by regional ret-
inal pigment epithelium (RPE) loss and the eventual degenera-
tion of overlying photoreceptors. CNV, also known as wet
roid through the Bruch membrane toward the retina. In the
case of CNV, bleeding and fluid leakage often occur due to
abnormal angiogenesis, resulting in acute loss of central vision.
Although the pathogenesis of AMD is still largely unknown, it
has been reported that abnormal vascular growth mediated by
vascular endothelial growth factor or other angiogenic factors
plays a pivotal role in the pathogenic angiogenesis process in
Our previous research has identified that HTRA1 on chro-
mosome 10q26 influences the risk of AMD (5). HTRA1 may
modulate AMD development through many potential path-
ways. Multiple substrates of HTRA1 have been identified,
* This work was supported, in whole or in part, by a National Institutes of
University and West China Hospital, a Veterans Affairs Merit Award,
lational Research (to K. Z.).
1Both authors contributed equally to this work.
2Supported by National Natural Science Foundation of China Grant
81025006. To whom correspondence may be addressed. E-mail: zliny@
3To whom correspondence may be addressed. E-mail: firstname.lastname@example.org.
4To whom correspondence may be addressed: 9500 Gilman Dr., La Jolla, CA
92093-0838. Tel.: 858-246-0823; Fax: 858-246-0961; E-mail: kang.zhang@
5The abbreviations used are: AMD, age-related macular degeneration; CNV,
choroidal neovascularization; RPE, retinal pigment epithelium; CARASIL,
cerebral autosomal recessive arteriopathy with subcortical infarcts and
leukoencephalopathy; SNP, single nucleotide polymorphism; IPL, inner
plexiform layer; GDF, growth differentiation factor; HTRA1, high tempera-
ture requirement factor A1.
THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 287, NO. 2, pp. 1520–1526, January 6, 2012
© 2012 by The American Society for Biochemistry and Molecular Biology, Inc.Published in the U.S.A.
1520 JOURNALOFBIOLOGICALCHEMISTRY VOLUME287•NUMBER2•JANUARY6,2012
at Biomedical Library, UCSD, on April 17, 2012
including aggrecan, decorin, biglycan, fibromodulin, fibronec-
tin, and transforming growth factor-? (TGF-?) family mem-
repress signaling by members of the TGF-? family (8, 9). A
recent study found that mutants of HTRA1 with diminished
small blood vessel in the brain, causing cerebral autosomal
recessive arteriopathy with subcortical infarcts and leukoen-
cephalopathy (CARASIL, a nonhypertensive cerebral small-
suppress TGF-? activity, leading to increased expression of
TGF-?1 in the tunica in affected small arteries. These findings
indicate that CARASIL is a vascular disease associated with
dysregulation of TGF-? signaling (10). As a secreted protein,
HTRA1 binds to several members of the TGF-? family, such as
TGF?1, TGF?2, activin, bone morphogenetic protein 4, and
growth differentiation factor 5 (GDF5) (8). Interaction of
HTRA1 with these factors alters the signaling pathways medi-
ated by the binding of these factors with their receptors (11).
GDF6 is a member of the TGF-? family and has been found to
express in the dorsal retina of developing mouse, Xenopus, and
ectoderm patterning (13) and control eye development by reg-
ulating neural and vascular development (14, 15). GDF6 muta-
tions have been associated with eye phenotypes such as
microphthalmia and anophthalmia (14).
In this report, we performed a genetic association study
and identified a single nucleotide polymorphism (SNP)
rs6982567A/G near GDF6 that is significantly associated with
AMD. The risk allele in rs6982567 was found to associate with
decreased levels of GDF6 and increased levels of HTRA1 gene
expression. Similarly, the HTRA1 AMD risk allele is also asso-
ciated with decreased GDF6 and increased HTRA1 expression.
In the retina of HTRA1 knock-out (htra1?/?) mice, we
observed decreased vascular development as well as significant
the RPE layer in comparison to the wild-type mice. Further-
more, as downstream effectors of GDF6 signaling, increased
levels of phosphorylated SMAD1/5/8 were detected in the
brain tissue of htra1?/?mice, suggesting that loss of HTRA1
affects regulation of signaling pathways involving the TGF-?
family members. Our data supports a critical role of HTRA1 in
the regulation of angiogenesis via TGF-? signaling and identi-
fied GDF6 as a novel disease gene for AMD.
Patients—This study was approved by the Institutional
Review Boards of the West China Hospital and University of
California, San Diego. Subjects gave informed consent prior to
participation. Participants underwent a standard examination,
which included visual acuity measurements, dilated slit lamp
biomicroscopy, and stereoscopic color fundus photography.
Grading was carried out with the classification established by
Age-related Eye Disease Study (AREDS) (16). Diagnosis of
advanced AMD was based on the presence of CNV (equivalent
to AREDS category 4 or 5). Control subjects were defined as
being ?60 years old, having fewer than 5 small drusen (?63
?m), and no RPE abnormalities. 2313 unrelated Caucasian
individuals of European descent comprising 1538 AMD
patients and 775 normal controls from the San Diego area
participated in this study. An independent cohort of 3307
unrelated Caucasian individuals comprising 2158 AMD
patients and 1149 normal controls was drawn from the
Michigan, Mayo, Age-related Eye Disease Study, Pennsylva-
Genotyping—SNPs were genotyped by SNaPshot on an ABI
(17). For rs6982567A/G, primers 5?-AAAGAGGTTCAGGG-
GATTTACA-3? and 5?-GGGCAGCTCAAGTCCTAATG-3?
were used to generate the amplicon encompassing the SNP
by standard PCR; 5?-GTTTGATCCTTTCATCTTGATTAG-
used as the SNaPshot primer. For rs10490924 T/G, primers 5?-
GCAAGTCTGTCCTCCTCGGT-3? and 5?-GTCTGGGGTA-
AGGCCTGATCAT-3? were used to generate the amplicon
encompassing the SNP by standard PCR. 5?-CAAACTGTCT-
TTATCACACTCCATGATCCCAGCT-3? was used as the
SNaPshot primer. Genotyping success rate was ?98% and
of the samples in case-control series.
Animals—All animal experiments followed the guidelines of
the Association for Research in Vision and Ophthalmology
(ARVO) Statement for the Use of Animals in Ophthalmic and
Vision Research and were approved by the Animal Care Com-
mittee of University of California, San Diego, and West China
Hospital, Sichuan University. Homozygous htra1?/?mice
were generated by standard gene targeting procedures as
described previously (18). Briefly, positive ES cell clones were
microinjected into C57BL/6J blastocysts to generate chimeras.
ras with C57BL/6J mice. The homozygous htra1?/?mice were
produced by inbreeding of the htra1?/?mice. For genotyping,
The wild-type allele was identified with a forward primer (5?-
CACTACGCATTGCAGCCCCTC-3?) and a reverse primer
(5?- CGTACCACGCTCCTGTCTTT-3?). Genotypes were
confirmed by PCR of genomic DNA extracted from tail snips.
Wild-type C57BL/6 littermate mice served as normal controls.
Histology of Retinal Vasculature—The animals were sacri-
ficed and eyes were fixed with 4% paraformaldehyde in PBS at
4 °C for 2 h (adult) or overnight (7-day postnatal). After three
washes with PBS, the eyes were then cut 1-mm behind the lim-
bus, and the anterior segments including cornea, iris, and lens
were removed under a microscope. After being separated from
the edge of the retina. The vitreous were gently removed from
the retinal surface using forceps. The retina were subsequently
incubated with 0.5% fluorescein isothiocyanate (FITC)-isolec-
tin B4 (Vector Laboratories, Burlingame, CA) overnight. Blood
vessels were visualized with blue argon laser wavelength (488
nm) using a scanning laser confocal microscope FV1000
(Olympus Corporation of the Americas, Center Valley, PA).
at Biomedical Library, UCSD, on April 17, 2012
nerve fiber layer, inner plexiform layer, and outer plexiform
layer were taken at the same distance from the optic head in
Retinal Vessel Density Examinations—We compared histol-
ogy images of retinal blood vessels from both wild-type and
htra1?/?mice at 7 days and 1 month postnatal. The images of
retinal sections in different histological layers were processed
using Adobe Photoshop (Adobe Systems Inc., San Jose, CA).
The images were cropped and converted to black and white
images. Pixel threshold of the images were determined, and
then a Gaussian blur under Filter function was used to elimi-
nate background signal and capture images of retinal vascula-
ture only. Layer subtraction function under Image-Calculation-
Subtract was used to calculate blood vessel density. The values
Immunohistochemistry of Mouse Retina—Eyes from wild-
type mice were sectioned and stained with rabbit anti-GDF6
radish peroxidase. The VectorStain Elite ABC substrate kit
were then counterstained with methyl green. Images were cap-
tured using an Axio Observer A1 microscope (Carl Zeiss
MicroImaging, Thornwood, NY).
Real-time PCR for Gene Expression—Total RNA was
extracted from mouse tissues or human lymphocytes using the
RNeasy Mini Kit (Qiagen Inc., Valencia, CA), and converted to
cDNA using the SuperScript III First Strand Synthesis System
with Power SYBR Green qPCR Master Mix (Applied Biosys-
tems, Foster City, CA) and analyzed with the 7500 Real-time
PCR Detection System (Applied Biosystems, Foster City, CA).
Primer sets used are listed in Table 1. Relative mRNA levels
compared with tissues from wild-type littermates or human
lymphocytes of protective genotypes.
Immunoblotting—Brain tissues were harvested from wild-
lysis buffer with phosphatase inhibitors (Cell Signaling Tech-
nology, Danvers, MA). Total protein concentrations were
determined by a BCA assay (Bio-Rad) and normalized. Protein
PVDF membrane and probed using anti-SMAD1 and anti-
phospho-SMAD1/5/8 (Cell Signaling Technology). Immuno-
blot signal was measured and analyzed using ImageJ (imagej.
nih.gov/ij) (19). Ratios of phosphor-SMAD/SMAD were
Statistical Analysis—The initial genetic association analysis
and controls. Subsequently, Pearson ?2statistics were calcu-
and genotypic tests, respectively. SNPs yielding a p value with
statistical significance were selected for further analysis. The
difference in vascular density and quantitative PCR data were
analyzed with a paired Student’s t test.
GDF6 Is Associated with Risk of AMD—One of the HTRA1
functions is to regulate signaling by TGF-? family members
tigated genetic associations of TGF-? family members with
AMD. We found a single significant association, SNP
rs6982567 located in 8q22.1 near the GDF6 gene. A total of
2313 unrelated Caucasian individuals of European descent
were genotyped as the first discovery cohort, which included
onstrated that the SNP rs6982567 was associated with AMD
with statistical significance (allelic p value of 0.017). This asso-
ciation was validated in an independent cohort of Caucasian of
European descent drawn from MMAP (2158 AMD cases and
1149 controls). After combining the Discovery cohort with the
MMAP cohort, SNP rs6982567 showed a highly significant
allele frequency being 20.9% in cases versus 16.6% in controls
Reciprocal Regulation of Gene Expression between GDF6 and
HTRA1—Because HTRA1 has been reported to regulate the
TGF-? family signaling pathway, we first examined gene
Primer sets used for real-time quantitative PCR
Forward sequence (5? 3 3?)
Reverse sequence (5? 3 3?)
Association between single nucleotide polymorphism (SNP) rs6982567 A/G and AMD in Discovery, MMAP-CEU, and combination cohorts
For the Discovery cohort, 2313 unrelated Caucasian individuals of European descent comprising 1538 AMD patients and 775 normal controls participated in this study in
San Diego area. For the MMAP-CEU cohort, 3307 unrelated Caucasian individuals comprising 2158 AMD patients and 1149 normal controls were drawn from the
Michigan, Mayo, AREDS, Pennsylvania (MMAP) study cohort.
MMAP-CEU A 2158 11490.2060.154
CombinedA 369619240.209 0.166
Risk allele AMDControl Affected frequencyControl frequencyAffected HWE
at Biomedical Library, UCSD, on April 17, 2012
expression levels of HTRA1 and GDF6 in human lymphocytes
stratified by their GDF6 (rs6982567 A/G) genotypes using
quantitative real-time PCR analysis. We found that for
samples of homozygous risk genotype AA than in the normal
ulated by 70% in the risk genotype (p ? 0.041, Fig. 1A). Gene
human lymphocytes stratified by their HTRA1 (rs10490924
T/G) genotypes. Similarly, the GDF6 mRNA level was 38%
lower in lymphocyte samples of homozygous risk TT genotype
than in the normal GG genotype (p ? 0.045), whereas HTRA1
expression was up-regulated by 94% in the risk genotype (p ?
0.022, Fig. 1B).
Reduction of Retinal Vasculature in htra1?/?Mice—Given
that human patients with recessive HTRA1 mutations exhib-
ited reduced cerebral small vessels and irregular vasculature in
mice (n ? 5) with the same number of wild-type littermate
controls (C57BL/6) for their retinal vessel density on three lay-
ers of retinal capillary beds. We observed reduced retinal cap-
sities for wild-type versus htra1?/?were 17.12 and 11.22 in
nerve fiber layer (p ? 0.018); 14.61 and 8.26 in the inner plexi-
form layer (p ? 0.047); and 14.36 and 9.11 in outer plexiform
layer (p ? 0.032), respectively (Fig. 2B). The same procedures
were performed on 7-day postnatal wild-type and htra1?/?
mice (n ? 8) to study early postnatal retinal vasculature. A
lower retinal vessel density was also observed in htra1?/?mice
(p ? 0.0011, Fig. 2, C and D).
Expression of gdf6 and Other AMD-related Genes in Mouse—
GDF6 protein was detected in the ganglion cell layer, inner
plexiform layer, and RPE in wild-type mouse retina by immu-
nohistochemistry (Fig. 3A). We further examined the in vivo
gene expression of gdf6 and htra1 as well as vegf in RPE layer,
retinal tissue, and brain tissue from wild-type and htra1?/?
mice. We found that removal of the htra1 gene in mice signifi-
cantly up-regulated gdf6 gene expression and down-regulated
vegf gene expression in the RPE layer (Fig. 3B). The removal of
the htra1 gene also significantly up-regulated gdf6 in mouse
retinal tissue and brain tissue, but had less impact on vegf level
in these tissues (Fig. 3, C and D).
Increased Phosphorylation Level of SMAD in htra1?/?Mice—
SMAD proteins are important downstream effectors of TGF-?
family signaling pathways. To determine the impact of loss of
HTRA1 on TGF-? family signaling, we assayed phosphoryla-
tion of SMAD proteins using brain tissue lysate from wild-type
and htra1?/?mice (n ? 6). Comparing to wild-type, a higher
samples from htra1?/?mice (p ? 0.0197, Fig. 4).
Previously, genetic association studies had identified a
HTRA1 polymorphism strongly associated with the risk of
18, 20). However, the precise molecular mechanism by which
HTRA1 causes AMD is still largely unknown. Increasing evi-
ment, cell growth, differentiation, and apoptosis. Dysfunctions
of HTRA1 may impact aging diseases such as arthritis (7) and
tein that can affect signaling by members of the TGF-? family,
which is closely associated with vascular angiogenesis and
regulation of vascular endothelial and smooth muscle cells,
depending on cellular and extracellular environment (10, 22).
As an example, mutations in three members of the TGF-?
superfamily, including bone morphogenetic protein 4, GDF3,
and GDF6, have been associated with microphthalmia and
anophthalmia (23, 24). GDF6 is also reported to affect eye
viduals comprising 1538 AMD patients and 775 normal con-
trols, and demonstrated that SNP rs6982567 is significantly
associated with AMD. The significant association of this SNP
with AMD was further validated in an independent replication
cohort from the MMAP study. When combining our cohort
with the independent cohort, the overall association is highly
significant (p ? 3.54 ? 10?8).
We observed reciprocal regulation patterns in GDF6 and
FIGURE 1. Gene expression pattern of GDF6 and HTRA1 from human lymphocytes stratified by their (A) GDF6 (rs6982567 A/G) or (B) HTRA1
(rs10490924 T/G) genotypes. Relative mRNA levels were calculated by normalizing with GAPDH and presented as relative levels to protective genotypes.
at Biomedical Library, UCSD, on April 17, 2012
HTRA1 (rs10490924 T/G) genotypes. In lymphocytes with a
homozygous AMD risk genotype in either GDF6 rs6982567 or
HTRA1 rs10490924, there was a down-regulation of GDF6
expression with concomitant up-regulation of HTRA1 expres-
sion. These findings suggest that as a member of the TGF-?
family, GDF6 appears to be regulated by HTRA1 activity, and
that dysregulation of a GDF6-related signaling pathway by
HTRA1 likely contributes to AMD pathogenesis.
Because dysfunction of HTRA1 affects TGF-? family signal-
ing in hereditary vascular disorders such as CARASIL, causing
abnormal angiogenesis (10, 25), and our data also suggests that
genetic polymorphisms resulting in up-regulation of HTRA1
impact GDF6 expression in vivo. Our study showed that the
is significantly attenuated compared with that in the wild-type
play in retinal vasculature development and maintenance.
In mouse eye, GDF6 expression was detected in the ganglion
an important factor regulating normal vascular growth during
development, possibly via down-regulation of GDF6. This is
consistent with a recent study showing that functional muta-
tions of HTRA1 resulted in dysregulation of TGF-? signaling,
resulting in a small blood vessel phenotype with ischemic cer-
ebral disease and stroke in the brain. Mutations in the TGF-?
signaling pathway have been reported to lead to hereditary
hemorrhagic telangiectasia, Marfan syndrome, and associated
vascular disorders (25–29). Our results further support a criti-
cal role of the TGF-? signaling pathway in vascular develop-
ment and diseases.
bourg et al. (30) to identify the downstream signaling effectors
tein/GDF-related ligands using genomic analyses and in vitro
assays. The authors demonstrated that GDF6 shares down-
stream signaling molecules with many other TGF-? ligands,
and functions specifically through the SMAD1/5/8 pathway by
induction of phosphorylation. It has also been reported that
GDF6 knockdown resulted in reduced SMAD1/5/8 phosphor-
We observed significantly higher levels of SMAD1/5/8 phos-
phorylation in htra1?/?mice, suggesting that the loss of
HTRA1 leads to up-regulated expression of GDF6 and
enhanced activation of TGF-? signaling.
Ours and others’ work suggest that an appropriate level of
HTRA1 is required for vascular development and maintenance
FIGURE 2. Retinal vasculature development in mature and developing wild-type (C57BL/6) and htra1?/?mice. A, histology images of retinal blood
vessels in 1-month-old mice visualized by a scanning laser confocal microscope after FITC-isolectin staining. Pictures for different layers of blood vessel,
including nerve fiber layer (NFL), inner plexiform layer (IPL), and outer plexiform layer (OPL) were taken at the same distance from the optic head in each
S.E. n ? 5 for 1-month-old and n ? 8 for 7-day-old mice.
1524 JOURNALOFBIOLOGICALCHEMISTRY VOLUME287•NUMBER2•JANUARY6,2012
at Biomedical Library, UCSD, on April 17, 2012