16. D. E. Angelaki, A. G. Shaikh, A. M. Green, J. D. Dickman,
Nature 430, 560 (2004).
17. W. O. Guldin, O. J. Grusser, Trends Neurosci. 21, 254
18. Materials and methods are available as supporting
material on Science Online.
19. H. Imamizu et al., Nature 403, 192 (2000).
20. A mean visuomotor processing time of 171 ms was
estimated for both 1g and –1g trials, on the basis of
methods described in (4, 22).
21. Both the 1g model and the t model use on-line visual
information about target position and velocity to trig-
ger an interceptive movement when TTC(t) [see (14)]
reaches a time threshold equal to the visuomotor
processing time (22). However, the 1g model com-
putes TTC(t) under the assumption that the target is
accelerated by gravity [Eq. 1 in (14)], whereas the t
model assumes that the target moves at constant
speed [TTC(t) 0 hˆw(t)=v ˆw(t), see (4, 23)]. Therefore, if
the brain uses the 1g model to predict target motion,
the responses to 1g trials will be correctly timed. By
contrast, the t model overestimates the TTC(t) of 1g
trials and predicts that the corresponding responses
should be systematically late. As a quantitative test,
the time of response was computed separately for
the five flight durations and was averaged across all
trials of all subjects for a given flight duration. For
1g trials, the response times predicted by the 1g
model were highly correlated with the experimen-
tal values in all subjects (mean r 0 0.998), whereas
those predicted by the t model systematically failed
to fit the experimental values (mean r 0 –0.5). The
root mean square error (RMSE) of the fit to 1g trials
was worse by a factor of 1.7 for the t model than
for the 1g model. For –1g trials, the response times
predicted by the t model were highly correlated
with the experimental values in all subjects (mean
r 0 0.973), and adequacy of the fit was demon-
strated by the random distribution of the residuals
around the mean over the data range. Instead, the
1g model fitted the responses to –1g trials less
adequately, as shown by a systematic bias of the
residuals around the mean (thus r value was not
computed for this fit). Accordingly, the RMSE of
the fit was 3.6 times worse for the 1g model than
for the t model.
22. M. Zago et al., J. Neurophysiol. 91, 1620 (2004).
23. D. N. Lee, P. E. Reddish, Nature 293, 293 (1981).
24. G. Bottini et al., Exp. Brain Res. 99, 164 (1994).
25. T. Brandt, M. Dieterich, A. Danek, Ann. Neurol. 35, 403
26. R. J. Peterka, C. C. Gianna-Poulin, L. H. Zupan, D. M.
Merfeld, J. Neurophysiol. 92, 2333 (2004).
27. S. Bense, T. Stephan, T. A. Yousry, T. Brandt, M. Dieterich,
J. Neurophysiol. 85, 886 (2001).
28. G. A. Orban et al., Neuropsychologia 41, 1757 (2003).
29. G. Rees, K. Friston, C. Koch, Nat. Neurosci. 3, 716 (2000).
30. H. Merchant, A. Battaglia-Mayer, A. P. Georgopoulos,
Cereb. Cortex 14, 314 (2004).
31. O. J. Grusser, M. Pause, U. Schreiter, J. Physiol. 430, 537
32. O. Blanke, S. Ortigue, T. Landis, M. Seeck, Nature 419,
33. V. Smetacek, Nature 415, 481 (2002).
34. R. N. Shepard, Psychol. Rev. 91, 417 (1984).
35. L. H. Zupan, D. M. Merfeld, C. Darlot, Biol. Cybern.
86, 209 (2002).
36. We thank R. Frackowiak, C. D. Frith, G. Galati, C. Porro,
G. Rizzolatti, and J. N. Sanes for comments and advice
on previous versions of this work. Supported by the
Italian Ministry of Health, the Italian Ministry of Uni-
versity and Research (FIRB and PRIN grants), and the
Italian Space Agency.
Supporting Online Material
Materials and Methods
Tables S1 to S5
References and Notes
29 November 2004; accepted 31 January 2005
Complement Factor H Variant
Increases the Risk of Age-Related
Jonathan L. Haines,1Michael A. Hauser,2Silke Schmidt,2
William K. Scott,2Lana M. Olson,1Paul Gallins,2Kylee L. Spencer,1
Shu Ying Kwan,2Maher Noureddine,2John R. Gilbert,2
Nathalie Schnetz-Boutaud,1Anita Agarwal,3Eric A. Postel,4
Margaret A. Pericak-Vance2*
Age-related macular degeneration (AMD) is a leading cause of visual impair-
ment and blindness in the elderly whose etiology remains largely unknown.
Previous studies identified chromosome 1q32 as harboring a susceptibility
locus for AMD. We used single-nucleotide polymorphisms to interrogate this
region and identified a strongly associated haplotype in two independent data
sets. DNA resequencing of the complement factor H gene within this haplo-
type revealed a common coding variant, Y402H, that significantly increases
the risk for AMD with odds ratios between 2.45 and 5.57. This common var-
iant likely explains È43% of AMD in older adults.
AMD causes progressive impairment of cen-
tral vision and is the leading cause of ir-
reversible vision loss in older Americans
(1). The most severe form of AMD involves
neovascular/exudative (wet) and/or atrophic
(dry) changes to the macula. Although the
etiology of AMD remains largely unknown,
implicated risk factors include age, ethnicity,
smoking, hypertension, obesity, and diet (2).
Familial aggregation (3), twin studies (4),
and segregation analysis (5) suggest that there
is also a substantial genetic contribution to
the disease. The candidate gene approach,
which focuses on testing biologically rele-
vant candidates, has implicated variants in
the ABCA4, FBLN6, and APOE genes as risk
factors for AMD. Replication of the ABCA4
and FBLN6 findings has been difficult, and
in toto these variants explain a small propor-
tion of AMD (6–8). The alternative genomic
approach uses a combination of genetic link-
age and association to identify previously
unknown genes involved in AMD. We par-
ticipated in a recent collaborative genomewide
linkage screen (9) in which chromosome 1q32
Table 1. CFH sequence variants identified in neovascular AMD cases and normal controls. All individuals
were homozygous for the AMD-associated GAGGT haplotype. The 24 affected individuals selected for
sequencing had severe neovascular disease (grade 5) (12) with diagnosis before age 74 (mean age at
diagnosis: 65.8 years). The 24 control individuals selected for sequencing had no evidence of AMD (grade
1) with age at exam after age 64 (mean age at exam: 69.8 years). The six previously identified SNPs are
labeled using standard nomenclature. The five previously unknown variants are labeled given their base
pair location on chromosome 1, Ensembl build 35. Five SNPs create nonsynonymous amino acid changes
within CFH, and five SNPs create synonymous changes. Exon 1 is not translated. n/a, not applicable.
Location SNP IDEffect
Minor allele frequency (%)
1Center for Human Genetics Research, Vanderbilt
University Medical Center, Nashville, TN 37232, USA.
2Center for Human Genetics and Department of
Medicine, Duke University Medical Center, DUMC
Box 3445, 595 LaSalle Street, Durham, NC 27710,
USA.3Vanderbilt Eye Institute, Vanderbilt University
Medical Center, Nashville, TN 37232, USA.
University Eye Center and Department of Ophthal-
mology, Duke University Medical Center, Durham,
NC 27710, USA.
*To whom correspondence should be addressed.
R E P O R T S
www.sciencemag.org SCIENCEVOL 30815 APRIL 2005
was identified as a likely region for an AMD
risk gene, a location also supported by other
studies (10, 11).
To identify the responsible gene on chro-
mosome 1q32, we initially genotyped 44
single nucleotide polymorphisms (SNPs) (12)
across the 24 megabases (Mb) incorporat-
ing this linkage region. We examined two
independent data sets: The first contained
182 families (111 multiplex and 71 discordant
sibpairs), and the second contained 495 AMD
cases and 185 controls. Each SNP was tested
for association independently in both data
sets. Two SNPs (rs2019724 and rs6428379)
in moderate linkage disequilibrium with each
other (r20 0.61) generated highly signifi-
cant associations with AMD in both the
family-based data set (rs2019724, P 0 0.0001;
rs6428379, P 0 0.0007) and in the case-
control data set (rs2019724, P G 0.0001;
rs6428379, P G 0.0001). These SNPs lie È263
kilobases (kb) apart.
To completely define the extent of linkage
disequilibrium, we genotyped an additional 17
SNPs across È655 kb flanked by rs1538687
and rs1537319 and encompassing the 263-kb
region. Two linkage disequilibrium blocks of
11 and 74 kb were identified and were sep-
arated by 176 kb (Fig. 1). The 11-kb block
contained rs2019724, and the 74-kb block con-
tained rs6428379. Association analysis of the
17 SNPs identified multiple additional SNPs
giving highly significant associations in one or
both of the family-based and case-control data
sets (Fig. 2). In the case-control data set, a
five-SNP haplotype (GAGGT, defined by SNPs
rs1831281, rs3753395, rs1853883, rs10494745,
and rs6428279, respectively) constituted 46%
of the case and 33% of the control chromo-
somes (P 0 0.0003). This same haplotype was
also significantly overtransmitted to affected
individuals in the family-based data set (P 0
0.00003). The convergence of the most sig-
nificant associations to this same haplotype in
the two independent data sets strongly sug-
gests that this region contains a commonly
inherited variant in an AMD risk gene.
The associated GAGGT haplotype spans
È261 kb. It contains the Complement Factor
H gene (CFH, OMIM #134370, accession
#NM_000186) and the five Complement
Factor H–related genes CFHL1 to CFHL5,
and lies within the Regulator of Complement
Activation (RCA) gene cluster. The most
consistent association results (Fig. 2) from
both the family-based and case-control data
sets converge within the CFH gene, implicat-
ing CFH as the AMD susceptibility gene. The
biological role of Complement Factor H as a
component of the innate immune system that
modulates inflammation through regulation of
complement (13) enhances its attractiveness
as a candidate AMD susceptibility gene.
Inflammation has been repeatedly implicated
in AMD pathology. C-reactive protein levels
are elevated in advanced disease (14), anti-
retinal autoantibodies have been detected in
AMD patients (15), macrophages are localized
near neovascular lesions (16), and the hallmark
drusen deposits contain many complement-
related proteins (17).
We screened for potential risk-associated
sequence variants in the coding region of
CFH by sequencing 24 cases with severe
neovascular disease and 24 controls with no
evidence of AMD. To maximize the likeli-
hood of identifying the risk-associated allele,
all sequenced cases and controls were homo-
zygous for the GAGGT haplotype. Five
previously unknown and six known sequence
variants were detected (Table 1). Only one
variant (rs1061170, sequence: T1277C; pro-
tein: Y402H) was present significantly more
often in cases than controls, occurring on 45
of 48 haplotypes in the cases and on 22 of 48
Fig. 1. Haploview plot defining haplotype block structure of AMD-associated region. The relative
physical position of each SNP is given in the upper diagram, and the pairwise linkage
disequilibrium (D¶) between all SNPs is given below each SNP combination. Dark red–shaded
squares indicated D¶ values 9 0.80. D¶ 0 1.0 when no number is given.
Fig. 2. Plot of family-based and case-control P values for all SNPs within the AMD-associated
haplotype. The genomic region spanning each gene is indicated in green. jlog10of the nominal
P values are plotted for each SNP. Results for both the family-based and case-control data sets
converge within the CFH gene.
R E P O R T S
15 APRIL 2005 VOL 308SCIENCEwww.sciencemag.org
haplotypes in the controls (P G 0.0001). The Download full-text
frequency of sequence variants within the
CFH coding region on the associated haplo-
type was significantly reduced in cases com-
pared to controls (12% versus 18%, P 0 0.002).
When the overrepresented T1277C variant
was removed from the analysis, this differ-
ence became more pronounced (3% versus
16%, P G 0.00001). Thus, T1277C is the pri-
mary DNA sequence variant differentiating
between the case and control haplotypes.
Complete genotyping of T1277C in the
family-based and case-control data sets re-
vealed a significant overtransmission in the
families (P 0 0.019) (12) and a highly sig-
nificant overrepresentation in the cases com-
pared to controls (P 0 0.00006). The odds
ratio for AMD was 2.45 E95% confidence
interval (CI): 1.41 to 4.25^ for carriers of one
C allele and 3.33 (95% CI: 1.79 to 6.20) for
carriers of two C alleles. When the analysis
was restricted to only neovascular AMD,
these odds ratios increased to 3.45 (95% CI:
1.72 to 6.92) and 5.57 (95% CI: 2.52 to
12.27), respectively. This apparent dose ef-
fect for risk associated with the C allele was
highly significant (P G 0.0001). There was
no apparent allelic or genotypic effect of
T1277C on age at AMD diagnosis (mean age
at diagnosis: TT, 76.5 years; TC, 77.5 years;
CC, 75.5 years). The population attributable
risk percent for carrying at least one C allele
was 43% (95% CI: 23 to 68%).
The Y402H variant is predicted to have
functional consequences consistent with AMD
pathology. Residue 402 is located within
binding sites for heparin (18) and C-reactive
protein (CRP) (19). Binding to either of these
partners increases the affinity of CFH for the
complement protein C3b (20, 21), augmenting
its ability to down-regulate complement_s ef-
fect. The observed colocalization of CFH,
CRP, and proteoglycans in the superficial
layer of the arterial intima suggests that CFH
may protect the host arterial wall from excess
complement activation (22). We hypothesize
that allele-specific changes in the activities of
the binding sites for heparin and CRP would
alter CFH_s ability to suppress complement-
related damage to arterial walls and might ul-
timately lead to vessel injury and subsequent
neovascular/exudative changes such as those
seen in neovascular AMD. Our data support
this hypothesis, because the risk associated
with the C allele is more pronounced when
the analyses are restricted to neovascular
AMD. Given the known functional interac-
tions of genes within the RCA gene cluster
(13), variants within these genes could inter-
act with or modify the effect of the T1277C
Plasma levels of CFH are known to de-
crease with smoking (23), a known risk
factor for AMD (2). This confluence of
genetic and environmental risk factors sug-
gests an integrated etiological model of
AMD involving chronic inflammation. Iden-
tification of the increased risk of AMD
associated with the T1277C variant should
enhance our ability to develop presympto-
matic tests for AMD, possibly allowing
earlier detection and better treatment of this
References and Notes
1. Centers for Disease Control and Prevention (CDC),
Morb. Mortal. Wkly. Rep. 53, 1069 (2004).
2. J. Ambati, B. K. Ambati, S. H. Yoo, S. Ianchulev, A. P.
Adamis, Surv. Ophthalmol. 48, 257 (2003).
3. C. C. Klaver et al., Arch. Ophthalmol. 116, 1646 (1998).
4. C. J. Hammond et al., Ophthalmology 109, 730
5. I. M. Heiba, R. C. Elston, B. E. Klein, R. Klein, Genet.
Epidemiol. 11, 51 (1994).
6. E. M. Stone et al., Nat. Genet. 20, 328 (1998).
7. S. Schmidt et al., Ophthalmic Genet. 23, 209 (2002).
8. D. W. Schultz et al., Hum. Mol. Genet. 12, 3315
9. D. E. Weeks et al., Am. J. Hum. Genet. 75, 174
10. G. R. Abecasis et al., Am. J. Hum. Genet. 74, 482
11. S. K. Iyengar et al., Am. J. Hum. Genet. 74, 20 (2004).
12. Materials and methods are available as supporting
material on Science Online.
13. S. Rodriguez de Cordoba, J. Esparza-Gordillo, E. Goicoechea
de Jorge, M. Lopez-Trascasa, P. Sanchez-Corral, Mol.
Immunol. 41, 355 (2004).
14. J. M. Seddon, G. Gensler, R. C. Milton, M. L. Klein,
N. Rifai, J. Am. Med. Assoc. 291, 704 (2004).
15. D. H. Gurne, M. O. Tso, D. P. Edward, H. Ripps,
Ophthalmology 98, 602 (1991).
16. M. C. Killingsworth, J. P. Sarks, S. H. Sarks, Eye 4, 613
17. R. F. Mullins, S. R. Russell, D. H. Anderson, G. S.
Hageman, FASEB J. 14, 835 (2000).
18. T. K. Blackmore, V. A. Fischetti, T. A. Sadlon, H. M.
Ward, D. L. Gordon, Infect. Immun. 66, 1427 (1998).
19. E. Giannakis et al., Eur. J. Immunol. 33, 962 (2003).
20. D. T. Fearon, Proc. Natl. Acad. Sci. U.S.A. 75, 1971
21. C. Mold, M. Kingzette, H. Gewurz, J. Immunol. 133,
22. R. Oksjoki et al., Arterioscler. Thromb. Vasc. Biol. 23,
23. J. Esparza-Gordillo et al., Immunogenetics 56, 77 (2004).
24. We thank all of the study participants and their rela-
tives; M. de la Paz, M. Klein, J. Caldwell, R. Domurath,
K. Haynes, V. Mitchell, M. Shaw, and J. Galloway for
participant ascertainment; K. Abramson, J. Benton, W.
Lambert, B. Love, T. Skelly, E. Tegnell, M. Allen, C.
Haynes, R. Chung, and J. Bunch for valuable technical
assistance; J. M. Vance and M. Summar for critical
reading of the manuscript; and D. J. M. Gass for
patient ascertainment and clinical expertise. We also
thank the following clinics and clinicians for referring
individuals to the study: Southern Retina, L.L.C (C. Harris);
Vitreo-Retinal Surgeons (M. Duan and C. Devine); Georgia
Retina, P.C.; and The Retina Group of Washington.
Supported by grants EY12118 (to M.A.P.-V. and J.L.H.)
and EY015216 (to S.S.) from the NIH/National Eye
Institute, grant AG11268 from the NIH/National Institute
on Aging (to H. Cohen), and grant M01 RR-00095 from
the NIH/National Center for Research Resources (to
Supporting Online Material
Materials and Methods
28 January 2005; accepted 23 February 2005
Published online 10 March 2005;
Include this information when citing this paper
Complement Factor H
Polymorphism and Age-Related
Albert O. Edwards,1*. Robert Ritter III,1Kenneth J. Abel,2
Alisa Manning,3Carolien Panhuysen,3,6Lindsay A. Farrer3,4,5,6,7
Age-related macular degeneration (AMD) is a common, late-onset, and com-
plex trait with multiple risk factors. Concentrating on a region harboring a
locus for AMD on 1q25-31, the ARMD1 locus, we tested single-nucleotide
polymorphisms for association with AMD in two independent case-control
populations. Significant association (P 0 4.95 ? 10j10) was identified
within the regulation of complement activation locus and was centered
over a tyrosine-402 Y histidine-402 protein polymorphism in the gene en-
coding complement factor H. Possession of at least one histidine at amino
acid position 402 increased the risk of AMD 2.7-fold and may account for
50% of the attributable risk of AMD.
AMD is a leading cause of blindness in older
individuals (1). It is a late-onset, complex
trait with hereditary, lifestyle, and medical
risk factors (2). The condition typically pre-
sents in the fifth decade of life with small
yellow deposits external to the outer retina
and retinal pigment epithelium (RPE) called
drusen. Large numbers of drusen and clini-
cal features of damage to the RPE markedly
increase the risk of complications (atrophy
of the RPE and abnormal neovasculariza-
tion of the outer retina), leading to severe
vision loss (1).
Although the primary pathogenic mecha-
nisms of AMD were previously unknown,
there is strong evidence that genetics plays a
role (3–9). The first locus for AMD (ARMD1)
was reported in a single extended family linked
to chromosome 1q25.3-31.3 (5). Because there
was strong evidence for linkage to this region
R E P O R T S
www.sciencemag.orgSCIENCE VOL 30815 APRIL 2005