A genome-wide association scan identifies the hepatic cholesterol transporter ABCG8 as a susceptibility factor for human gallstone disease

Abstract

With an overall prevalence of 10-20%, gallstone disease (cholelithiasis) represents one of the most frequent and economically relevant health problems of industrialized countries. We performed an association scan of >500,000 SNPs in 280 individuals with gallstones and 360 controls. A follow-up study of the 235 most significant SNPs in 1,105 affected individuals and 873 controls replicated the disease association of SNP A-1791411 in ABCG8 (allelic P value P(CCA) = 4.1 x 10(-9)), which was subsequently attributed to coding variant rs11887534 (D19H). Additional replication was achieved in 728 German (P = 2.8 x 10(-7)) and 167 Chilean subjects (P = 0.02). The overall odds ratio for D19H carriership was 2.2 (95% confidence interval: 1.8-2.6, P = 1.4 x 10(-14)) in the full German sample. Association was stronger in subjects with cholesterol gallstones (odds ratio = 3.3), suggesting that His19 might be associated with a more efficient transport of cholesterol into the bile.

Full text

A genome-wide association scan identifies the hepatic
cholesterol transporter ABCG8 as a susceptibility factor
for human gallstone disease
Stephan Buch
1–3,13
, Clemens Schafmayer
3,4,13
, Henry Vo
¨
lzke
5
, Christian Becker
6,7
, Andre Franke
2
,
Huberta von Eller-Eberstein
3
, Christian Kluck
6,7
, Ingelore Ba
¨
ssmann
6,7
, Mario Brosch
1
, Frank Lammert
8
,
Juan Francisco Miquel
9
, Flavio Nervi
9
, Michael Wittig
2
, Dieter Rosskopf
10
, Birgit Timm
3
, Christine Ho
¨
ll
3
,
Marcus Seeger
1
, Abdou ElSharawy
2
,TimLu
11
, Jan Egberts
4
, Fred Fa
¨
ndrich
4
, Ulrich R Fo
¨
lsch
1
,
Michael Krawczak
3,11
, Stefan Schreiber
2,3
, Peter Nu
¨
rnberg
6,12
,Ju
¨
rgen Tepel
4
& Jochen Hampe
1
With an overall prevalence of 10–20%, gallstone disease
(cholelithiasis) represents one of the most frequent and
economically relevant health problems of industrialized
countries
1,2
. We performed an association scan of
4500,000 SNPs in 280 individuals with gallstones and
360 controls. A follow-up study of the 235 most significant
SNPs in 1,105 affected individuals and 873 controls replicated
the disease association of SNP A-1791411 in ABCG8
(allelic P value P
CCA
¼ 4.1  10
–9
), which was subsequently
attributed to coding variant rs11887534 (D19H). Additional
replication was achieved in 728 German (P ¼ 2.8  10
–7
)
and 167 Chilean subjects (P ¼ 0.02). The overall odds ratio
for D19H carriership was 2.2 (95% confidence interval:
1.8–2.6, P ¼ 1.4  10
–14
) in the full German sample.
Association was stronger in subjects with cholesterol
gallstones (odds ratio ¼ 3.3), suggesting that His19 might
be associated with a more efficient transport of cholesterol
into the bile.
Genetic susceptibility to gallstones was first recognized in the 1930s
(ref. 3). Both familial clustering
4
and an increased concordance rate of
the trait in monozygotic twins
5
have since been reported. Genome-
wide studies of gallstone susceptibility in inbred mouse strains
established a number of Lith loci
6–9
. In humans, a genome-wide
linkage analysis of Mexican American families identified a gallstone
susceptibility locus on chromosome 1p
10
, and several suggestive
linkage signals overlapped with the mouse Lith loci.
For the present study, we genotyped over 500,000 SNPs in case-
control panel A (Ta ble 1) using the Affymetrix 500K array. Stone-free
controls were chosen so that the controls had a higher median age (61
years) than the affected individuals (42 years), bearing in mind that
gallstone prevalence increases with age
2
. Genotyping was successful for
464,585 SNPs, as defined by a call rate of 493% using the BRLMM
algorithm. The 382,492 SNPs with a minor allele frequency Z2% and
no significant departure from Hardy-Weinberg equilibrium (HWE)
(P
HWE
4 0.01 in controls) were analyzed further. We estimated that
the genomic control inflation factors in panels A and A* amounted to
a negligible 1.03 and 1.04, respectively, using PLINK (Supplementary
Fig. 1 online).
Association analyses were performed twice: once using all panel A
samples (Tabl e 1) and once using only a subset of affected individuals
(N ¼ 96) and controls (N ¼ 205), denoted panel A*, with a
DM-algorithm call rate Z93% over the whole array. An age at
diagnosis o50 years (median: 29 years) was also required for inclu-
sion into panel A*. A total of 235 SNPs were selected for SNPlex-based
follow-up, a number that was chosen to match the available laboratory
resources. Follow-up markers were taken from one of two distinct
categories. For category I (N ¼ 202), SNPs were first ranked according
to the P value of the genotypic (P
CCG
)orallelic(P
CCA
)case-control
comparison in panels A and A*, and the top-ranking SNPs were
deemed potential lead markers. For a SNP to qualify as an actual lead,
at least one additional association signal, defined as a P value less than
two orders of magnitude larger than that of the possible lead marker,
was required within a distance of 50 kb. As a result, we included lead
© 2007 Nature Publishing Group http://www.nature.com/naturegenetics
Received 23 February; accepted 12 June; published online 15 July 2007; doi:10.1038/ng2101
1
First Department of Medicine,
2
Institute for Clinical Molecular Biology,
3
POPGEN Biobank and
4
Department of General and Thoracic Surgery, University Hospital
Schleswig-Holstein, 24105 Kiel, Germany.
5
Institute for Community Medicine, University Hospital Greifswald, Walther Rathenau Str. 48, 17487 Greifswald, Germany.
6
Cologne Center for Genomics, University of Cologne, Zu
¨
lpicher Strasse 47, 50674 Cologne, Germany.
7
RZPD German Resource Center for Genome Research,
Heubnerweg 6, 14059 Berlin, Germany.
8
Department of Internal Medicine I, University Hospital Bonn, Sigmund Freud-Strasse 25, 53105 Bonn, Germany.
9
Departamento de Gastroenterologia, Facultad de Medicina, Pontificia Universidad Cato
´
lica de Chile, Santiago, Chile.
10
Institute of Pharmacology, Ernst-Moritz-Arndt
University Greifswald, Friedrich Loeffler Str. 23d, 17487 Greifswald, Germany.
11
Institute of Medical Statistics and Informatics, University Hospital Schleswig-
Holstein, 24105 Kiel, Germany.
12
Center for Molecular Medicine Cologne (CMMC), University of Cologne, Joseph-Stelzmann-Strasse 52, 50931 Ko
¨
ln, Germany.
13
These authors contributed equally to this work. Correspondence should be addressed to J.H. (jhampe@1med.uni-kiel.de).
NATURE GEN ETICS ADVAN CE ONLINE PUBLICATION 1
LETTERS

SNPs with P values of up to 5  10
–4
in the follow-up. The screening
stage of our study served to ascertain markers for further study
(Supplementary Fig. 2 online) rather than assessing the significance
of their potential disease association. The P value of SNP A-1791411,
corrected for multiple testing (2  382,492 tests), was 0.80, under-
lining the need for independent replication. However, to offset the
possible disregard of regions with low marker density on the 500K
array, a second marker category (II, N ¼ 33) was created after ranking
and prioritizing the remaining SNPs by their allelic and genotypic
P values alone. The full list of follow-up SNPs is provided in
Supplementary Table 1 online, and a subset is shown in Tabl e 2 .
The 235 SNPs selected for follow-up were genotyped in an
independent sample of 1,105 symptomatic gallstone carriers and 873
independent, sex-matched controls (panel B). We used the following
as criteria for formal replication: a P value of 2  10
–4
(corresponding
to P ¼ 0.05 after Bonferroni correction for 235 tests) in the allelic and
genotypic tests as well as in the logistic regression analysis controlling
for age, sex, BMI and diabetes. Only the disease association of the SNP
with Affymetrix identifier A-1791411 in the ABCG8 gene was formally
replicated. After Bonferroni correction, we found P
CCA
¼ 1.1  10
–4
,
P
CCG
¼ 4.1  10
–4
and P
regression
¼ 9.4  10
–3
. We had included
marker A-1791411 in the follow-up because of its significant disease
association in panel A*. In panel A, its call rate was only 87%.
Therefore, we confirmed the chip-based genotyping results for
A-1791411 with TaqMan (98% concordance, retrospective allelic
P ¼ 5.7  10
–6
for panel A).
We selected additional SNPs covering the ABCG8 and ABCG5 gene
region from HapMap data from individuals of European ancestry
(minor allele frequency Z 3%, pairwise r
2
Z 0.8, P
HWE
Z 0.05) using
Haploview
11
. When we genotyped these SNPs in panel B, the most
significant disease association was found for rs4299376 in intron 2 of
the ABCG8 gene (Fig. 1), with P
CCA
¼ 7.5  10
–9
and P
CCG
¼ 1.2 
10
–9
(Supplementary Table 2 online).
ABCG5 and ABCG8 have been the subject of extensive mutation
detection efforts
12–14
. Starting from all known coding SNPs in these
two genes in dbSNP and the literature, we genotyped those with a
minor allele frequency 41% in the general population of European
ancestry (Supplementary Table 2). The coding SNPs responsible for
amino acid changes L36P, Q340E, M429V and G575R were mono-
morphic. For the sake of completeness, we also included two sitostero-
lemia SNPs: the SNP in ABCG8 responsible for premature stop
R121X, which was monomorphic, and the SNP responsible for
premature stop W361X in ABCG8, for which we observed one
heterozygote each in panel B controls and affected individuals. Using
P o 0.01 in both the allelic and the genotypic test as a criterion for
statistical significance, only the SNP responsible for D19H in
the ABCG8 gene was significantly associated with gallstone disease
(P
CCA
¼ 8.8  10
–10
, P
CCG
¼ 3.0  10
–9
in panel B). We confirmed
SNPlex genotyping of the SNP responsible for D19H using a TaqMan
assay (100% genotype concordance in panel B). In a logistic regression
analysis of panel B, no other SNP significantly improved the statistical
model in the presence of D19H (P 4 0.05). Haplotype analysis
(Ta bl e 3) further showed that all significant risk haplotypes indeed
carried the H allele of this SNP. We used an additional panel (C) of
independent cases and controls partly overlapping with panel B
(Ta bl e 1 ) for the fine mapping and replication of the D19H associa-
tion. The strength and localization of the association signal, as well as
the haplotype assignment of D19H, were replicated in a panel of
728 independent German affected individuals (panel C, Ta ble 1 ,
P
CCA
¼ 2.8  10
–7
, P
CCG
¼ 9.2  10
–7
) and in 167 affected individuals
and 167 controls from Chile (panel D, Tab le 1, P
CCA
¼ 0.02).
In a combined post hoc analysis of all German affected individuals,
we estimated the odds ratio to be 7.1 (95% confidence interval (c.i.),
0.9–158.6) for HH homozygosity and 2.2 (95% c.i., 1.8–2.6) for
carriership of the H allele. The latter corresponds to a population
attributable risk of B11%. In the Chilean sample, the 19H carriership
odds ratio was 1.9. A sample of individuals with gallstones from
Greifswald, which included an approximately equal numbers of
symptomatic (N ¼ 374) and asymptomatic (N ¼ 354) gallstone
carriers, demonstrated a disease association of 19H carriership for
both groups of affected individuals, with odds ratios of 2.5 (P ¼ 4.8 
10
–6
)and2.4(P ¼ 1.7  10
–6
), respectively. From panels A–C and the
Chilean sample, the combined significance of rs11887534 was
calculated as P ¼ 7.74  10
–9
by Fisher’s technique, taking into
account multiple testing in panel B and excluding overlapping controls
from panel C.
We performed a logistic regression analysis in all German affected
individuals (N ¼ 2,113) and controls (N ¼ 1,524, including 129
hitherto unused male controls) for which information on age, sex,
BMI and type II diabetes status was available. Upon adjustment for
these covariates, the effect of 19H carriership remained highly sig-
nificant (Wald P o 0.0001) and was similar to that seen in the
unadjusted analysis (adjusted odds ratio ¼ 2.2). Furthermore, none of
the interaction terms was significant in the presence of D19H. For 495
affected individuals in panels A and B, stone composition data were
available from previous FTIR spectroscopy
15
. The odds ratio of 19H
carriership was 3.3 (95% c.i., 2.5–4.3) for stones with cholesterol as a
major constituent, defined by a cholesterol content 430% (wt/wt). In
contrast, allele 19H was completely lacking in the 12 affected indivi-
duals with pure pigment stones.
Mutations in ABCG8 have been shown to be responsible for
sitosterolemia, a rare autosomal recessive disorder caused by the
excessive accumulation of plant sterols
16
. In view of the low prevalence
of sitosterolemia, with fewer than 200 affected families reported
© 2007 Nature Publishing Group http://www.nature.com/naturegenetics
Table 1 Summary of study populations
Affected individuals Controls
Panel N Median age Age at diagnosis N Age Percentage male Percentage symptomatic
Screening panel (panel A) 280 42 38 360 61 50% 86
Replication I and fine mapping (panel B) 1,105 54 48 873 63 28% 100
Replication II and fine mapping (panel C) 728 60 n/a 732
a
65 40% 51
Replication III (Chile) (panel D) 167 54 n/a 167 48 0% 72
All samples were sex matched. All subjects in panel A* were symptomatic with gallstone disease. German affected individuals and controls were recruited in northern Germany
through clinical centers at Kiel University, the POPGEN project and the SHIP project at Greifswald University. No notable population genetic differences exist between these two
regions
30
, so we combined patient samples into genotyping panels on the basis of phenotypic criteria alone. Ages are given in years.
a
Controls in this panel include 570 individuals randomly selected from panel B to match the age and sex distribution of the affected individuals.
2 ADVANCE ONLINE PUBLICATION NATURE GENETICS
LETTERS

worldwide, it is clear that causative mutations for this mendelian
disorder cannot at the same time account for 10% of individuals with
a frequent condition such as gallstone disease. Indeed, the previously
characterized sitosterolemia mutations either were not found in our
study subjects or occurred at the same frequency in affected indivi-
duals and controls, thereby confirming that gallstone disease is allelic,
but not genetically identical, to sitosterolemia. On the contrary,
sitosterolemia may even confer some resistance to gallstone disease,
bearing in mind that individuals with sitosterolemia secrete less
cholesterol into the bile than expected from their increased levels of
intestinal cholesterol absorption
17
.
Logistic regression analysis using BMI, sex, age and type II diabetes
status as covariates showed that the observed genetic gallstone risk was
independent of these potential confounders
18
. This result corroborates
previous epidemiological evidence that gallstone susceptibility cannot
be explained by established risk factors for gallstone disease, such as
obesity, for example
18–20
.
A number of human linkage signals
10
colocalize with the association
leads (Tabl e 2 and Supplementary Table 1). The replicated marker
A-1791411 is not localized in one of the linkage regions of the Mexican
American scan. This lack of concordance may be due to population
differences or may reflect the lower power of nonparametric linkage
analysis for a relatively infrequent genetic risk variant like D19H. This
also underlines the need for further studies to elucidate the nature of
the remaining genetic risk in gallstone disease (Supplementary Note
online). With hindsight, the identification of the ABCG5/ABCG8 locus
as a risk factor for human gallstone disease seems almost too obvious
to have required a genome-wide association scan, because the human
ABCG5/ABCG8 locus is homologous to a previously reported mouse
susceptibility locus (Lith9) detected in strain PERA/Ei
9
. However,
earlier studies by our group investigating candidate human genes
© 2007 Nature Publishing Group http://www.nature.com/naturegenetics
Table 2 Replication study (panel B) of the top 235 SNPs from the genome-wide association scan (panels A and A*)
Screening (panel A) Screening (panel A*) Panels A and A* HapMap Replication (panel B)
# Panel dbSNP ID CR P
CCA
P
CCG
CR P
CCA
P
CCG
MAF
(ctrl)
MAF
(case) MAF CR P
CCA
P
CCG
MAF
(ctrl)
MAF
(case) P
REG
24 A* rs6716275
a
0.93 5.12  10
–4
2.89  10
–3
0.97 8.06  10
–5
4.09  10
–4
0.34 0.52 0.41 1 1.67  10
–1
4.15  10
–2
0.43 0.41 5.75  10
–1
25 A* rs2194448
b
0.97 1.15  10
–1
1.82  10
–1
0.99 2.60  10
–4
1.17  10
–4
0.47 0.63 0.47 1 6.72  10
–1
7.83  10
–1
0.49 0.48 8.61  10
–1
26 A rs1550992
b
0.95 3.42  10
–4
1.23  10
–3
0.98 2.77  10
–1
5.07  10
–1
0.21 0.30 0.35 1 6.03  10
–2
1.27  10
–1
0.21 0.24 9.52  10
–2
27 A rs11124408
b
17.15 10
–5
1.64  10
–4
13.55 10
–4
1.49  10
–3
0.09 0.16 0.18 1 2.06  10
–2
2.28  10
–2
0.10 0.13 3.00  10
–2
28 A* rs7571463
b
0.92 6.94  10
–1
7.88  10
–1
0.94 4.91  10
–4
1.81  10
–3
0.23 0.11 0.27 1 5.32  10
–3
2.03  10
–2
0.18 0.21 8.62  10
–3
29 A* n/a
b,c
0.87 5.30  10
–5
6.01  10
–6
0.94 2.11  10
–6
9.84  10
–6
0.05 0.18 n/a
d
1 4.07  10
–9
1.53  10
–8
0.05 0.10 3.43  10
–7
30 A* rs370068 1 1.48  10
–1
1.03  10
–1
13.04 10
–4
6.09  10
–4
0.30 0.45 0.27 1 7.85  10
–1
4.95  10
–1
0.33 0.33 7.45  10
–1
31 A* rs4852395 0.99 6.43  10
–2
5.30  10
–2
11.23 10
–3
3.56  10
–4
0.16 0.27 0.18 0.99 1.15  10
–1
2.86  10
–1
0.15 0.17 3.02  10
–1
32 A rs1517862 13.87 10
–4
2.95  10
–4
11.35 10
–2
1.20  10
–2
0.02 0.06 0.08 1 6.45  10
–1
4.47  10
–1
0.03 0.04 8.49  10
–2
SNPs fell into one of two categories: those selected on the basis of additional flanking marker support (I) and those selected based upon the P value of the lead marker alone (II). Note, all SNPs in this table are category I and are located
on chromosome 2. The analysis was further stratified by whether a SNP was found to rank among the most significant markers in panel A or A*. Markers are reported by chromosome and nucleotide position (NCBI build 35). The call rate
(CR) and P values for the genotype-based (P
CCG
) and allele-based case-control comparisons (P
CCA
) are reported alongside the minor allele frequencies (MAF) in each sample and in the HapMap CEU category, for comparison. Boldface
indicates results that led to the inclusion of a marker in the follow-up. Markers with P o 0.0002 in all of the allele-, genotype- and regression-based tests in panel B are underlined. For replication panel B, a genotypic logistic regression
P value (P
REG
) controlling for age, sex, BMI and type II diabetes status is included. Only markers neighboring A-1791411 are shown; the full table is available as Supplementary Table 1.
a
Localizes to human chromosome 2p24.2, corresponding to one of the linkage regions reported in ref. 10.
b
Corresponds to the mouse Lith locus Lith9, as reported in refs. 6–9.
c
No ‘rs’ identifier exists (probe set ID A-1791411).
d
No HapMap MAF available,
because no ‘rs’ identifier assignment exists.
44.02 Mb
121110 1398765432112345678910111213
ABCG8ABCG5
chr2
43.95 Mb
rs2954804
rs2954802
rs6544718
rs11124950
rs4148221
rs4148222
rs4148219
rs17031742
rs4953028
rs13405698
rs4953027
rs4245795
rs12468591
rs4148217
rs17409589
rs10174731
*
rs10221914
rs4245791
rs4299376
rs10177200
rs3806470
rs4131229
rs4289236
rs4073237
rs4245786
rs1864814
rs10439467
rs10201851
rs10205816
rs10208987
rs4148192
rs4148194
rs6720173
rs10180615
rs17424122
*
rs4953023*
rs4148211*
rs11887534*
rs4148187*
A-1791411*
Figure 1 Overview of the physical and genetic structure of the ABCG5/
ABCG8 region. The physical position of the investigated SNPs and a
schematic representation of the gene structures are shown above. The
D19H variant is marked by an arrow. Coordinates refer to NCBI genome
assembly build 35. Below is the linkage disequilibrium structure of the
locus (D¢), as generated by Haploview
11
from the control genotypes of
panel B. Haplotype blocks derived from the HapMap genotypes from
individuals of European ancestry (CEU) are outlined in black in the
D¢ chart. SNPs used in the haplotype analysis (see Table 3)arein
boldface and are marked by an asterisk (*).
NATURE GEN ETICS ADVAN CE ONLINE PUBLICATION 3
LETTERS

orthologous to mouse Lith1 (ref. 21) and Lith6 (ref. 22) loci were
consistently disappointing. Nevertheless, the ABCG5/ABCG8 region
clearly represented a prime candidate after Abcg5 and Abcg8 (both
located within Lith9) were suggested as possible mouse candidate
genes
9,23
. Furthermore, variants in the ABCG5/ABCG8 genes have
been reported recently to be associated with different plasma lipid
levels in siblings with gallstones, suggesting a potential alteration of
bile cholesterol levels by the ABCG5/ABCG8 genotype
24
.
Although the genetic gallstone disease risk is clearly attributable to
variant D19H in the ABCG8 gene in our study, we cannot exclude the
possibility that other as-yet-unidentified variants at the ABCG5/
ABCG8 locus might also contribute to gallstone susceptibility. The
ABCG8 protein transports cholesterol into both the biliary and the gut
lumen and has been extensively characterized after its identification as
the causative gene for sitosterolemia (see ref. 12 for a recent review).
Others
25
have observed a lower concentration of campesterol in 19H
carriers and hypothesized that the H allele leads to an increased
transporter activity of ABCG8. This hypothesis received further
support by the observation that 19H carriership is associated with
lower total serum cholesterol levels
24,26
. Hypothetically, 19H carrier-
ship may thus increase the efficiency of the cholesterol transport into
the bile lumen, causing cholesterol hypersaturation of the bile and
eventually promoting the formation of cholesterol gallstones
27
.How-
ever, additional functional studies are needed to determine the effect
of the D19H variant on the ABCG8/ABCG5 heterodimer.
METHODS
Study subjects and procedures. Written informed consent was obtained from
all study participants. Study protocols were approved by the ethics committees
of the Kiel and Greifswald University Hospitals. Details of the recruitment and
genotyping technique are provided in Supplementary Methods online.
Statistical analysis. All markers were tested for a possible deviation from HWE
in the controls before inclusion. Single-marker association tests were performed
using Haploview
11
and GENOMIZER
28
with a w
2
test or Fisher’s exact test for
contingency tables, as appropriate. Haplotype frequency estimates were
obtained from singletons using COCAPHASE
29
. Significance testing of haplo-
type frequency differences was performed on the basis of the COCAPHASE
results, making use of the fact that twice the log-likelihood ratio between two
nested data models approximately follows a w
2
distribution with k degrees of
freedom, where k is the difference in the number of parameters between the two
models. Logistic regression analysis was performed with SPSS version 11.0,
coding individual SNP genotypes as categorical variables.
Note: Supplementary information is available on the Nature Genetics website.
ACKNOWLEDGMENTS
The cooperation of all patients, their families and physicians is gratefully
acknowledged. The authors gratefully acknowledge the support by the following
heads of surgical departments: I. Vogel (Sta
¨
dtisches Krankenhaus Kiel),
H. Dittrich (Rendsburg), J. Belz (Husum), R. Qua
¨
schling (Eckernfo
¨
rde),
H. Shekarriz (Schleswig), V. Mendel (Flensburg), W. Neugebauer (Flensburg),
F. Kallinowski (Heide), J. Klima (Niebu
¨
ll), M. Sailer (Hamburg) and
A. Schafmayer (Lu
¨
neburg). Special thanks are given to C. Fu
¨
rstenau, T. Wesse,
B. Petersen, L. Bossen, T. Henke, S. Ehlers, A. Dietsch and V. Pucken for technical
assistance. This study was supported by the German Ministry of Education and
Research through the POPGEN biobank project (01GS0426, 01GR0468), the
MediGrid project and the National Genotyping Platforms in Kiel and Cologne
and by the German Research Council (Ha 3091/2-1, 4-1, La 997/3-1), Applied
Biosystems, Mucosaimmunologie gGmbH and the Medical Faculty Kiel. The
SHIP recruitment project is funded by the Federal Ministry of Education and
Research (ZZ9603), the Ministry of Cultural Affairs and the Social Ministry of the
Federal State of Mecklenburg-West Pomerania. The Chilean study was supported
by grants from FONDECYT (Fondo Nacional de Desarrollo Cientı
´
fico y
Te c n o l o g o
´
gico) (numbers 1040820 (to J.F.M.) and 1030744 (to F.N.)).
AUTHOR CONTRIBUTIONS
S.B. performed the SNP selection, genotyping and data analysis and prepared the
figures and tables; C.S. coordinated the Kiel recruitment, phenotyped patients and
helped write the paper; J.E., H.v.E., C.H., B.T. and M.S. recruited patients and
helped write the paper. C.B., I.B., C.K. and P.N. performed the chip genotyping
and chip data analysis; H.V. and D.R. coordinated the SHIP recruitment and
participated in experimental design; J.F.M., F.N. coordinated the recruitment in
Chile and participated in experimental design; A.F., M.B., A.E., T.L. and M.W.
helped with data analysis and genotyping and F.L., F.F., U.R.F., S.S., P.N. and J.T.
helped design the experiment, supported recruitment and helped write the paper.
M.K. supervised and performed the statistical analysis and edited the paper. J.H.
designed and supervised the experiment, performed data analysis and wrote the
manuscript. All authors have revised the manuscript for intellectual content.
COMPETING INTERESTS STATEMENT
The authors declare no competing financial interests.
Published online at http://www.nature.com/naturegenetics
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions
1. Lammert, F. & Sauerbruch, T. Mechanisms of disease: the genetic epidemiology of
gallbladder stones. Nat. Clin. Pract. Gastroenterol. Hepatol. 2, 423–433 (2005).
2. Volzke, H. et al. Independent risk factors for gallstone formation in a region with high
cholelithiasis prevalence. Digestion 71,97–105(2005).
3. Ko
¨
rner, G. U
¨
ber die famila
¨
re Ha
¨
ufung der Gallenblasenkrankheiten. Zeitschrift fu
¨
r
Menschliche Vererbungs- und Konstitutionslehre 20, 528–582 (1937).
4. Nu
¨
rnberg, D., Berndt, H. & Pannwitz, H. Familia
¨
re Ha
¨
ufung von Gallensteinen. Dtsch.
med. Wschr. 114, 1059–1063 (1989).
5. Katsika, D. et al. Genetic and environmental influences on symptomatic gallstone
disease: a Swedish study of 43,141 twin pairs. Hepatology 41, 1138–1143
(2005).
© 2007 Nature Publishing Group http://www.nature.com/naturegenetics
Table 3 Haplotype analysis of seven SNPs at the ABCG8 locus in panels B and C
Fine mapping, panel B Fine mapping, panel C
Haplotype f
cases
f
controls
OR
case-control
P
COCAPHASE
f
cases
f
controls
OR
case-control
P
COCAPHASE
C-G-A-T-G-T-C 0.340 0.382 0.8 0.01 0.349 0.392 0.8 0.02
T-G-G-T-G-T-G 0.342 0.325 1.1 0.24 0.340 0.325 1.1 0.38
C-G-G-T-G-T-G 0.066 0.071 0.9 0.53 0.077 0.069 1.1 0.42
C-G-A-T-G-T-G 0.067 0.068 1.0 0.89 0.060 0.068 0.9 0.33
C-
C-A-C-A-T-C 0.097 0.045 2.3 7.75  10
–7
0.080 0.038 2.2 8.76  10
–7
C-G-A-T-G-A-C 0.023 0.043 0.5 5.73  10
–4
0.028 0.042 0.7 0.03
T-G-A-T-G-A-C 0.025 0.033 0.8 0.12 0.020 0.033 0.6 0.03
T-G-A-T-G-A-C 0.028 0.024 1.2 0.53 0.036 0.028 1.3 0.305
SNPs included in the haplotype analysis (rs4148187, rs11887534 (D19H), rs4148211 (Y54C), SNP_A-1791411, rs4953023, rs17424122 and rs17409589) are marked by an
asterisk in Figure 1. All analyses were carried out using COCAPHASE
29
. Nonsynonymous SNP rs11887534 (D19H) is in boldface, and the risk allele is underlined. The sole risk
haplotype (C-
C-A-C-A-T-C) is fully defined by rs11887534 allele C, corresponding to histidine at the amino acid sequence level; all other haplotypes are protective and carry allele G.
This haplotype pattern strongly suggests that rs11887534 is the major risk variant at the ABCG8 locus. OR, odds ratio.
4 ADVANCE ONLINE PUBLICATION NATURE GENETICS
LETTERS

6. Khanuja, B. et al. Lith1, a major gene affecting cholesterol gallstone formation
among inbred strains of mice. Proc. Natl. Acad. Sci. USA 92, 7729–7733
(1995).
7. Paigen, B. et al. Quantitative trait loci mapping for cholesterol gallstones in AKR/J and
C57L/J strains of mice. Physiol. Genomics 4, 59–65 (2000).
8. Lammert, F. et al. Lith genes control mucin accumulation, cholesterol crystallization,
and gallstone formation in A/J and AKR/J inbred mice. Hepatology 36, 1145–1154
(2002).
9. Wittenburg, H. et al. FXR and ABCG5/ABCG8 as determinants of cholesterol gallstone
formation from quantitative trait locus mapping in mice. Gastroenterology 125,
868–881 (2003).
10. Puppala, S. et al. A genomewide search finds major susceptibility loci for gallbladder
disease on chromosome 1 in Mexican Americans. Am. J. Hum. Genet. 78, 377–392
(2006).
11. Barrett, J.C., Fry, B., Maller, J. & Daly, M.J. Haploview: analysis and visualization of LD
and haplotype maps. Bioinformatics 21, 263–265 (2005).
12. Hazard, S.E. & Patel, S.B. Sterolins ABCG5 and ABCG8: regulators of whole body
dietary sterols. Pflugers Arch. 453, 745–752 (2007).
13. Hubacek, J.A., Berge, K.E., Cohen, J.C. & Hobbs, H.H. Mutations in ATP-cassette
binding proteins G5 (ABCG5) and G8 (ABCG8) causing sitosterolemia. Hum. Mutat.
18, 359–360 (2001).
14. Iida, A. et al. Catalog of 605 single-nucleotide polymorphisms (SNPs) among 13 genes
encoding human ATP-binding cassette transporters: ABCA4, ABCA7, ABCA8, ABCD1,
ABCD3, ABCD4, ABCE1, ABCF1, ABCG1, ABCG2, ABCG4, ABCG5, and ABCG8.
J. Hum. Genet. 47, 285–310 (2002).
15. Schafmayer, C. et al. Predictors of gallstone composition in 1025 symptomatic
gallstones from Northern Germany. BMC Gastroenterol. 6, 36 (2006).
16. Berge, K.E. et al. Accumulation of dietary cholesterol in sitosterolemia caused by
mutations in adjacent ABC transporters. Science 290, 1771–1775 (2000).
17. Salen, G. et al. Sitosterolemia. J. Lipid Res. 33, 945–955 (1992).
18. Nakeeb, A. et al. Gallstones: genetics versus environment. Ann. Surg. 235, 842–849
(2002).
19. Sarin, S.K., Negi, V.S., Dewan, R., Sasan, S. & Saraya, A. High familial prevalence of
gallstones in the first-degree relatives of gallstone patients. Hepatology 22, 138–141
(1995).
20. Duggirala, R., Mitchell, B.D., Blangero, J. & Stern, M.P. Genetic determinants of
variation in gallbladder disease in the Mexican-American population. Genet. Epide-
miol. 16, 191–204 (1999).
21. Schafmayer, C. et al. Investigation of the Lith1 candidate genes ABCB11 and LXRA in
human gallstone disease. Hepatology 44, 650–657 (2006).
22. Schafmayer, C. et al. Investigation of the Lith6 candidate genes APOBEC1 and PPARG
in human gallstone disease. Liver Int. (in the press).
23. Wittenburg, H. et al. Association of a lithogenic Abcg5/Abcg8 allele on Chromosome
17 (Lith9) with cholesterol gallstone formation in PERA/EiJ mice. Mamm. Genome 16,
495–504 (2005).
24. Acalovschi, M. et al. Are plasma lipid levels related to ABCG5/ABCG8 polymorphisms?
A preliminary study in siblings with gallstones. Eur. J. Intern. Med. 17, 490–494
(2006).
25. Berge, K.E. et al. Heritability of plasma noncholesterol sterols and relationship to DNA
sequence polymorphism in ABCG5 and ABCG8. J. Lipid Res. 43, 486–494 (2002).
26. Gylling, H. et al. Polymorphisms in the ABCG5 and ABCG8 genes associate with
cholesterol absorption and insulin sensitivity. J. Lipid Res. 45, 1660–1665 (2004).
27. Wittenburg, H. & Carey, M.C. Biliary cholesterol secretion by the twinned sterol half-
transporters ABCG5 and ABCG8. J. Clin. Invest. 110, 605–609 (2002).
28. Franke, A. et al. GENOMIZER: an integrated analysis system for genome-wide
association data. Hum. Mutat. 27, 583–588 (2006).
29. Dudbridge, F. Pedigree disequilibrium tests for multilocus haplotypes. Genet. Epide-
miol. 25, 115–121 (2003).
30. Steffens, M. et al. SNP-based analysis of genetic substructure in the German
population. Hum. Hered. 62, 20–29 (2006).
© 2007 Nature Publishing Group http://www.nature.com/naturegenetics
NATURE GEN ETICS ADVAN CE ONLINE PUBLICATION 5
LETTERS