Sequence variants in the CLDN14 gene associate with
kidney stones and bone mineral density
Gudmar Thorleifsson1,14, Hilma Holm1,2,14, Vidar Edvardsson3, G Bragi Walters1, Unnur Styrkarsdottir1,
Daniel F Gudbjartsson1, Patrick Sulem1, Bjarni V Halldorsson1,4, Femmie de Vegt5, Frank C H d’Ancona6,
Martin den Heijer5,7, Leifur Franzson8, Claus Christiansen9, Peter Alexandersen10, Thorunn Rafnar1,
Kristleifur Kristjansson1, Gunnar Sigurdsson11, Lambertus A Kiemeney5,6, Magnus Bodvarsson12,
Olafur S Indridason12, Runolfur Palsson12, Augustine Kong1, Unnur Thorsteinsdottir1,13& Kari Stefansson1,13
Kidney stone disease is a common condition. To search
for sequence variants conferring risk of kidney stones, we
conducted a genome-wide association study in 3,773 cases
and 42,510 controls from Iceland and The Netherlands. We
discovered common, synonymous variants in the CLDN14 gene
that associate with kidney stones (OR ¼ 1.25 and P ¼ 4.0 ?
10?12for rs219780[C]). Approximately 62% of the general
population is homozygous for rs219780[C] and is estimated
to have 1.64 times greater risk of developing the disease
compared to noncarriers. The CLDN14 gene is expressed in
the kidney and regulates paracellular permeability at epithelial
tight junctions. The same variants were also found to associate
with reduced bone mineral density at the hip (P ¼ 0.00039)
and spine (P ¼ 0.0077).
Nephrolithiasis is a common disorder in the Western world. The life-
time risk of stone formation exceeds 12% in men and 5% inwomen in
the United States1and the recurrence rates are 50% within 5–10 years
of the initial event in untreated individuals2. Most kidney stones
(B80%) contain calcium, whereas other forms, including uric acid,
struvite and cystine stones, are far less common1,3. Metabolic risk
factors can be identified in 80–90% of adults with recurrent disease4,5.
Idiopathic hypercalciuria is the most common risk factor, present in
40–50% of adults with recurrent calcium stones4,6and 75–80% of
Figure 1 The 21q22 locus. (a) The pairwise correlation structure in a
500-kb interval (36.45–36.95 Mb, NCBI B36) on chromosome 21. The
upper plot shows pairwise D¢ for 454 common SNPs (with MAF 4 5%)
from the HapMap (v22) CEU dataset. The lower plot shows the
corresponding r2values. (b) Estimated sex-averaged recombination rates
(saRR) in cM/Mb from the HapMap Phase II data33. (c) Location of known
genes in the region. (d) Schematic view of the association with kidney
stones for all the markers tested in the GWA. All panels use the same
horizontal scale shown in d.
Received 4 March; accepted 2 June; published online 28 June 2009; doi:10.1038/ng.404
1DeCODE genetics, Sturlugata 8, 101 Reykjavik, Iceland.2Department of Internal Medicine, Division of Cardiology, Emory University School of Medicine, Atlanta,
Georgia, USA.3Department of Pediatrics, Landspitali University Hospital, University of Iceland, Reykjavik, Iceland.4Reykjavik University, Reykjavik, Iceland.
5Department of Epidemiology, Biostatistics and Health Technology Assessment,6Department of Urology and7Department of Endocrinology, Radboud University
Nijmegen Medical Centre, Nijmegen, The Netherlands.8Department of Genetics and Molecular Medicine, Landspitali University Hospital, University of Iceland,
Reykjavik, Iceland.9Nordic Bioscience A/S, Herlev, Denmark.10Center for Clinical and Basic Research A/S, Ballerup, Denmark.11Division of Endocrinology and
Metabolism and12Division of Nephrology, Landspitali University Hospital, University of Iceland, Reykjavik, Iceland.13University of Iceland, Faculty of Medicine,
Reykjavik, Iceland.14These authors contributed equally to this work. Correspondence should be addressed to G.T. (email@example.com) or
926VOLUME 41 [ NUMBER 8 [ AUGUST 2009 NATURE GENETICS
© 2009 Nature America, Inc. All rights reserved.
children with kidney stones7. Other rarer abnormalities linked to
calcium kidney stone formation include hyperoxaluria, hyperurico-
suria and hypocitraturia1.
Both kidney stone disease and primary hypercalciuria have a well-
recognized heritable contribution. Between 35% and 65% of hyper-
calciuric stone formers and up to 70% of subjects with hypercalciuria
have relatives with nephrolithiasis8,9, and twin studies have estimated
the heritability of kidney stones to be 56%10. Calcium stones are also
associated with several monogenic mendelian traits, many of which
affect ion channels and membrane transporters8.
To search for sequence variants conferring risk of kidney stone
disease, we conducted a genome-wide association study using the
Illumina HumanHap300 and HumanHapCNV370 bead chips.
After quality filtering, 303,120 SNPs were tested for association with
radiopaque kidney stones in a sample of 1,507 Icelandic cases and
34,033 population controls. The results were adjusted for relatedness
between individuals and potential population stratification using the
method of genomic control by dividing the w2statistic by 1.078.
Two highly correlated SNPs (r2¼ 1 based on the HapMap CEPH
Utah data), on chromosome 21q22.13, achieved genome-wide sig-
nificant association (P o 1.6 ? 10?7, Supplementary Fig. 1 and
Supplementary Table 1): rs219781[C], with OR ¼ 1.30 and P ¼ 3.2 ?
10?8, and rs219778[T], with OR ¼ 1.30 and P ¼ 3.2 ? 10?8. Both
at-risk alleles are very common, with a population frequency of 75%
in the Icelandic control set. After adjustment for the association with
rs219781, neither rs219778 nor any other SNP on the HumanHap300/
HumanHapCNV370 in the 21q22.13 region showed significant asso-
ciation with kidney stones (Supplementary Table 1). The two SNPs
are located in the same linkage disequilibrium (LD) block, 2,020
bases apart on either side of the last exon of the CLDN14 (claudin 14)
gene (Fig. 1). This gene encodes a protein with four transmembrane
domains and intracellular N and C termini and is involved in
regulation of paracellular permeability at epithelial tight junctions11.
We typed the two variants in 1,520 additional Icelandic kidney
stone cases and 4,726 controls, and in 746 kidney stone cases and
3,751 controls from The Netherlands. rs219778[T] was significantly
associated with kidney stones in both the Icelandic (OR ¼ 1.25 and
P ¼ 1.2 ? 10?5) and Dutch (OR ¼ 1.18 and P ¼ 0.015) replication
sets. The combined effect of rs219778[T] in the discovery and two
replication sets was OR ¼ 1.23 (95% CI ¼ 1.16–1.31) with a
corresponding P value of 1.7 ? 10?12. Similar results were obtained
for rs219781[C] (OR ¼ 1.23 (95% CI ¼ 1.16–1.30) and P ¼ 4.0 ?
10?12) (Table 1). As the variants associating with kidney stones are
common in populations of European origin, their population attri-
butable risk is substantial, at 27–30%.
Both rs219778 and rs219781 are non-exonic SNPs. To search for
causal variant(s), we sequenced the exons and their immediate
flanking regions of the four isoforms of the CLDN14 gene (total of
Table 1 Association between variants in CLDN14 and kidney
Cohort (NC/NA) Frequency
Variant ATControlsCases OR (95% CI)P
Iceland I (34,033/1,507)
Iceland II (4,726/1,520)
Iceland combined (38,759/3,027)
The Netherlands (3,751/746)
3.2 ? 10?8
3.2 ? 10?8
3.4 ? 10?8
3.3 ? 10?8
1.2 ? 10?5
1.5 ? 10?5
5.5 ? 10?5
6.6 ? 10?6
2.7 ? 10?11
4.5 ? 10?11
2.4 ? 10?11
1.7 ? 10?11
1.7 ? 10?12
4.0 ? 10?12
4.0 ? 10?12
1.7 ? 10?12
Association between kidney stones and the four variants, rs219778[T], rs219781[C],
rs219780[C] and rs219779[C], all located in or close to the CLDN14 gene on
chromosome 21q22. Results are shown for the Icelandic discovery dataset, an Icelandic
replication dataset, the two Icelandic datasets combined, a replication dataset from The
Netherlands and for all the datasets combined using a Mantel-Haenszel model. Shown
are the number of controls (NC) and cases (NA) for each study group, the frequency in controls
and in cases, the OR with 95% confidence intervals (CI) and P values according to the
multiplicative model. For the Icelandic study groups, the P values and CI were adjusted for
relatedness (see Online Methods).
Table 2 Correlation between kidney stone variants in CLDN14 and
bone mineral density
Dataset (N) Effect (s.e.)PEffect (s.e.)P
?0.074 (0.019) 0.000083 ?0.070 (0.021) 0.00071
0.002 (0.046) 0.97
?0.091 (0.021) 0.000013 ?0.091 (0.022) 0.000062
0.029 (0.049) 0.57
?0.069 (0.020) 0.00043
?0.062 (0.051) 0.23
?0.071 (0.021) 0.00088
?0.066 (0.021) 0.002
?0.075 (0.055) 0.17
?0.064 (0.023) 0.0055
?0.042 (0.027) 0.12
?0.041 (0.029) 0.17
?0.021 (0.027) 0.45
?0.009 (0.029) 0.78
Females (10,363) ?0.073 (0.016) 0.000011 ?0.073 (0.018) 0.000058
?0.052 (0.016) 0.001
Females (10,608) ?0.052 (0.017) 0.002
Correlation between the two CLDN14 exonic variants that associate with kidney stones,
rs219780[C] and rs219779[C], and BMD. The correlation was tested in a set of 8,450
Icelandic individuals of both sexes and 3,601 Danish females. The table shows the
correlation with sex-, age- and weight-adjusted values of BMD for hip and spine
separately, and for the Icelandic sample set for males, females and both sexes
combined. Shown are the estimated effects expressed as standardized values per copy
of the SNP allele and the corresponding standard errors (s.e.) and P values. For the
Icelandic sample set, both the standard errors and P values have been adjusted for
relatedness of the study individuals using simulations. Also shown is the result for
the two sample sets combined.
?0.063 (0.015) 0.000039 ?0.060 (0.017) 0.00039
?0.057 (0.022) 0.0077
?0.055 (0.023) 0.019
NATURE GENETICS VOLUME 41 [ NUMBER 8 [ AUGUST 2009927
© 2009 Nature America, Inc. All rights reserved.
seven exons) in a random subset of kidney stone cases (N ¼ 371)
and controls (N ¼ 270). The previously discovered rare mutations
(398delT and 254T4A) that associate with deafness12were not
identified. However, we identified several noncoding and intronic
variants, as well as some synonymous and nonsynonymous variants
(Supplementary Table 2). Of these, two common synonymous SNPs,
rs219779 (R81R) and rs219780 (T229T), both located in the last and
only translated exon of CLDN14, showed significant association with
kidney stones in the material sequenced. In all sample sets combined,
these two variants showed very significant association with kidney
stones: OR ¼ 1.23 (95% CI ¼ 1.16–1.31) and P ¼ 1.7 ? 10?12for
rs219779, and OR ¼ 1.25 (95% CI ¼ 1.17–1.33) and P ¼ 4.0 ? 10?12
for rs219780 (Table 1). As both rs219779 and rs219780 are highly
correlated with the variants rs219778 and rs219781 (r2¼ 1 between
rs219778 and rs219779 and r2¼ 0.77 between rs219778 and rs219780)
and have similar risk allele frequencies (0.74–0.79), any further
dissection of the signal was not possible (Supplementary Table 3).
In the remaining text we present only results for rs219780, although
we report results for both exonic SNPs in all relevant tables.
The estimated risk was similar when we restricted the analysis to
males or to females. For 1,835 Icelandic male cases and 16,372
Icelandic male controls, we obtained for rs219780[C] OR ¼ 1.26
(P ¼ 5.3 ? 10?7), whereas for 1,192 female cases and 22,385 female
controls the corresponding result was OR ¼ 1.28 (P ¼ 9.5 ? 10?6).
The observed effect for rs219780[C] was somewhat stronger for the
1,492 cases with confirmed calcium stones (OR ¼ 1.37 and P ¼ 4.3 ?
10?10) compared to the result for all Icelandic cases. Relative to
noncarriers, the genotype-specific OR for kidney stones was 1.33
and 1.64 for heterozygous and homozygous carriers of rs219780[C],
respectively (Supplementary Table 4).
The synonymous SNPs rs219779 and rs219780 are located within the
sequences encoding the first putative extracellular and the C-terminal
cytoplasmic domains of the claudin 14 protein, respectively. Whether
these two synonymous variants have any
functional relevance has not, to our knowl-
edge, been determined. This association with
kidney stones could be mediated through
their LD to noncoding undefined regulatory
elements or through their own effect on
mRNA stability or processing, as previously
documented for both synonymous and non-
synonymous coding mutations in genes such
as DRD1, ABCB1 and OPRM1 (refs. 13–15).
We did not, however, observe significant
correlation of the risk variants with mRNA
expression of CLDN14 in 673 RNA samples
from adipose tissue or in 1,002 samples from
peripheral blood16(data not shown).
The correlation between kidney stones and
bone mineral density (BMD) has been exten-
sively explored. The overall data suggests a
risk for bone loss in individuals with kidney
stones, primarily those with hypercalciuria,
possibly associated with enhanced bone turn-
over17. We therefore assessed the association
between the two variants, rs219779[C] and
rs219780[C], and BMD at the hip (total hip)
and spine (L2-L4) in a sample of 8,450 Ice-
landic individuals and 3,601 Danish women.
In the combined analysis of the two samples,
each copy of the risk variant rs219780[C]
corresponded to a decrease in hip BMD of an estimated 0.060 s.d.
(P ¼ 0.00039) and a decrease in spine BMD of 0.057 s.d. (P ¼ 0.0077)
(Table 2). Although not significant, the effect on spine BMD in men
was similar to that observed in women, whereas the effect on hip BMD
in men, although small, was in the opposite direction compared to that
in women. The effect of rs219780[C] on BMD at the hip was greater
among kidney stone formers, whereas the reverse was true for BMD at
the spine. These differences, however, are not statistically significant
(Supplementary Table 5). The association with BMD remained signif-
icant after excluding kidney stone cases, suggesting that the effect of the
variants on BMD is not through enrichment of kidney stone cases
among individuals with low BMD. Likewise, the associationwith kidney
stones did not change in the Icelandic sample when we excluded
individuals that fulfill the WHO criteria for osteoporosis18; exclusion
of 162 cases and 2,263 controls gave OR ¼ 1.26 (P ¼ 4.6 ? 10?10)
As the vast majority of radiopaque kidney stones are calcium stones,
we decided to test for correlation between the risk variants and several
biochemical values related to calcium metabolism (Table 3 and Sup-
plementary Table 6). We tested rs219779[C] and rs219780[C] in a set
of 1,026 Icelanders who had participated in a population-based study
of bone and mineral health19. We observed suggestive association with
concentrations of serum total CO2, urinary calcium (uCa) and serum
parathyroid hormone (PTH). Each copy of rs219780[C] corresponded
to a decrease in serum total CO2of 0.25 mEq/l (P ¼ 0.0087), an
increase in uCa of 0.27 mmol/l (P ¼ 0.005) and an increase in PTH of
2.48 pg/ml (P ¼ 0.0072). No association was observed between the
risk variants and serum measurements of creatinine, calcium, phos-
phate, vitamin D or pH. We also found correlation between rs219780[C]
and increased serum alkaline phosphatase, a marker of bone turnover,
and increased serum C-telopeptide of type I collagen (CTx), a marker of
bone resorption. Although we recognize that the correlations presented
here between the risk variants and the various biochemical parameters
Table 3 Correlation between kidney stone variants in CLDN14 and measurements of
Trait Effect (s.e.)P Effect (s.e.)P
Serum albumin (mg/l)
Serum alkaline phosphotase (Total) (U/l)
Serum calcium (mmol/l)
Serum ionized calcium (mmol/l)
Serum total CO2(mEq/l)
Serum creatinine (mmol/l)
Serum CTx (U/l)
Serum cystacin C (mg/l)
Serum magnesium (mmol/l)
Serum parathyroid hormone - ELECS (pg/ml)
Serum pH value
Serum phosphate (mmol/l)
Serum 1.25-(OH)2-D vitamin D (mmol/l)
Serum 25-hydroxy vitamin D (nmol/l)
Urine calcium (mmol/l)
Urine creatinine (mmol/l)
Correlation between the two CLDN14 exonic variants that associate with kidney stones, rs219780[C] and rs219779[C],
and various biochemical measurements related to calcium metabolism. Correlation was tested in a population-based set
of 1,026 Icelanders. The effect and the corresponding standard error (s.e.) were obtained by regressing the unadjusted
trait values on the number of risk alleles and individual carriers. The P value was calculated by regressing the sex
and age adjusted and inverse normal transformed trait values on the number of risk alleles carried. All P values and
standard errors were adjusted for relatedness by simulations as described in Online Methods.
928 VOLUME 41 [ NUMBER 8 [ AUGUST 2009 NATURE GENETICS
© 2009 Nature America, Inc. All rights reserved.
are individually not significant if we adjust for the number of traits
tested, they nevertheless provide insight into the possible mechanisms
by which the CLDN14 variants could predispose to kidney stones and
We and others have previously reported a number of sequence
variants that affect BMD20–25. To further examine the relationship
between kidney stones and BMD observed in this study, we assessed
the correlation of previously reported BMD-associated variants with
kidney stones. Of the 18 BMD-related variants we tested, only one
showed nominally significant association (Supplementary Table 7).
Furthermore, we did not observe consistent association between the
18 variants and uCa, serum total CO2and PTH concentrations (Sup-
plementary Table 8). These findings support the view that the observed
effect on both kidney stones and BMD in our study is specifically
mediated through the variants in CLDN14 and is not a reflection of a
more general relationship between these two phenotypes.
Claudin 14 is a member of the claudin family of membrane
proteins that regulate paracellular passage of ions and small solutes
at epithelial tight junctions26. The specific distribution and function
of the various claudins is an important determinant of the para-
cellular transport properties of different epithelia26. A renal disorder,
autosomal recessive familial hypomagnesemia with hypercalciuria and
nephrocalcinosis, commonly associated with kidney stones, is caused
by mutations in the CLDN16 and CLDN19 genes, both expressed in
the loop of Henle11,27,28. CLDN14 is also expressed in the kidney, both
in the loop of Henle and the proximal convoluted tubule, as well
as in the epithelia of several other organs, and has been observed to
selectively decrease cation paracellular permeability29,30. Mutations
in CLDN14 have been found in families with nonsyndromic auto-
somal recessive deafness, but hitherto neither kidney stones nor
kidney dysfunction have been described in humans or mice with
In CLDN11/CLDN14 double mutant mice a mild enhancement in
urine excretion of magnesium and calcium was observed29. In line
with this finding, we observed a suggestive correlation between the
CLDN14 variants and higher uCa. Furthermore, in our work, the same
alleles showed correlation with lower serum total CO2and reduced
BMD. We postulate that the primary abnormalities associating with
the variants in CLDN14 are decreased serum total CO2and increased
uCa, and that kidney stones and decreased BMD may be caused by
these metabolic abnormalities.
Previous studies have suggested a role for claudins in acid-base
homeostasis. For example, claudin 8, a negative regulator of paracellular
cation transport, similar to claudin 14, also seems to act specifically
as a negative regulator of protons, bicarbonate and ammonium ions
in the distal nephron31. The association between CLDN14 and
serum total CO2level in our study suggests a role for claudin 14 in
acid-base balance, possibly directly through regulation of paracellular
ion transport. CLDN14 has been found to be expressed in the
proximal tubule where the reclamation of the bulk of filtered bicar-
bonate occurs as well as synthesis of ammonium, a major component
of renal acid excretion. Metabolic acidosis is known to both
increase the release of calcium from bone32and reduce calcium
reabsorption in the kidney, causing hypercalciuria3, and thus con-
tributing to both kidney stone formation and decreased BMD.
Additionally, metabolic acidosis causes hypocitraturia, another impor-
tant risk factor for calcium stones3. As discussed, hypercalciuria is
the most common metabolic risk factor for kidney stone disease,
and reduced BMD has been linked to both kidney stones and
hypercalciuria. Whether the observed association between the
CLDN14 variants and uCa is direct or indirect is unclear. It is,
however, well recognized that acid-base balance and electrolyte trans-
port, including that of calcium, are intricately coupled in the kidney.
In conclusion, variants in CLDN14 associate with kidney stones and
reduced BMD. Further research is required to clarify the biological
pathways interconnecting the two phenotypes and CLDN14. However,
as the claudins have been shown to influence paracellular cation
transport, it is tempting to postulate that the primary defects relating
to the associated variants are increased urinary excretion of calcium
and decreased serum total CO2.
Methods and any associated references are available in the online
version of the paper at http://www.nature.com/naturegenetics/.
Accession codes. GenBank: exons for sequencing were based on CLDN14
accessions NM_144492, NM_012130, AJ566765 and AJ566766. Genpept:
the position of the amino acids within the claudin 14 protein was based
on accession NP_036262.
Note: Supplementary information is available on the Nature Genetics website.
The study was designed and results interpreted by G.T., H.H., G.B.W., V.E., O.S.I.,
R.P., M.B., K.K., U.S., T.R., U.T. and K.S. Statistical analysis was carried out by
G.T., D.F.G., P.S., B.V.H. and A.K. V.E., O.S.I., R.P., G.S. and L.F. collected the
Icelandic data. F.d.V., F.C.H.d’A., M.d.H. and L.A.K. collected the Dutch data.
C.C. and P.A. collected the Danish data. Authors H.H., G.T., G.B.W., U.T. and
K.S. wrote the first draft of the paper. All authors contributed to the final version.
COMPETING INTERESTS STATEMENT
The authors declare competing financial interests: details accompany the full-text
HTML version of the paper at http://www.nature.com/naturegenetics/.
Published online at http://www.nature.com/naturegenetics/.
Reprints and permissions information is available online at http://npg.nature.com/
1. Coe, F.L., Evan, A. & Worcester, E. Kidney stone disease. J. Clin. Invest. 115,
2. Trinchieri, A. et al. A prospective study of recurrence rate and risk factors for recurrence
after a first renal stone. J. Urol. 162, 27–30 (1999).
3. Moe, O.W. Kidney stones: pathophysiology and medical management. Lancet 367,
4. Hess, B., Hasler-Strub, U., Ackermann, D. & Jaeger, P. Metabolic evaluation of patients
with recurrent idiopathic calcium nephrolithiasis. Nephrol. Dial. Transplant. 12,
5. Tefekli, A. et al. Metabolic risk factors in pediatric and adult calcium oxalate urinary
stone formers: is there any difference? Urol. Int. 70, 273–277 (2003).
6. Levy, F.L., Adams-Huet, B. & Pak, C.Y. Ambulatory evaluation of nephrolithiasis: an
update of a 1980 protocol. Am. J. Med. 98, 50–59 (1995).
7. Edvardsson, V., Elidottir, H., Indridason, O.S. & Palsson, R. High incidence of kidney
stones in Icelandic children. Pediatr. Nephrol. 20, 940–944 (2005).
8. Stechman, M.J., Loh, N.Y. & Thakker, R.V. Genetics of hypercalciuric nephrolithiasis:
renal stone disease. Ann. NY Acad. Sci. 1116, 461–484 (2007).
9. Polito, C., La Manna, A., Nappi, B., Villani, J. & Di Toro, R. Idiopathic hypercalciuria
and hyperuricosuria: family prevalence of nephrolithiasis. Pediatr. Nephrol. 14,
10. Goldfarb, D.S., Fischer, M.E., Keich, Y. & Goldberg, J. A twin study of genetic and
dietary influences on nephrolithiasis: a report from the Vietnam Era Twin (VET)
Registry. Kidney Int. 67, 1053–1061 (2005).
11. Angelow, S., Ahlstrom, R. & Yu, A.S. Biology of claudins. Am. J. Physiol. Renal Physiol.
295, F867–F876 (2008).
12. Wilcox, E.R. et al. Mutations in the gene encoding tight junction claudin-14 cause
autosomal recessive deafness DFNB29. Cell 104, 165–172 (2001).
13. Duan, J. et al. Synonymous mutations in the human dopamine receptor D2 (DRD2) affect
mRNA stability and synthesis of the receptor. Hum. Mol. Genet. 12, 205–216 (2003).
14. Wang, D., Johnson, A.D., Papp, A.C., Kroetz, D.L. & Sadee, W. Multidrug resistance
polypeptide 1 (MDR1, ABCB1) variant 3435C4T affects mRNA stability. Pharmaco-
genet. Genomics 15, 693–704 (2005).
15. Zhang, H. et al. Association between two mu-opioid receptor gene (OPRM1) haplotype
blocks and drug or alcohol dependence. Hum. Mol. Genet. 15, 807–819 (2006).
16. Emilsson, V. et al. Genetics of gene expression and its effect on disease. Nature 452,
17. Asplin, J.R. et al. Bone mineral density and urine calcium excretion among subjects
with and without nephrolithiasis. Kidney Int. 63, 662–669 (2003).
NATURE GENETICS VOLUME 41 [ NUMBER 8 [ AUGUST 2009929
© 2009 Nature America, Inc. All rights reserved.
18. Kanis, J.A., Melton, L.J. III, Christiansen, C., Johnston, C.C. & Khaltaev, N. The
diagnosis of osteoporosis. J. Bone Miner. Res. 9, 1137–1141 (1994).
19. Gudmundsdottir, S.L., Indridason, O.S., Franzson, L. & Sigurdsson, G. Age-related
decline in bone mass measured by dual-energy X-ray absorptiometry and quantitative
ultrasound in a population-based sample of both sexes: identification of useful
ultrasound thresholds for osteoporosis screening. J. Clin. Densitom. 8, 80–86 (2005).
20. Styrkarsdottir, U. et al. Multiple genetic loci for bone mineral density and fractures.
N. Engl. J. Med. 358, 2355–2365 (2008).
21. Richards, J.B. et al. Bone mineral density, osteoporosis, and osteoporotic fractures: a
genome-wide association study. Lancet 371, 1505–1512 (2008).
22. Styrkarsdottir, U. et al. New sequence variants associated with bone mineral density.
Nat. Genet. 41, 15–17 (2009).
23. Timpson,N.J.et al.Common variantsinthe region aroundOsterix are associatedwithbone
mineral density and growth in childhood. Hum. Mol. Genet. 18, 1510–1517 (2009).
24. Uitterlinden, A.G. et al. Polymorphisms in the sclerosteosis/van Buchem disease gene
(SOST) region are associated with bone-mineral density in elderly whites. Am. J. Hum.
Genet. 75, 1032–1045 (2004).
25. Sims, A.M. et al. Genetic analyses in a sample of individuals with high or low BMD shows
association with multiple Wnt pathway genes. J. Bone Miner. Res. 23, 499–506 (2008).
26. Krause, G. et al. Structure and function of claudins. Biochim. Biophys. Acta 1778,
27. Simon, D.B. et al. Paracellin-1, a renal tight junction protein required for paracellular
Mg2+ resorption. Science 285, 103–106 (1999).
28. Konrad, M. et al. Mutations in the tight-junction gene claudin 19 (CLDN19) are
associated with renal magnesium wasting, renal failure, and severe ocular involvement.
Am. J. Hum. Genet. 79, 949–957 (2006).
29. Elkouby-Naor, L., Abassi, Z., Lagziel, A., Gow, A. & Ben-Yosef, T. Double gene deletion
reveals lack of cooperation between claudin 11 and claudin 14 tight junction proteins.
Cell Tissue Res. 333, 427–438 (2008).
30. Ben-Yosef, T. et al. Claudin 14 knockout mice, a model for autosomal recessive
deafness DFNB29, are deaf due to cochlear hair cell degeneration. Hum. Mol. Genet.
12, 2049–2061 (2003).
31. Angelow, S., Kim, K.J. & Yu, A.S. Claudin-8 modulates paracellular permeability to
acidic and basic ions in MDCK II cells. J. Physiol. (Lond.) 571, 15–26 (2006).
32. Lemann, J. Jr, Bushinsky, D.A. & Hamm, L.L. Bone buffering of acid and base in
humans. Am. J. Physiol. Renal Physiol. 285, F811–F832 (2003).
33. Frazer, K.A. et al. A second generation human haplotype map of over 3.1 million SNPs.
Nature 449, 851–861 (2007).
930 VOLUME 41 [ NUMBER 8 [ AUGUST 2009 NATURE GENETICS
© 2009 Nature America, Inc. All rights reserved.
Study subjects from Iceland. Kidney stone cases and controls. The Icelandic
kidney stone cases consisted of individuals with confirmed radiopaque kidney
stones, diagnosed in the years 1983–2003, from the Icelandic kidney stone
registry at Landspitali University Hospital, Reykjavik, Iceland. The study
included a total of 3,027 individuals with kidney stones who had donated blood
through various genetic programs at deCODE genetics. Information on stone
composition was available for a subset of 1,528 of the cases, of which 1,492
had calcium stones. The study was approved by the Icelandic Data Protection
Authority and the National Bioethics Committee of Iceland and all study
subjects gave written informed consent. The 34,033 Icelandic controls used in
the genome-wide association study and the 4,726 controls used in the replication
study were selected among individuals who had participated in various genetic
programs at deCODE genetics. Individuals who reported a history of kidney
stones were excluded from the control set. The breakdown of the controls
into the various genetic programs was approximately as follows: tremor (360),
preeclampsia (780), endometriosis (330), psoriasis (850), type 2 diabetes (1,450),
Alzheimer’s disease (700), osteoarthritis (3,800), schizophrenia (550), peripheral
artery disease (1,450), abdominal aortic aneurysm (380), chronic obstructive
pulmonary disease (860), stroke (1,830), osteoporosis (3,230), coronary artery
disease (4,040), hypertension (2,500), asthma (1,470), Parkinson’s disease (320),
sleep apnea (650), age-related macular degeneration (620), polycystic ovary
syndrome (1,380), rheumatoid arthritis (760), lung cancer (280), longevity
(1,580), benign prostatic hyperplasia (860), enuresis (890), migraine (1,280),
glaucoma (620), attention deficit hyperactivity disorder (560), prostate cancer
(1,280), infectious diseases (3,420), anxiety (1,180), expression studies (870),
autism (160), dyslexia (870), melanoma (550), colorectal cancer (890), deep
vein thrombosis (950), restless leg syndrome (560), studies on addiction (3,300)
and population controls (1,180). Some of the controls have participated in
more than one genetic program, in which case their genotypes are included
Study on bone and mineral health cohort. These individuals participated in
a cross-sectional study on bone and mineral health in a random sample of
community-dwelling adults aged 30 to 85 years, a total of 1,026 individuals, in
the Reykjavik area in Iceland19. All subjects had hip (total hip; combined values
at the femoral neck, trochanter and intertrochanter region) and lumbar spine
bone density measured by dual energy X-ray absorptiometry (DEXA, Hologic
QDR4500A), gave blood and urine samples, and answered a thorough ques-
tionnaire on medications and medical history. Subjects with diseases or taking
medication known to affect bone tissue were exluded from the analysis. The
National Bioethics Committee of Iceland approved the study, and all partici-
pants provided a written consent form.
Bone density samples. The Icelandic bone density sample set comprises 8,446
individuals who had bone density measurements (DEXA, Hologic QDR4500A)
at the lumbar spine (L2-L4) and/or the hip (total hip) and participated in the
osteoporosis genetics program or other genetic programs at deCODE genetics.
All participants gave written informed consent and the study was approved by
the Icelandic Data Protection Authority and the National Bioethics Committee
Study subjects from The Netherlands. The Dutch individuals with kidney stones
were recruited from two sources: the outpatient clinics of the Radboud University
Nijmegen Medical Center (RUNMC) and the Nijmegen Biomedical Study.
All individuals who present to the outpatient clinics of the RUNMC are
invited to participate in a study on the effects of genes and lifestyle on the
development of urological diseases. If they consent, they fill out a lifestyle
questionnaire and donate a blood sample for DNA isolation.
The details of the Nijmegen Biomedical Study were reported previously34.
Briefly, this is a population-based survey conducted by the Department of
Epidemiology and Biostatistics and the Department of Clinical Chemistry of
the Radboud University Nijmegen Medical Center (RUNMC), in which 9,371
individuals participated from a total of 22,500 age and sex stratified, randomly
selected inhabitants of Nijmegen. The subjects were invited to participate in a
study on gene-environment interactions in multifactorial diseases. All partici-
pants filled out a questionnaire on lifestyle and medical history at baseline and
6,500 of them donated blood samples for DNA isolation and biochemical
studies. The controls for the study were also taken from the biobank of the
Nijmegen Biomedical Study. All cases and controls were of self-reported
European descent and were fully informed about the goals and the procedures
of the study.
The study protocols for the recruitment of subjects from outpatient clinics
and the recruitment of participants to the Nijmegen Biomedical Study were
approved by the Institutional Review Board of the RUNMC. All study subjects
gave written informed consent.
Study subjects from Denmark. The Danish samples were derived from the
Prospective Epidemiological Risk Factor (PERF Study)35. These are postmeno-
pausal women (n ¼ 3,884), in the age group 55–86 years, who participated in
a prospective epidemiological study and in various clinical trials for osteo-
porosis at the Center for Clinical and Basic Research, Copenhagen. Baseline
DEXA bone density measurement (Hologic QDR2000) at the hip (total hip)
and lumbar spine (L2-L4) was used. The study was approved by the Ethics
Committee of Copenhagen County and was in accordance with the principles
of the Helsinki Declaration.
Illumina genome-wide genotyping. All Icelandic case and control samples
were assayed with the Illumina HumanHap300 or HumanHapCNV370 bead
chips, containing 317,503 and 370,404 haplotype-tagging SNPs derived from
Phase I of the International HapMap project. Only SNPs present on both chips
were included in the analysis and SNPs were excluded if they (i) had yield lower
than 95% in cases or controls, (ii) had minor allele frequency less than 1% in
the population or (iii) showed significant deviation from Hardy-Weinberg
equilibrium in the controls (P o 0.001). Any samples with a call rate below
98% were excluded from the analysis. The final analysis included 303,120 SNPs.
Single-SNP genotyping. Single-SNP genotyping for all samples was carried out
at deCODE genetics in Reykjavik, Iceland, applying the same platform to all
populations studied. All single-SNP genotyping was carried out using the
Centaurus (Nanogen) platform36. The quality of each Centaurus SNPassay was
evaluated by genotyping each assay on the CEU samples and comparing the
results with the HapMap data. All assays had mismatch rate o0.5%. Addi-
tionally, all markers were re-genotyped on more than 10% of samples typed
with the Illumina platform, resulting in an observed mismatch in less than
0.5% of samples.
Association analysis. For association analysis, we used a standard likelihood
ratio statistic, implemented in the NEMO software37to calculate two-sided
P values and odds ratios (ORs) for each individual allele, assuming a multi-
plicative model for risk, that is, that the risk of the two alleles a person carries
multiplies38. Allelic frequencies, rather than carrier frequencies, are presented
for the markers and P values are given after adjustment for the relatedness of
the subjects. When estimating genotype-specific OR, we estimated genotype
frequencies in the population assuming Hardy-Weinberg equilibrium.
In general, allele and haplotype frequencies are estimated by maximum
likelihood and tests of differences between cases and controls are performed
using a generalized likelihood ratio test38. This method is particularly useful in
situations where there are some missing genotypes for the marker of interest
and genotypes of another marker, which is in strong LD with the marker of
interest, are used to provide some partial information. This was used in the
association tests presented in Tables 1–3 and Supplementary Tables 1, 3–6 to
ensure that the comparison of the highly correlated markers was done using the
same number of individuals. To handle uncertainties with phase and missing
genotypes, we computed maximum likelihood estimates, likelihood ratios and
P values directly for the observed data, and hence the loss of information due to
uncertainty in phase and missing genotypes is automatically captured by the
Results from multiple case-control groups were combined using a Mantel-
Haenszel model39in which the groups were allowed to have different popula-
tion frequencies for alleles, haplotypes and genotypes but were assumed to have
common relative risks.
The correlation between the sequence variants and the various biochemical
traits was done by regressing the individual trait values on the number of copies
of the at-risk variant an individual carries. To obtain effect estimates, we carried
out the regression using unadjusted trait values as response variables, whereas
to estimate the significance of the correlation, we carried out the regression
© 2009 Nature America, Inc. All rights reserved.
using sex and age adjusted and standardized (by an inverse normal transform) Download full-text
trait values as response values to eliminate the effect of non-normality of the
distribution of trait values. For the regression between bone mineral density
and the sequence variants, both the effect estimate and the P values were
calculated by regressing standardized age-, sex- and weight-adjusted trait values
on the number of copies of the risk variant an individual carries.
Correction for relatedness of the subjects and genomic control. Some of the
individuals in both the Icelandic case and control groups are related to each
other, causing the w2test statistic to have a mean 41 and median 40.675. We
estimated the inflation factor for the genome-wide association by calculating
the average of the 303,120 w2statistics, which was a method of genomic
control40to adjust for both relatedness and potential population stratification.
The inflation factor was estimated as 1.078 and the results presented from the
genome-wide association and in Table 1 and Supplementary Table 1 are based
on adjusting the w2statistics by dividing each of them by 1.078. To adjust the
association results for the Icelandic replication sample set, and the combined
replication and discovery sample set, where association results for a genome-
wide set of SNPs is not available, we used a previously described procedure
where we simulated genotypes through the genealogy of 708,683 Icelanders
to estimate the adjustment factor41. The adjustment factors for the replica-
tion and combined set of kidney stone cases and controls were 1.043 and
The adjustment factors for relatedness for the various biochemical traits
were estimated by simulations as described above and ranged from 1 (no
adjustment) to 1.030 depending on the trait in question. Likewise, the
adjustment factors for correlation with BMD were 1.165 (1.188) for hip (spine)
for females, 1 (1.023) for hip (spine) for males and 1.168 (1.198) for hip (spine)
for the sexes combined.
Sequencing of CLDN14. We resequenced the exons of the two CLDN14 RefSeqs
as well as thepredicted exons of two GenBank registered human mRNAs and the
sequence flanking the exons in 371 Icelandic cases and 270 Icelandic controls. We
designed sequencing assays (ranging in amplimer size from 305 bp to 550 bp)
for each of the exons using NCBI assembly build 36 and the Winseq program
(developed at deCODE genetics based on the Primer3 software)42. The 5 ml
PCR amplification reactions were set up on the Zymark SciClone ALH 300
robotic workstation in a 384-well PCR plate and amplified on a MJR Tetrad
(M.J. Research). We used Ampure (Agencourt) 384 PCR filters to remove
unincorporated PCR primers and mononucleotides from the PCR reaction. The
5 ml cycle sequencing dye terminator reaction was set up on a Zymark SciClone
ALH 300 robotic workstation in a 384-well PCR plate and amplified on a
MJR Tetrad (M.J. Research). Dye terminator removal was set up on a Zymark
SciClone ALH 300 robotic workstation using CleanSEQ (Agencourt) 384 PCR
filters. Electrophoresis was performed on Applied Biosystems 3700 DNA
Analyzers (Perkin-Elmer). deCODE genetics SequenceMiner sequence assembly
software was used to manually analyze and edit the generated sequence.
Only the last exon of CLDN14 is translated. We found several, mostly rare,
noncoding and intronic sequence variants as well as some synonymous and
nonsynonymous variants (Supplementary Table 2). Two rare nonsynonymous
variants caused the amino acid changes T4M43and A163V. Of the protein
coding exonic variants only the synonymous rs219780 (T229T) and rs219779
(R81R) variants were sufficiently frequent to use in our analysis.
Biochemistry. Fasting morning blood and second morning void urine samples
were collected. We measured serum ionized calcium and serum total CO2with
an ion selective electrode (ABL 700, Radiometer Denmark); serum alkaline
phosphatase (ALP) and urinary calcium and creatinine with a dry chemistry
autoanalyzer (Vitros); intact serum parathyroid hormone and serum CTx with
ECLIA (ElectroChemiLuminscence Immuno Assay, Elecsys 2010, Roche Diag-
nostics); and serum 25(OH)D (RIA; DiaSorin) and serum cystatin C with PEIA
(Particle-Enhanced Immunoturbidometric Assay; DAKO).
34. Wetzels, J.F., Kiemeney, L.A., Swinkels, D.W., Willems, H.L. & den Heijer, M. Age- and
gender-specific reference values of estimated GFR in Caucasians: the Nijmegen
Biomedical Study. Kidney Int. 72, 632–637 (2007).
35. Bagger, Y.Z. et al. Links between cardiovascular disease and osteoporosis in
postmenopausal women: serum lipids or atherosclerosis per se? Osteoporos. Int. 18,
36. Kutyavin, I.V. et al. A novel endonuclease IV post-PCR genotyping system. Nucleic
Acids Res. 34, e128 (2006).
37. Gretarsdottir, S. et al. The gene encoding phosphodiesterase 4D confers risk of
ischemic stroke. Nat. Genet. 35, 131–138 (2003).
38. Rice, J.A. Mathematical Statistics and Data Analysis (Wadsworth, Belmont, California,
39. Mantel, N. & Haenszel, W. Statistical aspects of the analysis of data from retrospective
studies of disease. J. Natl. Cancer Inst. 22, 719–748 (1959).
40. Devlin, B. & Roeder, K. Genomic control for association studies. Biometrics 55,
41. Stefansson, H. et al. A common inversion under selection in Europeans. Nat. Genet.
37, 129–137 (2005).
42. Rozen, S. & Skaletsky, H. Primer3 on the WWW for general users and for biologist
programmers. Methods Mol. Biol. 132, 365–386 (2000).
43. Uyguner, O. et al. Frequencies of gap- and tight-junction mutations in Turkish
families with autosomal-recessive non-syndromic hearing loss. Clin. Genet. 64,
© 2009 Nature America, Inc. All rights reserved.