This page has been archived and is no longer updated
Genome-wide association study of restless legs syndrome identifies common variants in three genomic regions.
Author: J. Winkelmann
Keywords
Keywords for this Article
Add keywords to your Content
Save
|
Cancel
Share
|
Cancel
Revoke
|
Cancel
Rate & Certify
Rate Me...
Rate Me
!
Comment
Save
|
Cancel
Flag Inappropriate
The Content is
Objectionable
Explicit
Offensive
Inaccurate
Comment
Flag Content
|
Cancel
Delete Content
Reason
Delete
|
Cancel
Close
Full Screen
"Genome-wide association study of restless legs syndrome identifies common variants in three genomic regions Juliane Winkelmann 1?3 , Barbara Schormair 1,3 , Peter Lichtner 1,3 , Stephan Ripke 2 , Lan Xiong 4 , Shapour Jalilzadeh 1,3 , Stephany Fulda 2 , Benno Pu�tz 2 , Gertrud Eckstein 1,3 , Stephanie Hauk 1,3 , Claudia Trenkwalder 5 , Alexander Zimprich 6 , Karin Stiasny-Kolster 7 , Wolfgang Oertel 7 , Cornelius G Bachmann 8 , Walter Paulus 8 , Ines Peglau 9 , Ilonka Eisensehr 10 , Jacques Montplaisir 11,12 , Gustavo Turecki 13 , Guy Rouleau 4 , Christian Gieger 14 , Thomas Illig 14 , H-Erich Wichmann 14,15 , Florian Holsboer 2 , Bertram Mu�ller-Myhsok 2,16 & Thomas Meitinger 1,3,16 Restless legs syndrome (RLS) is a frequent neurological disorder characterized by an imperative urge to move the legs during night, unpleasant sensation in the lower limbs, disturbed sleep and increased cardiovascular morbidity. In a genome-wide association study we found highly significant associations between RLS and intronic variants in the homeobox gene MEIS1,theBTBD9 gene encoding a BTB(POZ) domain as well as variants in a third locus containing the genes encoding mitogen-activated protein kinase MAP2K5 and the transcription factor LBXCOR1 on chromosomes 2p, 6p and 15q, respectively. Two independent replications confirmed these association signals. Each genetic variant was associated with a more than 50% increase in risk for RLS, with the combined allelic variants conferring more than half of the risk. MEIS1 has been implicated in limb development, raising the possibility that RLS has components of a developmental disorder. Nightwalkers, as individuals with RLS call themselves, are forced to move their legs during periods of rest especially in the evening and night to relieve uncomfortable or painful sensations in the deep calf 1 . This diurnal variation leads to impaired sleep onset, and the periodic leg movements during sleep in the majority of patients contribute to sleep disruption and a reduced quality of life as a major consequence 2 . There are recognized secondary forms of RLS such as in iron deficiency, pregnancy and end-stage renal disease and associated morbidity such as increased cardiovascular risk 2,3 . RLS is one of the most common neurological disorders, with an age-dependent prevalence of up to 10% in the elderly in North America and Europe 2 . Dopaminergic agents originally developed for Parkinson?s disease have been used to treat RLS, with an unknown mode of action 2 . Neurophysiological, pharmacological and neuroimaging studies suggest that the characteristic symptoms originate in the central nervous system, yet the underlying neurobiology remains obscure 4 . A family history of RLS is present in more than 50% of affected individuals, and similar figures have been reported for heritability in twin studies 5,6 . Linkage analysis uncovered five loci based on recessive (RLS1) or dominant inheritance (RLS2?RLS5), but so far 500,568 SNPs Stage 1: genome-wide Threshold Stage 2: replication Joint analysis Stages 1 + 2ab 401 / 1,644 903 / 891 255 / 287 ba 3 loci 236,758 SNPs P nominal ? 10 ?5 P corr < 0.05 Figure1 Study overview. Numbers refer to cases and controls and SNPs genotyped and analyzed. The 13 most significant SNPs together with neighboring SNPs were replicated in a German (?a?) and a Canadian (?b?) case/control sample. Three loci were confirmed in both stage 2 samples of the study. Received 11 April; accepted 12 June; published online 18 July 2007; doi:10.1038/ng2099 1 Institute of Human Genetics, GSF National Research Center of Environment and Health, D-85764 Neuherberg, Munich, Germany. 2 Max Planck Institute of Psychiatry, D-80804 Munich, Germany. 3 Technical University, Institute of Human Genetics, D-81675 Munich, Germany. 4 Laboratoire d?e�tude des maladies du cerveau, Centre de recherche du CHUM, Ho?pital Notre-Dame, Universite� de Montre�al, Montre�al, Que�bec H2L 4M1, Canada. 5 Paracelsus-Elena-Hospital, 34128 Kassel, Germany. 6 Neurological Department, Medical University of Vienna, 1090 Vienna, Austria. 7 Philipps University Marburg, Department of Neurology, 35039 Marburg, Germany. 8 University of Go�ttingen, Department of Clinical Neurophysiology, 37070 Go�ttingen, Germany. 9 Neurologische Praxis, 10969 Berlin, Germany. 10 Neurologische Praxis Sendlingerstrasse, 80331 Munich, Germany. 11 Centre d?e�tude du sommeil, Ho?pital du Sacre�-C?ur de Montre�al, Montre�al, Que�bec H4J 1C5, Canada. 12 Centre de recherche en sciences neurologiques, Universite� de Montre�al, Montre�al, Que�bec H4J 1C5, Canada. 13 Departments of Psychiatry and Human Genetics, McGill University, Douglas Hospital, Montreal, Quebec H4H 1R3, Canada. 14 Institute of Epidemiology, GSF National Research Center for Environment and Health, 85764 Neuherberg, Munich, Germany. 15 Institute of Medical Informatics, Biometry and Epidemiology, Ludwig-Maximilians-Universita�t, 81377 Munich, Germany, 16 These authors contributed equally to this work. Correspondence should be addressed to J.W. (janew@mpipsykl.mpg.de) or T.M. (meitinger@gsf.de). 1000 VOLUME 39 [ NUMBER 8 [ AUGUST 2007 NATURE GENETICS LETTERS � 200 7 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics no causally related sequence variants have been identified 5?7 . With SNP arrays becoming a mature technology, we conducted a genome- wide association study (GWAS), typing 500,568 SNPs in individuals with RLS and in a large control cohort from the general population. Genome-wide association The study design involved an exploratory stage (stage 1) followed by replication in two further case-control samples (stages 2a and 2b) (Fig. 1). In stage 1, we performed a GWAS, typing cases and controls on a single platform with the Affymetrix 500K Array Set. To enrich for risk alleles and minimize phenotypic hetero- geneity, we selected subjects with familial RLS (n � 401). Controls were selected randomly from a population-based cohort (n � 1,644, from the KORA-S3/F3 survey, described pre- viously) 8 . For statistical analysis, we selected SNPs by including only high-quality geno- types to reduce the number of false-positive signals (Supplementary Table 1 online). A total of 236,758 SNPs passed all quality control filters (mean call rate � 99.48%). The effect of population stratification was negli- gible (inflation factor l � 1.09 via genomic control) 9 (Fig. 2). Eigenvalue-based analysis showed only minimal population substruc- ture (Fig. 2). An Armitage trend test un- covered four SNPs with P values o 10 ?6 (Fig. 3 and Supplementary Table 2 online). After correcting for multiple testing, we identified a single SNP within MEIS1 that reached genome-wide significance (rs2300478, P corrected o 0.0002). Replication of genome-wide findings For stage 2 replication, 13 SNPs passed our inclusion criteria based on P value, location within a linkage peak and visual inspection of clustering data. We selected these and 15 neighboring SNPs for replication. They mapped to six discernible regions. Of these 28 SNPs, 25 were successfully genotyped in stage 2a and 24 in stage 2b (Supplementary Table 3 online). Individuals in 2a (n � 903, familial or sporadic RLS) had been recruited separately using the sampling design of stage 1. Control subjects were selected from KORA-S4 (n � 891). In stage 2a, we found nominally significant evidence for association in five regions, of which three withstood correction for multiple testing (Fig. 4 and Supplementary Table 4 online). The first region was on 2p, located in a 32-kb linkage disequilibrium (LD) block containing exon 9 of MEIS1. Here, two of three SNPs showed significant association (P o 10 ?11 ). MEIS1 is a member of a family of highly conserved TALE homeobox genes. Heterodimers of MEIS1 with PBX and HOX proteins augment the affinity and specificity of DNA binding by HOX proteins 10 . MEIS1 has been found to be overexpressed in acute myeloid leukemia 10 , and studies in Xenopus laevis have shown involvement in neural crest development 11 .In addition, there are several potential links to RLS: during embryonic development, MEIS1 is essential for proximo-distal limb formation 12 , and children with restless legs syndrome are often described as having growing pains 13 . MEIS1 is part of a Hox transcriptional regulatory network that specifies spinal motor neuron pool identity and con- nectivity 14 . Notably, spinal hyperexcitability is an established compo- nent in the genesis of periodic leg movements found in individuals with RLS 15 . Specific functions of MEIS1 in postembryonic tissues still remain to be established. The protein is known to be expressed in the adult mouse brain in cerebellar granule cells, the forebrain and, notably, in dopaminergic neurons of the substantia nigra 16 . The second region with significant association was on chromosome 6p, within a 113-kb LD block in intron 5 of the BTBD9 gene. All five SNPs tested were significant, four of these with P values o10 ?5 . Little is known about BTBD9 other than that it belongs to the BTB(POZ) proteins. BTB stands for broad complex, tramtrack and bric a` brac, genes that in Drosophila melanogaster are required for embryonic 20 15 10 5 0 0 10203040 ? 2 (obs.) ? 2 (e xpected) Figure 2 Extent of population stratification. The distribution of expected (under the null hypothesis) versus observed w 2 values (all P values obtained in the analysis of sample 1, using Armitage trend test with age and sex as covariates) before (blue) and after (red) correction by division with l. Adherence to the diagonal indicates lack of inflation of the statistic. As can be seen in the uncorrected plot, there is evidence for a systematic deviation toward higher-than-expected values. After the correction, there is near- perfect adherence to the diagonal for most of the values obtained, indicating that the correction performed well. Chromosome 1 8 6 4 2 0 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 x Genomic position (Gb) ?log 10 (p) 0.5 1 1.5 2 2.5 3 Figure 3 Genome-wide association study for RLS susceptibility loci. The analysis compared 393 successfully genotyped RLS cases with 1,602 population-based KORA controls. The x-axis represents genomic position, and the y-axis shows ?log 10 (P). Thirteen SNPs that passed inclusion criteria for the replication study of stage 2 are highlighted in bold. Note that the P values of three SNPs on chromosome 15 are very similar, and these SNPs appear as one single dot. NATURE GENETICS VOLUME 39 [ NUMBER 8 [ AUGUST 2007 1001 LETTERS � 200 7 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics development, cell fate determination in the eye, metamorphosis and pattern formation in the limbs 17,18 . Functions of BTB(POZ) proteins include transcription repression, cytoskeleton regulation, tetrameriza- tion and gating of ion channels, as well as ubiquitin- dependent protein degradation 17 .Themodularnatureofthisprotein and the universal occurrence of the particular domains of BTBD9 make assignment of a specific function difficult at present. The third region, defined by seven SNPs tested on 15q, showed significant evidence for association, with P o 10 ?4 . This region contains a 48-kb LD block overlapping the 3� end of MAP2K5,a member of the mitogen-activated protein kinase family, and the adjacent LBXCOR1 gene. MAPK pathways are conserved from yeast to human and are activated by a signaling cascade that mediates the transduction of extracellular signals to cytoplasmic nuclear effec- tors 19 . MAP2K5 is a specific upstream activator of ERK5, and this pathway is activated by oxidative stress, hyperosmolarity and growth factors. In addition, MAP2K5 and ERK5 are abundantly expressed in heart and skeletal muscles, and the MAP2K5/ERK5 MAP kinase cascade is critical at early stages of muscle cell differentia- tion 19 . The possible link between RLS risk alleles and known biological functions of the MAP2K5-ERK5 pathway is of particular interest, as this pathway is important in neuroprotection of dopaminergic neurons 20 . LBXCOR1 is annotated as being downstream of MAP2K5 and acting as a transcriptional corepressor of LBX1. This homeobox gene is critical in the development of sensory pathways in the dorsal horn of the spinal cord that relay pain and touch 21 .Three SNPs within the PTPRD gene in the chromosome 9 linkage region (RLS3) and one SNP on chromosome 16 in the A2BP1 gene were nominally significant. In stage 2b, we genotyped the same SNPs in affected individuals (n � 255) and controls (n � 287) from a French-Canadian popula- tion. Here, we found nominally significant evidence for association in four regions (two SNPs on chromosome 2p, five SNPs on 6p, sevenSNPson15qandoneSNPon16p,Supplementary Table 5 online). The same three regions as in stages 1 and 2a remained significant after correction for multiple testing. Odds ratios (ORs) and risk alleles were very similar to those for stage 2a. Ta bl e 1 shows those nine SNPs in the three loci confirmed in all three sample sets and in joint analysis withstanding genome-wide correction for multiple testing. 12 10 8 ?log 10 ( P nom ) 6 4 2 0 66420200 66485200 66550200 66615200 66680200 66745200 Position MEIS1 rs12469063 rs2300478 12 10 8 ?log 10 ( P nom ) 6 4 2 0 38155100 ZFAND3 BTBD9 GLO1 38220100 38285100 38350100 38415100 38480100 38545100 38610100 38675100 38740100 Position rs3923809 rs4236060 rs4714156 rs9394492 rs9296249 rs9357271 12 10 8 ?log 10 ( P nom ) 6 4 2 0 65535000 65600000 IQCH MAP2K5 LBXCOR1 65665000 65730000 65795000 65860000 65925000 65990000 Position rs4489954 rs3784709 rs1026732 rs6494696 rs884202 rs11635424 rs12593813 a b c Figure 4 Pairwise linkage disequilibrium diagrams for three RLS-associated loci. (a) MEIS1.(b) BTBD9.(c) Region of MAP2K5 and LBXCOR1.TheP values based on the stage 1 Affymetrix data clearly delineate the regions of interest within a single LD block in the limits of the transcribed genomic unit for MEIS1 and three joint LD blocks in BTBD9. For the region of MAP2K5 and LBXCOR1, the region of interest is limited to a single LD block beginning in the transcribed unit of MAP2K5 and ending in the transcribed unit of LBXCOR1. Pairwise LD, measured as D�, was calculated from the stage 1 control data set using the methods of Gabriel as implemented in Haploview. Shading represents the magnitude and significance of pairwise LD, with a white- to-red gradient reflecting lower to higher LD values. Stage 1 Affymetrix SNPs are indicated by red squares, replication SNPs (Stage 2a) by black circles and fine mapping SNPs (Stage 2a) by blue triangles. x-axis shows genomic position, and y-axis shows ?log 10 (P). Transcriptional units are indicated by black arrows, with exons depicted as black bars. 1002 VOLUME 39 [ NUMBER 8 [ AUGUST 2007 NATURE GENETICS LETTERS � 200 7 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics Fine mapping, haplotype and risk analysis We genotyped tagging SNPs and all known coding and splice-site SNPs for fine mapping in the stage 2a samples. This confirmed the candidate regions defined by the explorative phase of the study (Fig. 4). Haplotype analysis for MEIS1 delineated a haplotype block (rs3890755 to rs12469063). A haplotype completely described by allele A (rs6710341) and allele G (rs12469063) was more strongly associated than each single SNP in this block (P� 5.87 C2 10 ?20 ,OR� 2.75 [95% confidence interval, 2.23?3.41]). This haplotype was also maximally associated in the Canadian sample (P � 8.51 C2 10 ?7 , OR � 2.36 [1.40?3.97], Fig. 5). For BTBD9 and the MAP2K5 and LBXCOR1 region haplotype analysis confirmed the results of single-SNP analysis. In exploratory analysis, we compared the ORs obtained under the allele dosage model to those obtained under the unrestricted model. For MEIS1 and BTBD9, we did not find any significant difference between the models tested (MEIS1: P � 0.714; BTBD9: P � 0.913), but the allele dosage model was more parsimonious. For the MAP2K5 and LBXCOR1 region, the allele dosage model was significantly less likely than the unrestricted model (P � 0.006). Estimates pointed to a recessive model. This model was significantly better than the allele dosage model (P � 0.009) and not worse than the unrestricted model (P � 0.395). There was no difference in effect estimates between samples (Supplementary Table 6 online). In the combined German samples, lower limits of the sequential attributable fraction (SAFs) 22,23 were estimated at 0.092, 0.303 and 0.079 for MEIS1, BTBD9 and the MAP2K5 and LBXCOR1 region respectively. Corresponding upper limits (equal to the population attributable risk fraction (ARF)) were 0.227, 0.492 and 0.201. In the Canadian sample, the lower limits of the SAFs were 0.075, 0.316 and 0.090, respectively, and we estimated the upper limits at 0.226, 0.550 and 0.258, respectively. We could not identify any statistical interaction between these loci, either in the individual samples or in the combined German or combined German-Canadian samples. Overall, although the single ARF and SAF estimates may be slightly overestimated, they clearly indicate that the three loci account for a large part of the phenotype in the populations studied. We estimated the ARF jointly attributable to the three loci at 68.6% in the German population and 74.2% for the Canadian population. A comparison of familial versus sporadic cases in the combined stage 1 and 2a data set demonstrated virtually indistinguishable ORs Table 1 Confirmed association results dbSNP ID Chr Genome position Gene MAF (cases) Risk allele MAF (controls) rs2300478 2p 66634957 MEIS1 0.367 (G) G 0.241 (G) rs9296249 6p 38473819 BTBD9 0.162 (C) T 0.235 (C) rs9357271 6p 38473851 BTBD9 0.165 (C) T 0.238 (C) rs12593813 15q 65823906 MAP2K5 0.258 (A) G 0.330 (A) rs11635424 15q 65824632 MAP2K5 0.257 (A) G 0.330 (A) rs4489954 15q 65859129 MAP2K5 0.239 (T) G 0.311 (T) rs3784709 15q 65859329 MAP2K5 0.251 (T) C 0.321 (T) rs1026732 15q 65882139 MAP2K5 0.252 (A) G 0.327 (A) rs6494696 15q 65890260 [MAP2K5/LBXCOR1] 0.253 (C) G 0.326 (C) dbSNP ID OR (95% c.i.) Stage 1 P nom Stage 2a P nom Stage 2b P nom Stage 1+2a+2b P nom Stage 1+2a+2b P corrected a rs2300478 1.74 (1.57?1.92) 4.89E?10 5.93E?12 2.19E?03 3.41E?28 8.08E?23 rs9296249 1.67 (1.49?1.89) 2.19E?06 1.61E?06 4.14E?03 3.99E?18 9.44E?13 rs9357271 1.66 (1.48?1.87) 5.48E?06 1.85E?06 2.48E?03 6.31E?18 1.50E?12 rs12593813 1.50 (1.36?1.66) 1.85E?06 4.95E?05 1.57E?02 1.06E?15 2.51E?10 rs11635424 1.51 (1.37?1.67) 1.77E?06 2.54E?05 6.60E?03 3.65E?16 8.64E?11 rs4489954 1.51 (1.36?1.67) 2.44E?06 2.60E?05 1.66E?02 2.68E?15 6.35E?10 rs3784709 1.52 (1.37?1.68) 3.56E?06 7.46E?05 1.79E?03 4.06E?16 9.61E?11 rs1026732 1.53 (1.39?1.70) 4.67E?07 2.78E?05 5.22E?03 6.09E?17 1.44E?11 rs6494696 1.52 (1.38?1.69) 1.79E?06 5.20E?05 5.22E?03 2.00E?16 4.74E?11 SNPs with significant association that were successfully genotyped in all three case-control samples, located in three different genomic regions. Genome positions refer to the human March 2006 (hg18) assembly. [MAP2K5/LBXCOR1] denotes an intergenic position of the SNP. MAF, minor allele frequency; OR, odds ratio; c.i., confidence interval; P nom � nominal P value. MAF refers to stage 2a data only; OR was calculated using combined data from all stages. P values for stage 1, 2a and combined analysis were calculated using logistic regression implementing an Armitage trend test and taking sex and age as covariates into account. P values in stage 1 and 2a resulting from this regression were further corrected for population stratification by dividing the resulting w 2 by the inflation factor l. a P corrected � P value corrected for multiple testing using Bonferroni?s method, correcting for 236,758 SNPs. Intron 8 MEIS1 C T T T C C A C C C C C T G G G G T A A A G A A C T T T T C A A G G A A 0.287 Cases 0.231 (0.239) Controls 0.102 (0.112) 0.264 0.168 0.133 0.074 0.061 19,705 bp rs3890755 rs4300815 rs4544423 rs6710341 rs7603236 rs12469063 F r equency Figure 5 Haplotype structure for MEIS1. A haplotype consisting of six SNPs (of which rs6710341 and rs12469063 fully tagged the risk haplotype) is associated with RLS with odds ratios of 2.75 and 2.36 in the stage 2a and 2b samples, respectively. Haplotype frequencies for all haplotypes occurring with these six SNPs are based on cases and controls jointly and are given for cases and controls separately for the risk haplotype. For the Canadian sample, the frequencies are given in brackets and are based on the two tagging SNPs. NATURE GENETICS VOLUME 39 [ NUMBER 8 [ AUGUST 2007 1003 LETTERS � 200 7 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics for the regions on 6p and 15q. For the region on 2p, the risk was higher in familial (rs2300478: OR � 1.82 [1.55?2.14]) than in sporadic cases (OR � 1.59 [1.34?1.90]). However, confidence intervals were overlapping with no significant difference in allele distributions (P � 0.22, Supplementary Table 7 online). The familial relative risk figures estimated by the risk to siblings l s were 1.13 for MEIS1,1.02 for BTBD9 and 1.03 for MAP2K5/LBXCOR1 in the German data set, with almost identical estimates in the Canadian data. The increasing medical attention to RLS in recent years is matched by our ignorance about its underlying molecular basis. The genetic heterogeneity of RLS has made linkage studies notoriously difficult and favors association approaches. In agreement with power calcula- tions, an initial genome-wide screen for common variants in 400 cases and 1,600 controls enabled us to detect risk alleles with odds ratios 41.5. Sample size in the replicate was twice as high as in the initial GWAS and provided unequivocal evidence for the signals. The effects were strong enough that a second replication in a small independent sample from Canada also yielded significant signals for all three regions. A particular feature of our study design is the use of a control group from the general population. This provided us with very accurate estimates of the genotype frequencies and it avoided any bias to which a disease-negative population is prone. The identification of significant signals in genes that have not been considered candidates from previous biological knowledge is a recur- ring theme in GWASs 24 . The current knowledge about MEIS1, BTBD9, MAP2K5 and LBXCOR1 opens new avenues of RLS research, and the involvement of developmental genes challenges us to rethink our basic concept of this widespread disease. A major proportion of the risk for RLS is explained by variants in the loci identified. We could not derive any different contributions from any of these loci to familial versus sporadic RLS. The associated variants all convey very low familial relative risk (l s o 1.15 in all cases). The lack of positive results within the known linkage regions does not argue against the validity of the linkage results. The nominally significant signals detected in the RLS3 linkage region might indicate an allelic series of variants conferring weak and strong effects within the same gene. This study is not exhaustive in identifying genetic factors contribut- ing to RLS, and further investigations will provide a better picture of what constitutes the genetic architecture of the complex phenotype of restless legs syndrome. Future studies should investigate endopheno- types or secondary RLS cases, which might show alternative signal patterns. An interesting question is also whether the loci identified have a role in other dopaminergic disorders such as Parkinson?s disease or in other associated disorders such as attention deficit hyperactivity disorder or sleep disorders. Further experimental advances might include features such as higher sample numbers in the exploratory stage, higher SNP density, modification of clustering algorithms 25 , inclusion of lower frequency polymorphisms, investigation of copy number changes and use of lower statistical thresholds using a priori information. METHODS Study population and phenotype assessment. Cases of stages 1 and 2a were of European descent and were diagnosed according to standard criteria 2 in a personal interview. Familial RLS was defined by at least one affected first-degree relative. We excluded subjects with secondary RLS due to uremia, dialysis and iron deficiency. Controls of stage 1 and 2a were of European descent and from the KORA S3/ F3 and S4 surveys, representative of the general population. KORA procedures have been described 8 . For stage 1, we included 1,644 subjects from S3/F3, ages 35?84 years, and for stage 2a, 891 age- and sex-matched subjects from S4. In 2a, 102 affected individuals were outside the age range of KORA and were matched to the next age group. Affected individuals and controls of stage 2b were of French-Canadian ancestry. Affected individuals (n � 255) were diagnosed according to standard criteria 2 , and polysomnography was performed in 156 subjects; of those, 82.1% (n � 128) showed significant periodic leg movements during sleep. Controls were recruited from the general population (n � 287). Secondary cases were excluded. Studies were performed according to the declaration of Helsinki and approved by institutional review boards in Germany, Austria, and Canada. Written informed consent was obtained from participants. For demographic data of successfully genotyped samples, see Supplementary Table 8 online. Genome-wide assays, SNP genotyping and quality control. Stage 1 genotyp- ing was performed using the Affymetrix 500K Array Set. Genotypes were determined using the BRLMM algorithm with cases and controls undergoing a joint cluster analysis. From 500,568 SNPs, a total of 236,758 were selected for subsequent analyses based on stringent quality control criteria. Exclusion criteria were call rate o98% (n � 146,297), minor allele frequency (MAF) o10% (n � 151,583), deviations from HWE (P o 0.00001, n � 22,536) and low number of heterozygotes (o10, n � 33,122). 14,069 SNPs were mono- morphic. For a detailed breakdown, see Supplementary Table 1. For the 13 SNPs passing the inclusion criteria for genotyping in stages 2, visual inspection of clustering was performed using the Affymetrix SNP Signaling Tool 1.0.0.12. All clusters passed this test. To validate the stage 1 experiment, we genotyped 15 SNPs in 400 samples on another platform (Sequenom MassArray system) with a genotype discordance rate of 0.2%. Stage 2 and fine-scale mapping were performed using MALDI-TOF mass spectrometry on a Sequenom system (Autoflex HT and SpectroTYPER RT 3.4 analysis software). Assays were designed using AssayDesign 3.1.2.2 with iPLEX Gold chemistry default parameters. Supplementary Table 9 online lists oligonucleotide sequences of replication and fine mapping. SNP quality control criteria leading to exclusion were call rate o97%, MAF o10% and Po0.001 for deviations from HWE in controls. This resulted in an exclusion of one SNP (rs2110974) in stage 2a, two SNPs (rs2110974, rs7881785) in stage 2b and 51 SNPs in fine mapping. All coding SNPs were monomorphic. A total of 28 affected individuals and 55 controls in stage 2a and 44 affected individuals and 46 controls in stage 2b were excluded owing to low call rate (o90%) of all SNPs within a single DNA sample. SNP selection for stage 2. We used the following inclusion criteria: (i) P o 10 ?6 in stage 1 analysis (four SNPs); (ii) Pr 10 ?5 with two neighboring SNPs (� 100 kb) with P r 10 ?3 (eight SNPs); (iii) P r 10 ?4 for SNPs within described linkage peaks (one SNP in RLS3). For these 13 SNPs, we chose 15 additional neighboring SNPs based on LD structure for genotyping in the replication samples 2a and 2b (Supplementary Table 3). SNP selection for fine mapping. SNPs in the coding regions and 10 kb of flanking sequences were selected using the Tagger algorithm (r 2 � 0.8) implemented in HAPLOVIEW 3.3.2 (ref. 26). In addition, all coding-region SNPs and splice-site SNPs were included. This led to 41 SNPs on chromosome 2p (38 tagging, 1 synonymous and 2 nonsynonymous), 77 SNPs on chromo- some 6p (tagging only) and 46 SNPs on chromosome 15q (37 tagging, 1 synonymous, 4 nonsynonymous, 2 splice site, 2 frameshift coding). In total, 164 SNPs were selected, of which 163 were converted into genotyping assays, and 103 with a MAF 410% were analyzed. Analysis of genetic effects. To test and correct for possible population stratification, we performed an EIGENSOFT 27,28 analysis. We used a random sample of 16,000 SNPs passing the quality criteria for the stage 1 sample and allowed for ten rounds of outlier removal. In the first six rounds, a total of 50 outliers (8 cases and 42 controls) were removed, with none removed in the remaining rounds. To assess stratification, we compared the expected distribution of P values for association versus the expected w 2 distribution with one degree of freedom 29 . We compared the empirically observed mean of the lower 90% of the distribution of the statistics observed and divided it by its expectation 9 . This led to an inflation factor (l) of 1.09 (Fig. 2). 1004 VOLUME 39 [ NUMBER 8 [ AUGUST 2007 NATURE GENETICS LETTERS � 200 7 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics We performed logistic regression analysis coding the number of minor alleles as the dependent variables, thus implementing Armitage?s trend test, including age and sex as covariates and allowing for interactions between age, sex and the number of alleles. Odds ratios and confidence interval limits were obtained through logistic regression analysis. The w 2 values resulting from these analyses (stage 1, 2a and fine mapping) were divided by l, assuming similar conditions for both German samples. Haplotype analysis was performed using HAPLOVIEW 3.3.2 (refs. 26,29), with the fraction of strong LD informative comparisons set at 0.9, and using UNPHASED 3.0.8, which allows the incorporation of age and sex as covariates 30 . Haplotype blocks were delineated using the method of Gabriel implemented in HAPLOVIEW 26 . ORs were obtained using logistic regression with age and sex as covariates in the stage 2 samples. w 2 and P values in 2a were l corrected. Differences between familial and sporadic cases were tested using Fisher?s exact test. Familial attributable risks were calculated using the power calculator described in ref. 22. Multiple testing. Using WG-PERMER, a program for rapid permutation of genome-wide data, preliminary analysis showed that P values after Westfall- Young and Bonferroni correction, with the number of tests set at the number of SNPs tested (n � 236,758), were in good agreement. This may reflect stringent criteria for SNPs to enter the analysis, resulting in low average r 2 values between SNPs. To maintain comparability across results, we show Bonferroni-corrected P values for stage 1 and the combined analysis. For stages 2a and 2b, we give Westfall-Young?corrected P values based on 10,000 permutations, as only a few candidate regions were tested with high LD between them, and thus Bonferroni would be conservative. Power analysis. Power analysis for the combined German sample was per- formed using the Genetic Power Calculator (http://pngu.mgh.harvard.edu/ ~purcell/gpc/). The power of any SNP tested with MAF Z 0.2 and OR Z 1.5 (or 1/1.5 or lower) was beyond 90%. The P value used was the P value required for a significant result after Bonferroni correction (P� 0.05/236,758 � 2.112 C2 10 ?7 ). Testing the mode of inheritance. OR values and likelihoods were obtained using logistic regression analysis with age, sex and samples as covariates in the combined stage 1, 2a and 2b samples. Significance testing between models was done using the likelihood ratio test. Attributable risk fraction. To quantify the contribution of these loci to RLS, we estimated the population attributable risk fraction (ARF) 22 and the sequential attributable fraction (SAF) 23 . We used the allele dosage model for MEIS1 and BTBD9 and the recessive model for MAP2K5 and LBXCOR1 and calculated upper and lower limits of SAFs 23 . For each locus, we used the SNP with the lowest P value, aware of the fact that this may lead to slight overestimation of the ARF. The ARF for the three loci combined was calculated by allowing for the possibility of simultaneous exposure to several of the risk genotypes. Requests for materials: janew@mpipsykl.mpg.de or meitinger@gsf.de. Note: Supplementary information is available on the Nature Genetics website. ACKNOWLEDGMENTS We are grateful to all patients who participated in this study. The authors also thank T.M. Strom, J. Favor, D. Vogt-Weisenhorn, W. Wurst and I. Tews for discussions and R. Feldmann, J. Golic, K. Junghans, B. Schmick, N. Trapp, M. Petzold, G. Fischer and M. Putz for technical assistance. We acknowledge L. Habersack, H. Rhese and J. Schmidt-Evers from the German RLS patient organization for supporting this study. Part of this work was financed by the National Genome Research Network (NGFN). The KORA study group consists of H.-E. Wichmann (speaker), R. Holle, J. John, T. Illig, C. Meisinger, A. Peters and their co-workers, who are responsible for the design and conduct of the KORA studies. The KORA (Cooperative Research in the Region of Augsburg) research project was initiated and financed by the National Research Centre for Environment and Health (GSF), which is funded by the German Federal Ministry of Education and Research and by the State of Bavaria. S.H. was partly supported by a grant from the German RLS patient organization. J.W. was partly supported by a grant form the Bavarian Ministry of Science, Culture and Art. The Canadian part of the study was supported by a Canadian Institutes of Health Research (CIHR) grant to G.R, J.M and G.T. AUTHOR CONTRIBUTIONS Study design: J.W., P.L., G.R., F.H., B.M.-M., T.M.; recruitment and biobanking of individuals with RLS: J.W., S.H., C.T., A.Z., K.S.-K., W.O., C.B., W.P., I.P., I.E., T.M.; recruitment and biobanking of KORA controls: C.G., T.I., H.-E.W.; recruitment and biobanking of Canadian affected individuals and controls: L.X., J.M., G.T., G.R.; Affymetrix genotyping: B.S., P.L., G.E.; Sequenom genotyping: B.S., P.L., S.J.; supervision of typing of all markers: J.W., P.L.; software development and data processing: S.R.,B.P.; statistical analysis: S.R., B.P., B.M.-M.; clustering of Affymetrix genotypes: S.R., B.M.-M.; manuscript writing: J.W., B.S., S.F., L.X., F.H., B.M.-M., T.M. 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/ reprintsandpermissions 1. Walters, A.S. et al. A questionnaire study of 138 patients with restless legs sydrome: the ?night-walkers? survey. Neurology 46, 92?95 (1996). 2. Allen, R.P. et al. Restless legs syndrome: diagnostic criteria, special consider- ations, and epidemiology. A report from the restless legs syndrome diagnosis and epidemiology workshop at the National Institutes of Health. Sleep Med. 4, 101?119 (2003). 3. Winkelman, J.W., Finn, L. & Young, T. Prevalence and correlates of restless legs syndrome symptoms in the Wisconsin Sleep Cohort. Sleep Med. 7, 545?552 (2006). 4. Barrie`re, G., Cazalets, J.R., Bioulac, B., Tison, F. & Ghorayeb, I. The restless legs syndrome. Prog. Neurobiol. 77, 139?165 (2005). 5. Winkelmann, J. et al. Genetics of restless legs syndrome (RLS): state-of-the-art and future directions. Mov. Disord., published online 7 June 2007 (doi:10.1002/ mds.21587). 6. Rao, S., Winkelmann, J. & Wang, Q.K. in Restless Legs Syndrome. Diagnosis and Treatment (ed. Ondo,W.G.) 111?123 (Informa Healthcare, New York, 2007). 7. Kemlink, D. et al. Family-based association study of the restless legs syndrome loci 2 and 3 in a European population. Mov. Disord. 22, 207?212 (2007). 8. Wichmann, H.E., Gieger, C., Illig, T. & MONICA/KORA Study Group KORA-gen- resource for population genetics, controls and a broad spectrum of disease phenotypes. Gesundheitswesen 67, S26?S30 (2005). 9. Clayton, D.G. et al. Population structure, differential bias and genomic control in a large-scale, case-control association study. Nat. Genet. 37, 1243?1246 (2005). 10. Azcoitia, V., Aracil, M., Martinez, A.C. & Torres, M. The homeodomain protein Meis1 is essential for definitive hematopoiesis and vascular patterning in the mouse embryo. Dev. Biol. 280, 307?320 (2005). 11. Maeda, R. et al. Xmeis1, a protooncogene involved in specifying neural crest cell fate in Xenopus embryos. Oncogene 20, 1329?1342 (2001). 12. Mercader, N. et al. Conserved regulation of proximodistal limb axis development by Meis/Hth. Nature 402, 425?429 (1999). 13. Rajaram, S.S., Walters, A.S., England, S.J., Mehta, D. & Nizam, F. Some children with growing pains may actually have restless legs syndrome. Sleep 27, 767?773 (2004). 14. Dasen, J.S., Tice, B.C., Brenner-Morton, S. & Jessel, T.M. A Hox regulatory network establishes motor neuron pool identity and target-muscle connectivity. Cell 123, 477?491 (2005). 15. Bara-Jimenez, W., Aksu, M., Graham, B., Sato, S. & Hallett, M. Periodic limb movements in sleep. State-dependent excitability of the spinal flexor reflex. Neurology 54, 1609?1615 (2000). 16. Allen Institute for Brain Science. Allen Brain Atlas. http://www.brain-map.org (2004). 17. Stogios, P.J., Downs, G.S., Jauhal, J.J., Nandra, S.K. & Prive, G.G. Sequence and structural analysis of BTB domain proteins. Genome Biol. 6, R82 (2005). 18. Godt, D., Couderc, J.L., Cramton, S.E. & Laski, F.A. Pattern formation in the limbs of Drosophila:brica` brac is expressed in both a gradient and a wave-like pattern and is required for specification and proper segmentation of the tarsus. Development 119, 799?812 (1993). 19. Dinev, D. et al. Extracellular signal regulated kinase 5 (ERK5) is required for the differentiation of muscle cells. EMBO Rep. 2, 829?834 (2001). 20. Cavanaugh, J.E., Jaumotte, J.D., Lakoski, J.M. & Zigmond, M.J. Neuroprotective role of ERK1/2 and ERK5 in a dopaminergic cell line under basal conditions and in response to oxidative stress. J. Neurosci. Res. 84, 1367?1375 (2006). 21. Gross, M.K., Dottori, M. & Goulding, M. Lbx1 specifies somatosensory association interneurons in the dorsal spinal cord. Neuron 34, 535?549 (2002). NATURE GENETICS VOLUME 39 [ NUMBER 8 [ AUGUST 2007 1005 LETTERS � 200 7 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics 22. Skol, A.D., Scott, L.J., Abecasis, G.R. & Boehnke, M. Joint analysis is more efficient than replication-based analysis for two-stage genome-wide association. Nat. Genet. 38, 209?213 (2006). 23. Rowe, A.K., Powell, K.E. & Flanders, W.D. Why population attributable fractions can sum to more than one. Am. J. Prev. Med. 26, 243?249 (2004). 24. WTCCC. Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls. Nature 447, 661?678 (2007). 25. Plagnol, V., Cooper, A.D., Todd, J.A. & Clayton, D.G. A method to address differential bias in genotyping in large?scale association studies. Plos Genetics 3, e74 10.1371/ journal.pgen.0030074 (2007). 26. Barrett, J.C., Fry, B., Maller, J. & Daly, M.J. Haploview: analysis and visualisation of LD and haplotype maps. Bioinformatics 21, 263?265 (2005). 27. Patterson, N., Price, A.L. & Reich, D. Population structure and Eigenanalysis. Plos Genetics 2, e190 (2006). 28. Price, A.L. et al. Principal components analysis corrects for stratification in genome- wide association studies. Nat. Genet. 38, 904?909 (2006). 29. Devlin, B. & Roeder, K. Genomic control for association studies. Biometrics 55, 997?1004 (1999). 30. Dudbridge, F. UNPHASED user guide. Technical Report 2006/5. MRC Biostatistics Unit, Cambridge, UK, (2006). 1006 VOLUME 39 [ NUMBER 8 [ AUGUST 2007 NATURE GENETICS LETTERS � 200 7 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics "
Add Content to Group
|
Bookmark
|
Keywords
|
Flag Inappropriate
share
Close
Digg
Facebook
MySpace
Google+
Comments
Close
Please Post Your Comment
*
The Comment you have entered exceeds the maximum length.
Submit
|
Cancel
*
Required
Comments
Please Post Your Comment
No comments yet.
Save Note
Note
View
Public
Private
Friends & Groups
Friends
Groups
Save
|
Cancel
|
Delete
Please provide your notes.
Next
|
Prev
|
Close
|
Edit
|
Delete
Genetics
Gene Inheritance and Transmission
Gene Expression and Regulation
Nucleic Acid Structure and Function
Chromosomes and Cytogenetics
Evolutionary Genetics
Population and Quantitative Genetics
Genomics
Genes and Disease
Genetics and Society
Cell Biology
Cell Origins and Metabolism
Proteins and Gene Expression
Subcellular Compartments
Cell Communication
Cell Cycle and Cell Division
Scientific Communication
Career Planning
Loading ...
Scitable Chat
Register
|
Sign In
Visual Browse
Close
Comments
CloseComments
Please Post Your Comment