OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a...

17
EXTENDED REPORT Novel genetic variants associated with lumbar disc degeneration in northern Europeans: a meta-analysis of 4600 subjects Frances M K Williams, 1 Aruna T Bansal, 2 Joyce B van Meurs, 3,4 Jordana T Bell, 1 Ingrid Meulenbelt, 4,5 Pradeep Suri, 6 Fernando Rivadeneira, 3,4 Philip N Sambrook, 7 Albert Hofman, 8 Sita Bierma-Zeinstra, 9 Cristina Menni, 1 Margreet Kloppenburg, 10 P Eline Slagboom, 4,5 David J Hunter, 7 Alex J MacGregor, 1 Andre G Uitterlinden, 3 Tim D Spector 1 Additional data are published online only. To view this le please visit the journal online (http://dx.doi.org/10. 1136/annrheumdis-2012- 201551). For numbered afliations see end of article. Correspondence to Dr Frances M K Williams, Department Twin Research and Genetic Epidemiology, Kings College London, St Thomas Hospital, Westminster Bridge Road, London SE1 7EH, UK; [email protected] ATB and JBM contributed equally. Accepted 19 July 2012 Published Online First 19 September 2012 ABSTRACT Objective Lumbar disc degeneration (LDD) is an important cause of low back pain, which is a common and costly problem. LDD is characterised by disc space narrowing and osteophyte growth at the circumference of the disc. To date, the agnostic search of the genome by genome-wide association (GWA) to identify common variants associated with LDD has not been fruitful. This study is the rst GWA meta-analysis of LDD. Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable on all forms of imaging (plain radiograph, CT scan and MRI) and performed a meta-analysis of ve cohorts of Northern European extraction each having GWA dataimputed to HapMap V.2. Results This study of 4600 individuals identied four single nucleotide polymorphisms with p<5×10 8 , the threshold set for genome-wide signicance. We identied a variant in the PARK2 gene (p=2.8×10 8 ) associated with LDD. Differential methylation at one CpG island of the PARK2 promoter was observed in a small subset of subjects (β=8.74×10 4 , p=0.006). Conclusions LDD accounts for a considerable proportion of low back pain and the pathogenesis of LDD is poorly understood. This work provides evidence of association of the PARK2 gene and suggests that methylation of the PARK2 promoter may inuence degeneration of the intervertebral disc. This gene has not previously been considered a candidate in LDD and further functional work is needed on this hitherto unsuspected pathway. INTRODUCTION Lumbar disc degeneration (LDD) is a common, age-related trait: 1 over a third of middle aged women have at least one degenerate disc. LDD contributes to low back pain 23 and as low back pain is common in the general population and costly to society, 4 LDD is of considerable public health importance. Discrete biochemical, histo- logical, metabolic and functional changes occur in LDD, such that discs become dehydrated, lose disc height and there is accompanying outgrowth of osteophytes from the vertebral body margin. 5 There are similarities with peripheral joint osteoarthritis (OA). LDD has been shown to be heritable, with estimates of 65%80% 67 and so a considerable proportion of the variance in LDD is explained by genetic factors. Yet to date, candidate gene studies have detected only a small number of convincing associations of genetic variants with LDD (reviewed by Ryder et al 8 ). A number of studies show conicting results: these are likely due to small sample size or may reect ethnic dif- ferences between Northern European and Asian populations, as seen in OA. 9 That some published genome-wide associations (GWAs) in common complex traits fail to replicate candidate gene nd- ings suggests limitations to the candidate gene method. 10 As in other common complex traits, 11 a considerable proportion of the genetic variance in LDD remains unexplained. While GWA studies do not capture all variation in the genome, the approach does offer an agnostic search of the genome for variants associated with common complex traits. Their main limitation results from the inherent multiple testing in their design, meaning that power is lost and large samples are needed to address this. In order to optimise sample size in the present study we performed meta-analysis of GWAS using a number of cohorts having the LDD phenotype. PATIENTS AND METHODS The cohorts available for inclusion in this study were all population samples, except Genetics of osteoARthrosis and Progression study (GARP) which specically recruited participants having a diagnosis of OA. A variable was derived from mea- sures of disc height and osteophytes obtained from lateral images on MR, CT scan or plain radiograph. Summing this variable over the lumbar discs pro- vided a continuous measure of disc degeneration. GWA of this summary variable was performed by each individual study group and summary statis- tics were sent to and collated by KCL. Meta-analysis was performed of imputed GWA data from ve population cohorts (Framingham, GARP, Rotterdam study 1 and 3 and TwinsUK (TUK)) having imaging of the spine (see below). All cohorts had obtained fully informed consent Open Access Scan to access more free content OPEN ACCESS Ann Rheum Dis 2013;72:11411148. doi:10.1136/annrheumdis-2012-201551 1141 Clinical and epidemiological research

Transcript of OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a...

Page 1: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

EXTENDED REPORT

Novel genetic variants associated with lumbar discdegeneration in northern Europeans a meta-analysisof 4600 subjectsFrances M K Williams1 Aruna T Bansal2 Joyce B van Meurs34 Jordana T Bell1

Ingrid Meulenbelt45 Pradeep Suri6 Fernando Rivadeneira34 Philip N Sambrook7

Albert Hofman8 Sita Bierma-Zeinstra9 Cristina Menni1 Margreet Kloppenburg10

P Eline Slagboom45 David J Hunter7 Alex J MacGregor1 Andre G Uitterlinden3

Tim D Spector1

Additional data arepublished online only To viewthis file please visit the journalonline (httpdxdoiorg101136annrheumdis-2012-201551)

For numbered affiliations seeend of article

Correspondence toDr Frances M K WilliamsDepartment Twin Researchand Genetic EpidemiologyKingrsquos College LondonSt Thomas HospitalWestminster Bridge RoadLondon SE1 7EH UKfranceswilliamskclacuk

ATB and JBM contributedequally

Accepted 19 July 2012Published Online First19 September 2012

ABSTRACTObjective Lumbar disc degeneration (LDD) is animportant cause of low back pain which is a commonand costly problem LDD is characterised by disc spacenarrowing and osteophyte growth at the circumference ofthe disc To date the agnostic search of the genome bygenome-wide association (GWA) to identify commonvariants associated with LDD has not been fruitful Thisstudy is the first GWA meta-analysis of LDDMethods We have developed a continuous trait basedon disc space narrowing and osteophytes growth whichis measurable on all forms of imaging (plain radiographCT scan and MRI) and performed a meta-analysis of fivecohorts of Northern European extraction each havingGWA data imputed to HapMap V2Results This study of 4600 individuals identified foursingle nucleotide polymorphisms with plt5times10minus8 thethreshold set for genome-wide significance We identifieda variant in the PARK2 gene (p=28times10minus8) associatedwith LDD Differential methylation at one CpG island ofthe PARK2 promoter was observed in a small subsetof subjects (β=874times10minus4 p=0006)Conclusions LDD accounts for a considerableproportion of low back pain and the pathogenesis ofLDD is poorly understood This work provides evidenceof association of the PARK2 gene and suggests thatmethylation of the PARK2 promoter may influencedegeneration of the intervertebral disc This gene hasnot previously been considered a candidate in LDD andfurther functional work is needed on this hithertounsuspected pathway

INTRODUCTIONLumbar disc degeneration (LDD) is a commonage-related trait1 over a third of middle agedwomen have at least one degenerate disc LDDcontributes to low back pain2 3 and as low backpain is common in the general population andcostly to society4 LDD is of considerable publichealth importance Discrete biochemical histo-logical metabolic and functional changes occur inLDD such that discs become dehydrated lose discheight and there is accompanying outgrowth ofosteophytes from the vertebral body margin5

There are similarities with peripheral joint

osteoarthritis (OA) LDD has been shown to beheritable with estimates of 65ndash806 7 and so aconsiderable proportion of the variance in LDD isexplained by genetic factors Yet to date candidategene studies have detected only a small number ofconvincing associations of genetic variants withLDD (reviewed by Ryder et al8) A number ofstudies show conflicting results these are likelydue to small sample size or may reflect ethnic dif-ferences between Northern European and Asianpopulations as seen in OA9 That some publishedgenome-wide associations (GWAs) in commoncomplex traits fail to replicate candidate gene find-ings suggests limitations to the candidate genemethod10 As in other common complex traits11 aconsiderable proportion of the genetic variance inLDD remains unexplained While GWA studies donot capture all variation in the genome theapproach does offer an agnostic search of thegenome for variants associated with commoncomplex traits Their main limitation results fromthe inherent multiple testing in their designmeaning that power is lost and large samples areneeded to address this In order to optimise samplesize in the present study we performedmeta-analysis of GWAS using a number of cohortshaving the LDD phenotype

PATIENTS AND METHODSThe cohorts available for inclusion in this studywere all population samples except Genetics ofosteoARthrosis and Progression study (GARP)which specifically recruited participants having adiagnosis of OA A variable was derived from mea-sures of disc height and osteophytes obtained fromlateral images on MR CT scan or plain radiographSumming this variable over the lumbar discs pro-vided a continuous measure of disc degenerationGWA of this summary variable was performed byeach individual study group and summary statis-tics were sent to and collated by KCLMeta-analysis was performed of imputed GWAdata from five population cohorts (FraminghamGARP Rotterdam study 1 and 3 and TwinsUK(TUK)) having imaging of the spine (see below)All cohorts had obtained fully informed consent

Open AccessScan to access more

free content

OPEN ACCESS

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1141

Clinical and epidemiological research

from their participants and appropriate ethics committeeapproval In all studies a cumulative degeneration score wasconstructed from the sum of scores of degenerative change ateach level (disc space narrowing coded 0ndash3 and osteophyteseither anterior or posterior or both coded 0ndash3) In thosecohorts where only four disc levels were read (FraminghamHeart study (FHS)) a fifth level was imputed by taking themean reading for four discs as a surrogate for the fifth disc andsumming over five discs The data underwent inverse normaltransformation to generate a normally distributed variable

Phenotyping the cohortsThe FHS is a longitudinal cohort of a defined populationin Massachusetts initiated in 1948 (httpwwwframinghamheartstudyorg) It began as a study sample of 5209Framingham men and women between the ages of 30 and 60Subsequently offspring and third generation subjects wereincorporated Every other year after an extensive baselineexamination subjects undergo testing that includes a medicalhistory blood profile echocardiogram and bone eye and othertests The subset of the Framingham subjects covered by thecurrent analysis comprised 366 subjects from the Offspring andGeneration three arms of the study who had undergone CTscanning of the spine and the recruitment conduct and specifi-cations of CT scanning have been reported elsewhere12

Measurement of the lumbar spine CTs for disc height andscoring (0ndash3) for anterior and posterior osteophytes was per-formed by a spine specialist using the mid-sagittal plane atspinal levels L2-L3 L3-L4 L4-L5 and L5-S1 by author PS usingthe atlas of Jarosz et al7 Using sagittal CT reformatting themid-sagittal plane was determined at each spinal level and mea-surements of disc height in millimetres were made The mea-sured values for disc height (mm) were converted to 0ndash3categorical scale for disc height loss Using an imputed valuefor the fifth lumbar disc based on the mean value of the mea-sured four discs values for disc height loss and anterior andposterior osteophytes were summed over five lumbar disclevels

The GARP study comprises white sibling pairs of Dutchorigin affected by OA at multiple sites and is aimed at identify-ing determinants of OA susceptibility and progressionProbands (ages 40ndash70 years) and their siblings had OA at mul-tiple joint sites of the hand or in two or more of the followingjoint sites hand spine (cervical or lumbar) knee or hip asdescribed previously13 Subjects included in this study hadundergone lateral radiographs of the spine (T4-S1) Each inter-vertebral disc level from L12 to L5S1 was reviewed for thepresence and severity of osteophytes (anterior) and disc narrow-ing using the Lane atlas14 where 0=none grade 1=mild grade2=moderate and grade 3=severe The score at each level foranterior osteophytes and disc height loss were summed overthe five lumbar levels

The Rotterdam study is a prospective population-basedfollow-up study of the determinants and prognosis of chronicdiseases in the elderly15 16 All persons living in Ommoord asuburb of Rotterdam who were aged 55 years and over wereinvited to participate A total of 7983 participants were exam-ined For the current analysis two subsets of the data wereconsidered Rotterdam cohort 1 (RS1) consists of 2440 subjectsRotterdam cohort 3 (RS3) consists of 974 subjects Subjects ori-ginating from the Rotterdam study underwent plain radiog-raphy and scoring of LDD as previously described2 In brieflateral lumbar radiographs were scored by a single observer forthe presence of the individual radiographic features of disc

degeneration Each intervertebral disc from L12 to L5S1 wasreviewed for the presence and severity of osteophytes (anterior)and disc narrowing using the Lane atlas as described above14

The scores for the two traits over the five lumbar discs weresummed

The TUK registry was described previously17 The register wasstarted in 1993 and now comprises of approximately 10 000monozygotic and dizygotic adult Caucasian twins aged16ndash85 years from all over the UK plus some parents and sib-lings It now incorporates previous twin registries from theInstitute of Psychiatry and Aberdeen University This is a vol-unteer sample recruited by successive media campaignswithout selecting for particular diseases or traits All twinsreceive a series of detailed disease and environment question-naires The majority of twins have been assessed in detail clin-ically at several time points for several hundred phenotypesrelated to common diseases or intermediate traits The subsetof TUK covered by the current analysis consisted of 744 sub-jects who had participated in the spine MR study (scanned1996ndash2000) using a Siemens MR machine with (MunichGermany) 10-tesla superconducting magnet Serial sagittalimages of the cervical thoraco-lumbar junction and lumbarspine (T9-L5) were obtained7 Images were coded for discheight loss and anterior osteophytes using a 0ndash3 scale in eachcase where 0 is normal and 3 maximal degeneration as per theatlas of Jarosz et al7 All five lumbar discs were scored and thescores summed to give a combined LDD variable

Genotyping and imputationFHS subjects were genotyped using Affymetrix GeneCHipHuman Mapping 500 K array set (Affymetrix Santa Clara CAUSA) andor the 100 K array set andor the 50 K arrayMethods and quality controls have been described previously18

GARP subjects were genotyped using Illumina Human660WQuad BeadChips (HumanHap550v3 HumanHap610 IlluminaSan Diego CA USA) Genotyping was performed at the geno-typing Rotterdam Genotyping Centre Positive strand geno-types were called by clustering in Genome studio andimputation was performed using IMPUTE software andhapmap phase II V2119 20 Strict selection criteria were appliedto the measured genotypes using a high information content(r2 of gt95) and a minor allele frequency gt00025 Associationanalyses were performed using an inhouse developed softwarepackage that allows the analyses of family data using all infor-mation available in the cases and controls by extending theCochran-Armitage trend test21

RSI and RS3 subjects in the Rotterdam Study sets were geno-typed on the HumanHap550v3 (RS1) or HumanHap610 (RS3)Genotyping BeadCHip (Illumina San Diego California USA)The following sample quality control criteria were appliedsample call rate gt975 gender mismatch with typed X-linkedmarkers evidence for DNA contamination in the samples usingthe mean of the autosomal heterozygosity gt033 exclusion ofduplicates or first-degree relatives identified using Identity byState probabilities and exclusion of outliers (four SD awayfrom the population mean using multidimensional scaling ana-lysis with four principal components) Filtering criteria forimputation are summarised in supplementary table S1

TUK subjects were genotyped using a combination ofIllumina arrays (Human Hap300 and the Human Hap610Q)Genotyping was performed by the Wellcome Trust SangerInstitute using the Infinium assay (Illumina) across threegenome-wide single nucleotide polymorphism (SNP) sets asdescribed previously22 Genotyping results had been sent to

1142 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

KCL for collation and analysis using the statistical packageSTATA (StataCorp)23 Strict quality control was applied314 075 SNPs were retained for analysis (987) 733 wereexcluded because their call rates were le90 and 725 SNPs hadminor allele frequency lt001 In TUK significant populationsubstructure was excluded using the STRUCTURE program24

GWA analysisAll analyses were performed on inverse normal transformedsummary LDD score as described above Each study performedGWA analysis for LDD scores using either MACH2QTL (httpwwwsphumicheducsgabecasisMACHindexhtml) (RS1and RS3) or SNPTEST (httpwwwstatsoxacuk~marchinisoftwaregwassnptesthtml) (GARP) which use genotypedosage value as continuous additive predictors of LDD score ina linear regression framework or ProbABEL25 using an additivegenetic model while accounting for relatedness between themembers of a family Analysis of imputed genotype dataaccounted for uncertainty in each genotype prediction by usingeither the dosage information from MACH26 or the genotypeprobabilities from IMPUTE19

Meta-analysis of the five study groupsGenotypes for 25ndash3 million autosomal SNPs were imputed sep-arately to increase coverage using HapMap V2 (httpwwwhapmaporg) as the reference panel In GARP and TUK imput-ation was performed with IMPUTE V219 and in the otherstudies with MACH26 The common reference panel led to thereporting of results for the positive strand for all cohorts In add-ition allele pairs were compared between cohorts and no detect-able strand-flips were found the minor allele frequency was alsocompared between datasets The distributions of β values of thecohorts were found to be similar and therefore suitable formeta-analysis All directly genotyped or imputed autosomalSNPs having information from more than one study group(n=2 552 511) were included in the meta-analysis Associationresults were combined using inverse variance weighted fixedeffects meta-analysis using PLINK V106 (httppngumghharvardedupurcellplink) Two meta-analyses were run thefirst was unadjusted the second was adjusted for age and sex asboth are known risk factors for LDD and each risk factor wascorrelated with LDD in each study group Heterogeneity of esti-mated effect was expressed using Q (weighted sum of squares)and I2 (ratio of true heterogeneity to total observed variation)SNPs were excluded from the meta-analysis if the cohort-specificimputation quality as assessed by r2 (MACH) or InformationScore (IMPUTE) metric was lt040 On this basis one markerwas excluded from the unadjusted association and one from theadjusted association

DNA methylation data and analysisWhole blood DNA methylation levels were obtained for 38individuals in the TUK cohort using the IlluminaHumanMethylation27 DNA Analysis BeadChip assay as previ-ously described27 The sample included four monozygotic twinpairs eight dizygotic twin pairs and 14 unrelated individualsAt each CpG site within an individual the methylation levelwas presented as β which represents the ratio of intensitysignal obtained from the methylated beads in the array overthe sum of methylated and unmethylated bead signalsFollowing quality control checks we obtained DNA methyla-tion at three CpG sites in the promoter region of the PARK2gene within 2kb of the transcription start site The threeprobes (cg15832436 cg21926612 and cg24816866) mappeduniquely to the human genome (hg18) within two mismatches(see Bell et al28) We fitted linear mixed effects models to assessassociation between DNA methylation levels at the three CpGsites in the PARK2 promoter and LDD We regressed the rawmethylation levels on fixed-effect terms including methylationchip and LDD and random-effect terms denoting family struc-ture and zygosity and compared the association of differen-tially methylated regions with a null model which excludedLDD from the fixed-effects terms We also repeated the associ-ation analyses by normalising the methylation values at eachCpG site to N(0 1)

RESULTSThe study samples for the meta-analysis included 4683 indivi-duals of European ancestries Table 1 shows sample size demo-graphic characteristics LDD and lumbar spine imaging methodfor each independent cohort The majority of participants werefemale subjects (670) and the samples had a mean age of577 years Across the cohorts the mean level of LDD variedfrom 0011 to 346 reflecting differences in imaging methodsHowever the variance of the LDD variables was broadlysimilar (range 0958ndash114) as were the distributions of the esti-mated genetic effect sizes (β) The genotyping and imputationmethods are shown in online supplementary table S1

Quantilendashquantile plots for the unadjusted LDD GWA meta-analysis are presented in figure 1 (see online supplementaryfigure S1 adjusted) Test statistic inflation post meta-analysisas measured by the genomic control statistic29 was low(λGC unadjusted=102 λGC adjusted=103) Results of theunadjusted and adjusted association analyses were broadlysimilar with the p values of the adjusted analysis somewhatattenuated A Manhattan plot for the unadjusted analysis isshown in figure 2 with numeric results in table 2 (unadjusted)and online supplementary table S2 (adjusted) for SNPs havingplt10minus5

Table 1 Characteristics of the study samples

FHS GARP RS1 RS3 TUK

N 330 192 2440 974 744Age (years) 543 (110) 603 (71) 657 (67) 547 (34) 536 (83)Women () 422 797 57 59 968BMI (kgm2) 281 (51) 2644 (48) 263 (34) 2712 (46) 249 (44)Lumbar spine imaging CT Radiograph Radiograph Radiograph MRILDD variable 249 (097) 002 (0958) 0006 (0978) 0011 (0965) 346 (114)

Values are mean (SD) unless specified otherwiseBMI body mass index FHS Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study LDD Lumbar disc degeneration RS1 Rotterdam study cohort 1RS3 Rotterdam study cohort 3 TUK TwinsUK

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1143

Clinical and epidemiological research

Four markers achieved genome-wide significance in theunadjusted GWAS of which three were on chromosome 6(rs926849 rs2187689 rs7767277) and an intergenic marker onchromosome 3 (rs17034687) The results of the meta-analysisadjusted for age and gender were broadly similar the strongestsignal was for SNP rs926849 This SNP lies on an intronicregion of the Parkinson protein 2 E3 ubiquitin protein ligase(PARK2) gene on chromosome 6 A Forest plot of the groupsrsquoresults and their meta-analysis is shown in figure 3 Data were

available from four study groups and the range of estimatedminor allele frequencies was 023ndash032 Imputation quality washigh for all four studies contributing this SNP (gt090 table 2)and the estimated between-study heterogeneity was acceptable(I2=0 p(Q)=067) The results of a four-study meta-analysiswhich excluded GARP show the marker to have p=95eminus8

(unadjusted) The minor or C allele of rs926849 was associatedwith a lower level of LDD implying that the minor allele is pro-tective Figure 4 shows a regional plot of both genotyped andimputed SNPs within 200 Kb of the PARK2 gene along withrecombination rates

Two of the other strongly associated SNPs are in perfectlinkage disequilibrium (LD) rs2187689 and rs7767277 onchromosome 6 Data were available for four studies and therange of estimated allele frequency was 006ndash009 Imputationquality was high for all four studies (gt090) Both SNPs are instrong LD (r2=076) with an intronic marker on the prote-asome subunit β type 9 large multifunctional peptidase 2 gene(PSMB9) that is located in the class II region of the majorhistocompatibility complex (MHC) Both genotyped andimputed SNPs within 400 Kb of rs2187689 along with recom-bination rates are shown in a regional plot in onlinesupplementary figure S3 None of these top SNPs is in LD withknown functional SNPs in either PARK2 or PSMB9

We tested for an association between LDD and DNA methy-lation variants at three CpG sites in the PARK2 promoterA significant association between DNA methylation at CpGsite cg15832436 and LDD (β=874times10minus4 SE=249times10minus4p=0006) was observed The pattern of hypermethylation withincreasing LDD levels was reflected at the remaining promoterCpG sites however these did not reach nominal significance(cg21926612 β=0003 p=014 cg24816866 β=676times10minus4p=039) We repeated the analyses using normalised methyla-tion levels and observed that the association betweencg15832436 and LDD remained nominally significant

DISCUSSIONGWA offers an unbiased scan of common genetic variants(minor allele frequency gt5) and thus may deliver novel var-iants in genes not hitherto suspected of playing a role in discdegeneration This work is among the first to report on agenome-wide meta-analysis being conducted for LDD LDD isan age-related process which occurs in all people to someextent and may be detected as early as the teenage years1 LDDis known to have genetic determinants7 16 and its expression isalso influenced by gender (women develop LDD later) bodymass index30ndash34 and smoking35 Occupational factors also playa small role in LDD36 37 LDD as determined by MRI has beenimplicated in the development of episodes of severe and disab-ling low back pain3 We undertook this large meta-analysis inorder to identify novel genetic variants associated with LDDand to shed light on the underlying pathology of discdegeneration

GWA data obtained using differing chip technology may bereadily compared using imputation with HapMap In total2 552 511 overlapping markers were available in each cohortWe identified four markers having significant association withthe LDD phenotype (plt5times10minus8) There was similarity in theresults obtained with and without adjustment for the covari-ates age and sex A total of 26 markers had plt10minus5 in bothmeta-analyses As expected results of the adjusted analyseshad slightly attenuated p values (see online supplementarytable S2) which likely reflect the confounding effect of age Inboth analyses there were multiple associations to the Human

Figure 1 Quantilendashquantile plot of observed versus expecteddistribution of p values for the genome-wide association (GWA)meta-analysis The plots show GWA meta-analysis quantilendashquantileplot of observed against expected results unadjusted for covariates

Figure 2 Manhattan plot for meta-analysis of unadjustedgenome-wide association results Plot shows combined results for thefive studies included in the meta-analysis unadjusted results The blueand red horizontal lines mark the levels of suggestive and likelysignificance respectively

1144 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

Leukocyte Antigen (HLA) region and to markers in PARK2(Parkinson protein 2 E3 ubiquitin protein ligase) Among themost significant findings (table 2) is SNP rs926849 that lies at6q252ndash27 within an intron in the PARK2 gene a large gene of13 Mb comprising 12 exons The SNP encodes a change of basefrom T to C and is reported to have a minor allele frequency of023ndash034 in dbSNP which is keeping with the findings in ourstudy groups (table 2 figures 3 and 4) Although this SNP hasnot been directly genotyped by any study group estimates

suggest imputation to be accurate for rs926849 (range 95ndash

99 table 2) PARK2 encodes a protein called parkin which isa component of a multiprotein E3 ubiquitin ligase complexthat mediates the targeting of unwanted proteins for proteaso-mal degradation This complex also controls the level of pro-teins involved in cell activities such as cell division and growthit may be a tumour suppressor protein Parkin is also involvedin mitophagy it translocates from the cytosol and promotesthe degradation of uncoupled mitochondria38 Parkin is widely

Table 2 Results of the genome-wide association meta-analysis (unadjusted showing those SNPs having plt10minus5)

MAF Imputation quality

SNP Chr Gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP Eff all β SE p Value

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 023 0038 182E-09rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 023 0041 272E-08rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 023 0041 281E-08rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C minus013 0024 325E-08rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 02 0037 558E-08rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 02 0037 559E-08rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 019 0036 936E-08rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 019 0036 939E-08rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 019 0036 104E-07rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 019 0037 146E-07rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 019 0037 147E-07rs10214886 6 NA 4479 01 009 007 011 NA 098 1 092 097 NA A 019 0036 232E-07rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 019 0037 322E-07rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A minus012 0024 361E-07rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A minus012 0023 368E-07rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 019 0038 458E-07rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C minus026 0054 947E-07rs7896691 10 PFKP 4652 01 01 009 013 007 099 1 096 067 9761 C 017 0035 202E-06rs6602024 10 PFKP 4673 01 01 009 013 007 099 1 098 067 9762 A 017 0035 220E-06rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C minus011 0023 356E-06rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A minus053 0114 359E-06rs17132175 10 PFKP 4659 01 01 009 013 007 098 099 097 063 9713 C 016 0036 373E-06rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 011 0023 375E-06rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C minus013 0028 399E-06rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A minus011 0025 432E-06rs2484990 10 NA 4429 001 001 0 002 NA 056 082 047 093 NA C 068 0148 439E-06rs1250307 10 NA 4415 001 001 0 002 NA 056 082 045 094 NA A 068 0148 439E-06rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 011 0023 443E-06rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A minus013 0029 455E-06rs2484992 10 NA 4440 001 001 0 002 NA 056 082 048 093 NA C 068 0149 481E-06rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A minus028 0061 517E-06rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0023 543E-06rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0047 564E-06rs11918654 3 ARL8B 4310 027 027 029 027 NA 097 098 09 099 NA C minus011 0024 710E-06rs4881085 10 PFKP 4490 01 01 01 013 NA 1 1 099 067 NA A 016 0036 744E-06rs2657195 8 NA 4316 022 021 018 021 NA 085 085 087 074 NA A minus013 0028 754E-06rs11754641 6 NA 4478 003 003 002 004 NA 093 095 091 1 NA C 029 0064 784E-06rs12805875 11 NA 4674 042 041 042 043 041 1 1 1 1 9812 A 009 0021 851E-06rs980238 8 NA 4491 03 029 031 029 NA 1 1 1 1 NA A minus01 0023 911E-06rs7103004 11 NA 4667 042 041 042 043 041 1 1 099 1 9709 C 009 0021 920E-06rs4554859 11 NA 4674 042 041 042 043 041 1 1 1 1 9812 G 009 0021 924E-06rs7118412 11 NA 4669 042 041 042 043 041 1 1 099 1 9709 A 009 0021 937E-06rs2017567 16 NA 4683 042 043 042 042 047 099 1 1 045 100 C 01 0021 938E-06rs710924 16 PIGQ 4683 042 043 042 042 047 1 1 1 045 9964 C 009 0021 993E-06

r2 from MACH for RS1 RS3 FHS and GARP information score from IMPUTE for TUKStudies contributing data are denotedSNP single nucleiotide polymorphism Chr chromosome N number of subjects studied MAF minor allele frequency RS1 Rotterdam study cohort 1 RS3 Rotterdam studycohort 3 TUK TwinsUK FHS Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study Eff all effect allele β effect size SE SE of β

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1145

Clinical and epidemiological research

expressed in solid organs as well as skeletal muscle (httpwwwproteinatlasorg) Mutations within PARK2 are associated withdiverse conditions including autosomal recessive juvenileParkinsonrsquos disease Alzheimer rsquos disease diabetes mellitus andseveral solid tumours (reviewed in39) Parkin may account for theinverse relation between Parkinsonrsquos disease and cancer inci-dence40 Our findings of hypermethylation with increasing LDDscore suggest that PARK2 expression is reduced with increasingdisc degeneration but functional studies of intervertebral disc andother spine tissues are needed

Three further markers in the unadjusted meta-analysis hadplt5times10minus8 Marker rs17034687 is an intergenic marker onchromosome 3 Based on One Thousand Genomes (1KG)CEUdata it is not in LD (r2gt 03) with any known gene-basedmarkers Markers rs2187689 (supplementary figure S2) andrs7767277 are HLA region markers neither of which is includedin the 1KG pilot data Using data from HapMap V3 (release 2)rs2187689 and rs776277 are in perfect LD with each other andin LD (r2= 076) with an intronic marker in PSMB9 (prote-asome (prosome macropain) subunit β type 9 large

multifunctional peptidase 2) Proteasomes are distributedthroughout eukaroytic cells at high concentration and cleavepeptides in an ATPubiquitin-dependent process in a non-lysosomal pathway The gene is located in the class II region ofthe MHC Expression of the gene is induced by interferon γand this gene product replaces catalytic subunit 1 (proteasomeβ 6 subunit) in the immunoproteasome

While lumbar degeneration is not considered an inflamma-tory process and has not been reported to be auto-immune inaetiology there is evidence of pro-inflammatory cytokine acti-vation in herniated lumbar discs41 and anti-TNF has been usedsuccessfully to treat disc herniation-induced sciatica42 Of notethe COL11A2 gene lies 169 KB upstream from rs2187689 AnSNP (rs2076311) within this candidate gene has been shown tobe associated with MR determined disc signal intensity in acandidate gene study of Finnish male twins43 SNP rs2076311is not however in LD with our top hit rs2187689 (r2=0017)and so it seems unlikely that this collagen-encoding geneaccounts for our observed association Many published GWAstudies have identified SNPs in intergenic regions and the

Figure 4 Regional plot of association results and recombination rates for the PARK2 gene unadjusted results minuslog10 p values (y-axis) of the singlenucleotide polymorphisms (SNPs) are shown according to their chromosomal positions (x-axis) with lead SNP shown as a purple diamond Thecolour intensity of each symbol depicting an SNP reflects the extent of LD with the rs926849 coloured red (r2gt08) through to blue (r2lt02)Genetic recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build36 (NCBI) of the human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn toshow the relative positions and therefore the maps are not to physical scale

Figure 3 Forest plot of rs926849 in PARK2 unadjusted for covariates The contribution of the studies included in the meta-analysis is shown in thisfixed effects model The C allele is considered Heterogeneity I2=0 p(Q)=067 TE treatment effect seTE standard error treatment effect

1146 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

precise role of these regions is yet to be defined Long rangeenhancers for example could operate here and so an influenceon COL11A2 expression cannot be ruled out

Of suggestive significance is SNP rs4802666 (p=376times10minus6adjusted meta-analysis) which lies within the MYH14 genewhich encodes myosin heavy chain 14 non-muscle It isexpressed in cell lines derived from bone (httpwwwproteinatlasorg) and is implicated in autosomal dominanthearing impairment It is of interest in LDD because it lies onchromosome 19 under the linkage peak we have reported intwins for LDD44 and a peak reported by the Framinghamgroup for hand OA45 As there is a known relationship betweenthese two phenotypes this region on chromosome 19 forms ahighly plausible candidate region for OA It is not impossiblethat a muscle-expressed protein plays a role in LDD throughmechanisms similar to those proposed for OA considered bysome to be a multi-tissue syndrome rather than simply adisease of cartilage46

The main limitation of the study is one of obtaining anaccurate phenotype on individuals which is known to be animportant factor in the success of GWA47 There is at presentno agreed gold standard imaging method in the determinationof LDD although it is recognised that MRI offers the most sen-sitive widely available tool Even so MR is relatively expensiveand many of the largest spine cohorts in the world have plainradiographs which offer more limited phenotypic informationThe coding method applied to the imaging is also yet to be for-mally standardised our interest in the individual subtraits ofLDD led us to devise a coding method in which they wereseparated as reported previously7 In order to obtain sufficientsample size a number of cohorts contributed having differentimaging methods but traits were selected to enable comparisonacross the cohorts Thus study groups recoded their imagingwhere necessary to meet uniform requirements for inclusionWe included measures of disc height (coded 0ndash3) and anteriorosteophytes in RS1 RS3 GARP and TUK (also coded 0ndash3) andposterior osteophytes in FHS (coded 0ndash3) These subpheno-types were summed over the five discs and underwent inversenormal transformation to give a normal distribution A furtherlimitation is that four cohorts are population samples whileGARP is derived from OA-affected sibling pairs We includedGARP because it has made a contribution to similar analysesperformed for OA48 and with adjustment for relatedness pro-vides data comparable with other studies While the differingmethods of imaging provide different amounts of informationso the LDD variable has lower mean in those cohorts withradiographs the variance is comparable Where GARP samplesmade a contribution to the meta-analysis (a number of thesignificant SNPs did not include a contribution from GARPtable 2) the minor allele frequency was similar to those ofother groups The TUK group has a disproportionate numberof women for historical reasons The men were retainedhowever as they did not differ significantly from women inthe LDD variable or body mass index (data not shown) Thisstudy lacks a replication group A second sample of similar sizeto the first is considered important to show that the findingsof the first sample are true positives Unfortunately there areto our knowledge no other collections of Northern Europeanshaving spine imaging which together would approach oursample size There is considerable evidence in the literaturethat the genetic predisposition between Northern Europeansand Asians to OA is different49 and given the similaritiesbetween OA and LDD we felt replication should be made inNorthern Europeans We elected to include all the subjects in

a single powerful study rather than split the sample andreduce the chances of finding significant novel loci associatedwith LDD

In conclusion this is the first large-scale GWA study of LDDand we have identified several novel variants in the PARK2gene and in PSMB9 within MHC class 2 We have shown in asmall subset that methylation at one of the PARK2 promotersis associated with MRI determined LDD Both loci meritfurther investigation to shed light on the important clinicalendpoint of low back pain

Author affiliations1Department Twin Research and Genetic Epidemiology Kingrsquos College LondonLondon UK2Acclarogen Ltd St Johnrsquos Innovation Centre Cambridge UK3Department of Internal Medicine Erasmus MC Rotterdam The Netherlands4The Netherlands Genomics Initiative sponsored Netherlands Consortium for HealthyAging (NGI-NCHA) LeidenRotterdam The Netherlands5Department of Molecular Epidemiology Leiden University Medical Center LeidenRotterdam The Netherlands6VA Boston Healthcare System Harvard Medical School Spaulding RehabilitationHospital New England Baptist Hospital Boston USA7Department of Rheumatology University of Sydney and Royal North Shore HospitalSydney Australia8Department of Epidemiology Erasmus MC Rotterdam The Netherlands9Department of General Practice Erasmus MC Rotterdam The Netherlands10Department of Rheumatology and Department of Clinical Epidemiology LeidenUniversity Medical Center Leiden The Netherlands

Acknowledgements FMKW is a Wellcome Trust Intermediate Fellow The authorswould like to acknowledge the contribution of Maria Popham to the collection andanalysis of TwinsUK data The GWA study was funded by the NetherlandsOrganisation of Scientific Research NWO Investments (nr 1750102005011911-03-012) the Research Institute for Diseases in the Elderly (014-93-015 RIDE2)and the Netherlands Genomics Initiative (NGI)the Netherlands Consortium forHealthy Aging (NCHA) project nr 050-060-810 We thank Pascal Arp Mila JhamaiDr Michael Moorhouse Marijn Verkerk and Sander Bervoets for their help in creatingthe GWAS database The Rotterdam Study is funded by Erasmus Medical Center andErasmus University Rotterdam the Netherlands Organisation for the Health Researchand Development (ZonMw) the Research Institute for Diseases in the Elderly (RIDE)the Ministry of Education Culture and Science the Ministry for Health Welfare andSports the European Commission (DG XII) and the Municipality of Rotterdam Theauthors are very grateful to the participants and staff from the Rotterdam Study theparticipating general practitioners and the pharmacists We would like to thank DrTobias A Knoch Luc V de Zeeuw Anis Abuseiris and Rob de Graaf as well as theirinstitutions the Erasmus Computing Grid Rotterdam The Netherlands and especiallythe national German MediGRID and ServicesMediGRID part of the German D-Gridboth funded by the German Bundesministerium fuer Forschung und Technology undergrants 01 AK 803 A-H and 01 IG 07015 G for access to their grid resourcesThe GARP study was supported by the Leiden University Medical Centre and theDutch Arthritis Association Pfizer Inc Groton CT USA supported the inclusion ofthe GARP study The genotypic work was supported by the Netherlands Organisationof Scientific Research (MW 904-61-095 911-03-016 917 66344 and 911-03-012)Leiden University Medical Centre and the Centre of Medical System Biology and theNetherlands Consortium for Healthy Aging both in the framework of the NetherlandsGenomics Initiative (NGI) The research leading to these studies has received fundingfrom the European Unionrsquos Seventh Framework Programme (FP72007-2011) undergrant agreement ndeg 259679 Pradeep Suri was funded by K12HD01097 from theNational Institute of Health (US) and the Rehabilitation Medicine Scientist TrainingProgramme (RMSTP) The Framingham Study is supported by the National HeartLung and Blood Institutersquos Framingham Heart Study contract (No N01-HC-25195) forthe recruitment enrolment and examination of the Offspring and Third Generationcohorts and the imaging by CT scan

Contributors The authors listed fulfilled the criteria for authorship

Funding Fellowship to FMKW who led the project

Competing interests None

Ethics approval All bodies appropriate to the groups contributing data

Provenance and peer review Not commissioned externally peer reviewed

Open Access This is an Open Access article distributed in accordance with theCreative Commons Attribution Non Commercial (CC BY-NC 30) license whichpermits others to distribute remix adapt build upon this work non-commercially andlicense their derivative works on different terms provided the original work is

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1147

Clinical and epidemiological research

properly cited and the use is non-commercial See httpcreativecommonsorglicensesby-nc30

REFERENCES1 Powell MC Wilson M Szypryt P et al Prevalence of lumbar disc degeneration

observed by magnetic resonance in symptomless women Lancet 198621366ndash72 de Schepper EI Damen J van Meurs JB et al The association between lumbar

disc degeneration and low back pain the influence of age gender and individualradiographic features Spine (Phila Pa 1976) 201035531ndash6

3 MacGregor AJ Andrew T Sambrook PN et al Structural psychological andgenetic influences on low back and neck pain a study of adult female twinsArthritis Rheum 200451160ndash7

4 Phillips CJ Economic burden of chronic pain Expert Rev Pharmacoecon OutcomesRes 20066591ndash601

5 Roberts S Evans H Trivedi J et al Histology and pathology of the humanintervertebral disc J Bone Joint Surg Am 200688(Suppl 2)10ndash14

6 Battie MC Levalahti E Videman T et al Heritability of lumbar flexibility and therole of disc degeneration and body weight J Appl Physiol 2008104379ndash85

7 Sambrook PN MacGregor AJ Spector TD Genetic influences on cervical andlumbar disc degeneration a magnetic resonance imaging study in twins ArthritisRheum 199942366ndash72

8 Ryder JJ Garrison K Song F et al Genetic associations in peripheral jointosteoarthritis and spinal degenerative disease a systematic review Ann Rheum Dis200767584ndash91

9 Valdes AM Loughlin J Oene MV et al Sex and ethnic differences in theassociation of ASPN CALM1 COL2A1 COMP and FRZB with genetic susceptibilityto osteoarthritis of the knee Arthritis Rheum 200656137ndash46

10 Richards JB Kavvoura FK Rivadeneira F et al Collaborative meta-analysisassociations of 150 candidate genes with osteoporosis and osteoporotic fractureAnn Intern Med 2009151528ndash37

11 Manolio TA Collins FS Cox NJ et al Finding the missing heritability of complexdiseases Nature 2009461747ndash53

12 Hoffmann U Siebert U Bull-Stewart A et al Evidence for lower variability ofcoronary artery calcium mineral mass measurements by multi-detector computedtomography in a community-based cohortndashconsequences for progression studiesEur J Radiol 200657396ndash402

13 Riyazi N Meulenbelt I Kroon HM et al Evidence for familial aggregation of handhip and spine but not knee osteoarthritis in siblings with multiple joint involvementthe GARP study Ann Rheum Dis 200564438ndash43

14 Lane NE Nevitt MC Genant HK et al Reliability of new indices of radiographicosteoarthritis of the hand and hip and lumbar disc degeneration J Rheumatol1993201911ndash18

15 Hofman A Breteler MM van Duijn CM The Rotterdam Study 2010 objectives anddesign update Eur J Epidemiol 200924553ndash72

16 Bijkerk C Houwing-Duistermaat JJ Valkenburg HA et al Heritabilities of radiologicosteoarthritis in peripheral joints and of disc degeneration of the spine ArthritisRheum 1999421729ndash35

17 Spector TD Williams FM The UK Adult Twin Registry (TwinsUK) Twin Res HumGenet 20069899ndash906

18 Suchindran S Rivedal D Guyton JR et al Genome-wide association study of Lp-PLA(2) activity and mass in the Framingham Heart Study PLoS Genet 20106e1000928

19 Howie BN Donnelly P Marchini J A flexible and accurate genotype imputation method forthe next generation of genome-wide association studies PLoS Genet 20095e1000529

20 Marchini J Howie B Myers S et al A new multipoint method for genome-wideassociation studies by imputation of genotypes Nat Genet 200739906ndash13

21 Uh HW Wijk HJ Houwing-Duistermaat JJ Testing for genetic associationtaking into account phenotypic information of relatives BMC Proc 20093(Suppl 7)S123

22 Richards JB Rivadeneira F Inouye M et al Bone mineral density osteoporosisand osteoporotic fractures a genome-wide association study Lancet20083711505ndash12

23 Stata for Windows Version 10 edn College Station Texas USA Statacorp 200724 Pritchard JK Stephens M Donnelly P Inference of population structure using

multilocus genotype data Genetics 2000155945ndash59

25 Aulchenko YS Struchalin MV van Duijn CM ProbABEL package for genome-wideassociation analysis of imputed data BMC Bioinformatics 201011134

26 Li Y Willer CJ Ding J et al MaCH using sequence and genotype data to estimatehaplotypes and unobserved genotypes Genet Epidemiol 201034816ndash34

27 Bell JT Tsai PC Yang TP et al Epigenome-wide scans identify differentiallymethylated regions for age and age-related phenotypes in a healthy ageingpopulation PLoS Genet 20128e1002629

28 Bell JT Pai AA Pickrell JK et al DNA methylation patterns associate withgenetic and gene expression variation in HapMap cell lines Genome Biol 201112R10

29 Devlin B Roeder K Genomic control for association studies Biometrics199955997ndash1004

30 Liuke M Solovieva S Lamminen A et al Disc degeneration of the lumbar spine inrelation to overweight IntJObes (Lond) 200529903ndash8

31 Pye SR Reid DM Adams JE et al Influence of weight body mass index andlifestyle factors on radiographic features of lumbar disc degeneration Ann RheumDis 200766426ndash7

32 Symmons DP van Hemert AM Vandenbroucke JP et al A longitudinal study ofback pain and radiological changes in the lumbar spines of middle aged women IIRadiographic findings Ann Rheum Dis 199150162ndash6

33 Williams FM Popham M Livshits G et al A response to Videman et allsquochallenging the cumulative injury model positive effects of greater body mass ondisc degenerationrsquo Spine J 201010571ndash2

34 Yoshimura N Dennison E Wilman C et al Epidemiology of chronic discdegeneration and osteoarthritis of the lumbar spine in Britain and Japan acomparative study J Rheumatol 200027429ndash33

35 Battie MC Videman T Gill K et al 1991 Volvo Award in clinical sciences Smokingand lumbar intervertebral disc degeneration an MRI study of identical twins Spine1991161015ndash21

36 Videman T Battie MC The influence of occupation on lumbar degeneration Spine1999241164ndash8

37 Williams FM Sambrook PN Neck and back pain and intervertebral discdegeneration role of occupational factors Best Pract Res Clin Rheumatol20112569ndash79

38 Tanaka A Parkin-mediated selective mitochondrial autophagy mitophagy Parkinpurges damaged organelles from the vital mitochondrial network FEBS Lett20105841386ndash92

39 Kay DM Stevens CF Hamza TH et al A comprehensive analysis of deletionsmultiplications and copy number variations in PARK2 Neurology 2010751189ndash94

40 Gao X Ning Y Cancer and Parkinsons disease the odd couple Drugs Today (Barc)201147215ndash22

41 Kang JD Georgescu HI McIntyre-Larkin L et al Herniated lumbar intervertebraldiscs spontaneously produce matrix metalloproteinases nitric oxide interleukin-6and prostaglandin E2 Spine (Phila Pa 1976) 199621271ndash7

42 Korhonen T Karppinen J Paimela L et al The treatment of disc-herniation-inducedsciatica with infliximab one-year follow-up results of FIRST II a randomizedcontrolled trial Spine (Phila Pa 1976) 2006312759ndash66

43 Videman T Saarela J Kaprio J et al Associations of 25 structural degradativeand inflammatory candidate genes with lumbar disc desiccation bulging and heightnarrowing Arthritis Rheum 200960470ndash81

44 Williams FM Kato BS Livshits G et al Lumbar disc disease shows linkage tochromosome 19 overlapping with a QTL for hand OA Ann Rheum Dis200867117ndash19

45 Demissie S Cupples LA Myers R et al Genome scan for quantity of handosteoarthritis the Framingham Study Arthritis Rheum 200246946ndash52

46 Brandt KD Radin EL Dieppe PA et al Yet more evidence that osteoarthritis is nota cartilage disease Ann Rheum Dis 2006651261ndash4

47 Amos CI Successful design and conduct of genome-wide association studies HumMol Genet 200716R220ndashR225

48 Zhai G van Meurs JB Livshits G et al A genome-wide association studysuggests that a locus within the ataxin 2 binding protein 1 gene is associatedwith hand osteoarthritis the Treat-OA consortium J Med Genet 200946614ndash16

49 Valdes AM Spector TD Genetic epidemiology of hip and knee osteoarthritisNat Rev Rheumatol 2011723ndash32

1148 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

Supplementary Table 1 Genotyping and imputation methods by study

FHS GARP RS1 RS3 TUK

Genotyping

Platform Affymetrix Illumina Illumina Illumina Illumina

Chip 500K or 100K or 10 K Human660W

quad Human Hap550 Human Hap550

Human Hap300 or Human

Hap610Q

Call rate gt097 gt 095 gt 0975 gt 0975 gt090

MAF gt 005 gt 00025 gt 001 gt 001 gt001

HWE p-value gt 10-6

gt 10-6

gt 10-6

gt 10-6

gt 10-4

Imputation

Variants included for imputation 534982 462361 512349 514073 314075

Imputation software MACH vs1 IMPUTE MACH MACH IMPUTE vs 2

Genome build

HapMap Phase II

release 22

Hapmap Phase II

release 21

Hapmap Phase II

release 22

Hapmap Phase II

release 22

Hapmap Phase II

release 36

Total number of SNPs available 2543887 1830498 2543887 2543887 3044064

Legend to Supplementary Table 1

FHS represents Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study RS1 Rotterdam study cohort 1 RS3 Rotterdam study

cohort 3 TUK TwinsUK BMI body mass index MAF Minor Allele Frequency HWE Hardy-Weinberg equilibrium

Supplementary Table 2 Results of the GWA meta-analysis adjusted for age and sex showing those SNPs having plt10-5

MAF Imputation Quality

SNP Chr gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP

Eff

All BETA SE P

rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C -012 0023 950E-08

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 019 0037 367E-07

rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 019 0039 663E-07

rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 019 0039 664E-07

rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A -028 0058 730E-07

rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A -011 0022 742E-07

rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C -025 0051 135E-06

rs11136566 8 CSMD1 4610 039 039 036 039 046 099 098 097 096 9739 A -01 0021 224E-06

rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 016 0035 307E-06

rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 016 0035 307E-06

rs365962 16 NA 4608 044 044 048 047 042 099 099 096 069 9861 C -01 0020 310E-06

rs387953 16 NA 4683 046 045 048 046 043 1 099 1 067 9999 G -009 0020 337E-06

rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A -011 0023 357E-06

rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A -013 0027 376E-06

rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C -012 0026 468E-06

rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0045 477E-06

rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 016 0035 497E-06

rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 016 0035 593E-06

rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 016 0035 593E-06

rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A -01 0023 624E-06

rs7234567 18 NA 4682 026 026 022 022 029 1 1 1 089 9767 A -01 0023 634E-06

rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 016 0035 652E-06

rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C -01 0022 654E-06

rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 01 0022 679E-06

rs10890236 1 NA 4683 009 009 009 009 009 1 1 1 093 100 C 016 0035 707E-06

rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 016 0037 708E-06

rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0022 728E-06

rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 015 0034 821E-06

rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 01 0022 823E-06

rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 015 0035 873E-06

rs7205409 16 RAB40C 4590 042 043 045 043 047 099 099 096 044 9933 C 009 0021 878E-06

rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A -047 0107 970E-06

Legend to Supplementary Table 2

Studies contributing data are denoted RS1 Rotterdam study cohort 1 RS3 Rotterdam study cohort 3 TUK TwinsUK BMI body mass index FHS

Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study

SNP single nucleotide polymorphism Chr chromosome position SNP location in base pairs MAF minor allele frequency

R2 from MACH for RS1 RS3 FHS and GARP Information score from IMPUTE for TUK

Eff All effect allele beta effect size SE standard error of beta p p value

Supplementary Figure 1 Quantile-quantile plot of meta-analysis adjusted GWA results

Legend to Supplementary Figure 1

The plots show GWA meta-analysis quantile-quantile plot of observed against expected results adjusted for age and sex

Supplementary Figure 2 Manhatten plot for meta-analysis of adjusted GWA results

Legend to Supplementary Figure 2

Plot shows combined results for the 5 studies included in the meta-analysis adjusted for age and sex The blue and red horizontal lines mark the levels of

suggestive and likely significance respectively

Supplementary Figure 3 Regional plot of association results and recombination rates for the class II MHC region unadjusted for age and sex

Legend to Supplementary Figure 3

minuslog10 P values (y axis) of the SNPs are shown according to their chromosomal positions (x axis) with lead SNP shown as a purple diamond The colour

intensity of each symbol depicting a SNP reflects the extent of LD with the rs926849 coloured red (r2 gt 08) through to blue (r

2 lt 02) Genetic

recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build 36 (NCBI) of the

human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn to show the relative

positions and therefore the maps are not to physical scale

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 2: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

from their participants and appropriate ethics committeeapproval In all studies a cumulative degeneration score wasconstructed from the sum of scores of degenerative change ateach level (disc space narrowing coded 0ndash3 and osteophyteseither anterior or posterior or both coded 0ndash3) In thosecohorts where only four disc levels were read (FraminghamHeart study (FHS)) a fifth level was imputed by taking themean reading for four discs as a surrogate for the fifth disc andsumming over five discs The data underwent inverse normaltransformation to generate a normally distributed variable

Phenotyping the cohortsThe FHS is a longitudinal cohort of a defined populationin Massachusetts initiated in 1948 (httpwwwframinghamheartstudyorg) It began as a study sample of 5209Framingham men and women between the ages of 30 and 60Subsequently offspring and third generation subjects wereincorporated Every other year after an extensive baselineexamination subjects undergo testing that includes a medicalhistory blood profile echocardiogram and bone eye and othertests The subset of the Framingham subjects covered by thecurrent analysis comprised 366 subjects from the Offspring andGeneration three arms of the study who had undergone CTscanning of the spine and the recruitment conduct and specifi-cations of CT scanning have been reported elsewhere12

Measurement of the lumbar spine CTs for disc height andscoring (0ndash3) for anterior and posterior osteophytes was per-formed by a spine specialist using the mid-sagittal plane atspinal levels L2-L3 L3-L4 L4-L5 and L5-S1 by author PS usingthe atlas of Jarosz et al7 Using sagittal CT reformatting themid-sagittal plane was determined at each spinal level and mea-surements of disc height in millimetres were made The mea-sured values for disc height (mm) were converted to 0ndash3categorical scale for disc height loss Using an imputed valuefor the fifth lumbar disc based on the mean value of the mea-sured four discs values for disc height loss and anterior andposterior osteophytes were summed over five lumbar disclevels

The GARP study comprises white sibling pairs of Dutchorigin affected by OA at multiple sites and is aimed at identify-ing determinants of OA susceptibility and progressionProbands (ages 40ndash70 years) and their siblings had OA at mul-tiple joint sites of the hand or in two or more of the followingjoint sites hand spine (cervical or lumbar) knee or hip asdescribed previously13 Subjects included in this study hadundergone lateral radiographs of the spine (T4-S1) Each inter-vertebral disc level from L12 to L5S1 was reviewed for thepresence and severity of osteophytes (anterior) and disc narrow-ing using the Lane atlas14 where 0=none grade 1=mild grade2=moderate and grade 3=severe The score at each level foranterior osteophytes and disc height loss were summed overthe five lumbar levels

The Rotterdam study is a prospective population-basedfollow-up study of the determinants and prognosis of chronicdiseases in the elderly15 16 All persons living in Ommoord asuburb of Rotterdam who were aged 55 years and over wereinvited to participate A total of 7983 participants were exam-ined For the current analysis two subsets of the data wereconsidered Rotterdam cohort 1 (RS1) consists of 2440 subjectsRotterdam cohort 3 (RS3) consists of 974 subjects Subjects ori-ginating from the Rotterdam study underwent plain radiog-raphy and scoring of LDD as previously described2 In brieflateral lumbar radiographs were scored by a single observer forthe presence of the individual radiographic features of disc

degeneration Each intervertebral disc from L12 to L5S1 wasreviewed for the presence and severity of osteophytes (anterior)and disc narrowing using the Lane atlas as described above14

The scores for the two traits over the five lumbar discs weresummed

The TUK registry was described previously17 The register wasstarted in 1993 and now comprises of approximately 10 000monozygotic and dizygotic adult Caucasian twins aged16ndash85 years from all over the UK plus some parents and sib-lings It now incorporates previous twin registries from theInstitute of Psychiatry and Aberdeen University This is a vol-unteer sample recruited by successive media campaignswithout selecting for particular diseases or traits All twinsreceive a series of detailed disease and environment question-naires The majority of twins have been assessed in detail clin-ically at several time points for several hundred phenotypesrelated to common diseases or intermediate traits The subsetof TUK covered by the current analysis consisted of 744 sub-jects who had participated in the spine MR study (scanned1996ndash2000) using a Siemens MR machine with (MunichGermany) 10-tesla superconducting magnet Serial sagittalimages of the cervical thoraco-lumbar junction and lumbarspine (T9-L5) were obtained7 Images were coded for discheight loss and anterior osteophytes using a 0ndash3 scale in eachcase where 0 is normal and 3 maximal degeneration as per theatlas of Jarosz et al7 All five lumbar discs were scored and thescores summed to give a combined LDD variable

Genotyping and imputationFHS subjects were genotyped using Affymetrix GeneCHipHuman Mapping 500 K array set (Affymetrix Santa Clara CAUSA) andor the 100 K array set andor the 50 K arrayMethods and quality controls have been described previously18

GARP subjects were genotyped using Illumina Human660WQuad BeadChips (HumanHap550v3 HumanHap610 IlluminaSan Diego CA USA) Genotyping was performed at the geno-typing Rotterdam Genotyping Centre Positive strand geno-types were called by clustering in Genome studio andimputation was performed using IMPUTE software andhapmap phase II V2119 20 Strict selection criteria were appliedto the measured genotypes using a high information content(r2 of gt95) and a minor allele frequency gt00025 Associationanalyses were performed using an inhouse developed softwarepackage that allows the analyses of family data using all infor-mation available in the cases and controls by extending theCochran-Armitage trend test21

RSI and RS3 subjects in the Rotterdam Study sets were geno-typed on the HumanHap550v3 (RS1) or HumanHap610 (RS3)Genotyping BeadCHip (Illumina San Diego California USA)The following sample quality control criteria were appliedsample call rate gt975 gender mismatch with typed X-linkedmarkers evidence for DNA contamination in the samples usingthe mean of the autosomal heterozygosity gt033 exclusion ofduplicates or first-degree relatives identified using Identity byState probabilities and exclusion of outliers (four SD awayfrom the population mean using multidimensional scaling ana-lysis with four principal components) Filtering criteria forimputation are summarised in supplementary table S1

TUK subjects were genotyped using a combination ofIllumina arrays (Human Hap300 and the Human Hap610Q)Genotyping was performed by the Wellcome Trust SangerInstitute using the Infinium assay (Illumina) across threegenome-wide single nucleotide polymorphism (SNP) sets asdescribed previously22 Genotyping results had been sent to

1142 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

KCL for collation and analysis using the statistical packageSTATA (StataCorp)23 Strict quality control was applied314 075 SNPs were retained for analysis (987) 733 wereexcluded because their call rates were le90 and 725 SNPs hadminor allele frequency lt001 In TUK significant populationsubstructure was excluded using the STRUCTURE program24

GWA analysisAll analyses were performed on inverse normal transformedsummary LDD score as described above Each study performedGWA analysis for LDD scores using either MACH2QTL (httpwwwsphumicheducsgabecasisMACHindexhtml) (RS1and RS3) or SNPTEST (httpwwwstatsoxacuk~marchinisoftwaregwassnptesthtml) (GARP) which use genotypedosage value as continuous additive predictors of LDD score ina linear regression framework or ProbABEL25 using an additivegenetic model while accounting for relatedness between themembers of a family Analysis of imputed genotype dataaccounted for uncertainty in each genotype prediction by usingeither the dosage information from MACH26 or the genotypeprobabilities from IMPUTE19

Meta-analysis of the five study groupsGenotypes for 25ndash3 million autosomal SNPs were imputed sep-arately to increase coverage using HapMap V2 (httpwwwhapmaporg) as the reference panel In GARP and TUK imput-ation was performed with IMPUTE V219 and in the otherstudies with MACH26 The common reference panel led to thereporting of results for the positive strand for all cohorts In add-ition allele pairs were compared between cohorts and no detect-able strand-flips were found the minor allele frequency was alsocompared between datasets The distributions of β values of thecohorts were found to be similar and therefore suitable formeta-analysis All directly genotyped or imputed autosomalSNPs having information from more than one study group(n=2 552 511) were included in the meta-analysis Associationresults were combined using inverse variance weighted fixedeffects meta-analysis using PLINK V106 (httppngumghharvardedupurcellplink) Two meta-analyses were run thefirst was unadjusted the second was adjusted for age and sex asboth are known risk factors for LDD and each risk factor wascorrelated with LDD in each study group Heterogeneity of esti-mated effect was expressed using Q (weighted sum of squares)and I2 (ratio of true heterogeneity to total observed variation)SNPs were excluded from the meta-analysis if the cohort-specificimputation quality as assessed by r2 (MACH) or InformationScore (IMPUTE) metric was lt040 On this basis one markerwas excluded from the unadjusted association and one from theadjusted association

DNA methylation data and analysisWhole blood DNA methylation levels were obtained for 38individuals in the TUK cohort using the IlluminaHumanMethylation27 DNA Analysis BeadChip assay as previ-ously described27 The sample included four monozygotic twinpairs eight dizygotic twin pairs and 14 unrelated individualsAt each CpG site within an individual the methylation levelwas presented as β which represents the ratio of intensitysignal obtained from the methylated beads in the array overthe sum of methylated and unmethylated bead signalsFollowing quality control checks we obtained DNA methyla-tion at three CpG sites in the promoter region of the PARK2gene within 2kb of the transcription start site The threeprobes (cg15832436 cg21926612 and cg24816866) mappeduniquely to the human genome (hg18) within two mismatches(see Bell et al28) We fitted linear mixed effects models to assessassociation between DNA methylation levels at the three CpGsites in the PARK2 promoter and LDD We regressed the rawmethylation levels on fixed-effect terms including methylationchip and LDD and random-effect terms denoting family struc-ture and zygosity and compared the association of differen-tially methylated regions with a null model which excludedLDD from the fixed-effects terms We also repeated the associ-ation analyses by normalising the methylation values at eachCpG site to N(0 1)

RESULTSThe study samples for the meta-analysis included 4683 indivi-duals of European ancestries Table 1 shows sample size demo-graphic characteristics LDD and lumbar spine imaging methodfor each independent cohort The majority of participants werefemale subjects (670) and the samples had a mean age of577 years Across the cohorts the mean level of LDD variedfrom 0011 to 346 reflecting differences in imaging methodsHowever the variance of the LDD variables was broadlysimilar (range 0958ndash114) as were the distributions of the esti-mated genetic effect sizes (β) The genotyping and imputationmethods are shown in online supplementary table S1

Quantilendashquantile plots for the unadjusted LDD GWA meta-analysis are presented in figure 1 (see online supplementaryfigure S1 adjusted) Test statistic inflation post meta-analysisas measured by the genomic control statistic29 was low(λGC unadjusted=102 λGC adjusted=103) Results of theunadjusted and adjusted association analyses were broadlysimilar with the p values of the adjusted analysis somewhatattenuated A Manhattan plot for the unadjusted analysis isshown in figure 2 with numeric results in table 2 (unadjusted)and online supplementary table S2 (adjusted) for SNPs havingplt10minus5

Table 1 Characteristics of the study samples

FHS GARP RS1 RS3 TUK

N 330 192 2440 974 744Age (years) 543 (110) 603 (71) 657 (67) 547 (34) 536 (83)Women () 422 797 57 59 968BMI (kgm2) 281 (51) 2644 (48) 263 (34) 2712 (46) 249 (44)Lumbar spine imaging CT Radiograph Radiograph Radiograph MRILDD variable 249 (097) 002 (0958) 0006 (0978) 0011 (0965) 346 (114)

Values are mean (SD) unless specified otherwiseBMI body mass index FHS Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study LDD Lumbar disc degeneration RS1 Rotterdam study cohort 1RS3 Rotterdam study cohort 3 TUK TwinsUK

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1143

Clinical and epidemiological research

Four markers achieved genome-wide significance in theunadjusted GWAS of which three were on chromosome 6(rs926849 rs2187689 rs7767277) and an intergenic marker onchromosome 3 (rs17034687) The results of the meta-analysisadjusted for age and gender were broadly similar the strongestsignal was for SNP rs926849 This SNP lies on an intronicregion of the Parkinson protein 2 E3 ubiquitin protein ligase(PARK2) gene on chromosome 6 A Forest plot of the groupsrsquoresults and their meta-analysis is shown in figure 3 Data were

available from four study groups and the range of estimatedminor allele frequencies was 023ndash032 Imputation quality washigh for all four studies contributing this SNP (gt090 table 2)and the estimated between-study heterogeneity was acceptable(I2=0 p(Q)=067) The results of a four-study meta-analysiswhich excluded GARP show the marker to have p=95eminus8

(unadjusted) The minor or C allele of rs926849 was associatedwith a lower level of LDD implying that the minor allele is pro-tective Figure 4 shows a regional plot of both genotyped andimputed SNPs within 200 Kb of the PARK2 gene along withrecombination rates

Two of the other strongly associated SNPs are in perfectlinkage disequilibrium (LD) rs2187689 and rs7767277 onchromosome 6 Data were available for four studies and therange of estimated allele frequency was 006ndash009 Imputationquality was high for all four studies (gt090) Both SNPs are instrong LD (r2=076) with an intronic marker on the prote-asome subunit β type 9 large multifunctional peptidase 2 gene(PSMB9) that is located in the class II region of the majorhistocompatibility complex (MHC) Both genotyped andimputed SNPs within 400 Kb of rs2187689 along with recom-bination rates are shown in a regional plot in onlinesupplementary figure S3 None of these top SNPs is in LD withknown functional SNPs in either PARK2 or PSMB9

We tested for an association between LDD and DNA methy-lation variants at three CpG sites in the PARK2 promoterA significant association between DNA methylation at CpGsite cg15832436 and LDD (β=874times10minus4 SE=249times10minus4p=0006) was observed The pattern of hypermethylation withincreasing LDD levels was reflected at the remaining promoterCpG sites however these did not reach nominal significance(cg21926612 β=0003 p=014 cg24816866 β=676times10minus4p=039) We repeated the analyses using normalised methyla-tion levels and observed that the association betweencg15832436 and LDD remained nominally significant

DISCUSSIONGWA offers an unbiased scan of common genetic variants(minor allele frequency gt5) and thus may deliver novel var-iants in genes not hitherto suspected of playing a role in discdegeneration This work is among the first to report on agenome-wide meta-analysis being conducted for LDD LDD isan age-related process which occurs in all people to someextent and may be detected as early as the teenage years1 LDDis known to have genetic determinants7 16 and its expression isalso influenced by gender (women develop LDD later) bodymass index30ndash34 and smoking35 Occupational factors also playa small role in LDD36 37 LDD as determined by MRI has beenimplicated in the development of episodes of severe and disab-ling low back pain3 We undertook this large meta-analysis inorder to identify novel genetic variants associated with LDDand to shed light on the underlying pathology of discdegeneration

GWA data obtained using differing chip technology may bereadily compared using imputation with HapMap In total2 552 511 overlapping markers were available in each cohortWe identified four markers having significant association withthe LDD phenotype (plt5times10minus8) There was similarity in theresults obtained with and without adjustment for the covari-ates age and sex A total of 26 markers had plt10minus5 in bothmeta-analyses As expected results of the adjusted analyseshad slightly attenuated p values (see online supplementarytable S2) which likely reflect the confounding effect of age Inboth analyses there were multiple associations to the Human

Figure 1 Quantilendashquantile plot of observed versus expecteddistribution of p values for the genome-wide association (GWA)meta-analysis The plots show GWA meta-analysis quantilendashquantileplot of observed against expected results unadjusted for covariates

Figure 2 Manhattan plot for meta-analysis of unadjustedgenome-wide association results Plot shows combined results for thefive studies included in the meta-analysis unadjusted results The blueand red horizontal lines mark the levels of suggestive and likelysignificance respectively

1144 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

Leukocyte Antigen (HLA) region and to markers in PARK2(Parkinson protein 2 E3 ubiquitin protein ligase) Among themost significant findings (table 2) is SNP rs926849 that lies at6q252ndash27 within an intron in the PARK2 gene a large gene of13 Mb comprising 12 exons The SNP encodes a change of basefrom T to C and is reported to have a minor allele frequency of023ndash034 in dbSNP which is keeping with the findings in ourstudy groups (table 2 figures 3 and 4) Although this SNP hasnot been directly genotyped by any study group estimates

suggest imputation to be accurate for rs926849 (range 95ndash

99 table 2) PARK2 encodes a protein called parkin which isa component of a multiprotein E3 ubiquitin ligase complexthat mediates the targeting of unwanted proteins for proteaso-mal degradation This complex also controls the level of pro-teins involved in cell activities such as cell division and growthit may be a tumour suppressor protein Parkin is also involvedin mitophagy it translocates from the cytosol and promotesthe degradation of uncoupled mitochondria38 Parkin is widely

Table 2 Results of the genome-wide association meta-analysis (unadjusted showing those SNPs having plt10minus5)

MAF Imputation quality

SNP Chr Gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP Eff all β SE p Value

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 023 0038 182E-09rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 023 0041 272E-08rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 023 0041 281E-08rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C minus013 0024 325E-08rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 02 0037 558E-08rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 02 0037 559E-08rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 019 0036 936E-08rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 019 0036 939E-08rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 019 0036 104E-07rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 019 0037 146E-07rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 019 0037 147E-07rs10214886 6 NA 4479 01 009 007 011 NA 098 1 092 097 NA A 019 0036 232E-07rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 019 0037 322E-07rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A minus012 0024 361E-07rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A minus012 0023 368E-07rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 019 0038 458E-07rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C minus026 0054 947E-07rs7896691 10 PFKP 4652 01 01 009 013 007 099 1 096 067 9761 C 017 0035 202E-06rs6602024 10 PFKP 4673 01 01 009 013 007 099 1 098 067 9762 A 017 0035 220E-06rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C minus011 0023 356E-06rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A minus053 0114 359E-06rs17132175 10 PFKP 4659 01 01 009 013 007 098 099 097 063 9713 C 016 0036 373E-06rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 011 0023 375E-06rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C minus013 0028 399E-06rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A minus011 0025 432E-06rs2484990 10 NA 4429 001 001 0 002 NA 056 082 047 093 NA C 068 0148 439E-06rs1250307 10 NA 4415 001 001 0 002 NA 056 082 045 094 NA A 068 0148 439E-06rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 011 0023 443E-06rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A minus013 0029 455E-06rs2484992 10 NA 4440 001 001 0 002 NA 056 082 048 093 NA C 068 0149 481E-06rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A minus028 0061 517E-06rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0023 543E-06rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0047 564E-06rs11918654 3 ARL8B 4310 027 027 029 027 NA 097 098 09 099 NA C minus011 0024 710E-06rs4881085 10 PFKP 4490 01 01 01 013 NA 1 1 099 067 NA A 016 0036 744E-06rs2657195 8 NA 4316 022 021 018 021 NA 085 085 087 074 NA A minus013 0028 754E-06rs11754641 6 NA 4478 003 003 002 004 NA 093 095 091 1 NA C 029 0064 784E-06rs12805875 11 NA 4674 042 041 042 043 041 1 1 1 1 9812 A 009 0021 851E-06rs980238 8 NA 4491 03 029 031 029 NA 1 1 1 1 NA A minus01 0023 911E-06rs7103004 11 NA 4667 042 041 042 043 041 1 1 099 1 9709 C 009 0021 920E-06rs4554859 11 NA 4674 042 041 042 043 041 1 1 1 1 9812 G 009 0021 924E-06rs7118412 11 NA 4669 042 041 042 043 041 1 1 099 1 9709 A 009 0021 937E-06rs2017567 16 NA 4683 042 043 042 042 047 099 1 1 045 100 C 01 0021 938E-06rs710924 16 PIGQ 4683 042 043 042 042 047 1 1 1 045 9964 C 009 0021 993E-06

r2 from MACH for RS1 RS3 FHS and GARP information score from IMPUTE for TUKStudies contributing data are denotedSNP single nucleiotide polymorphism Chr chromosome N number of subjects studied MAF minor allele frequency RS1 Rotterdam study cohort 1 RS3 Rotterdam studycohort 3 TUK TwinsUK FHS Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study Eff all effect allele β effect size SE SE of β

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1145

Clinical and epidemiological research

expressed in solid organs as well as skeletal muscle (httpwwwproteinatlasorg) Mutations within PARK2 are associated withdiverse conditions including autosomal recessive juvenileParkinsonrsquos disease Alzheimer rsquos disease diabetes mellitus andseveral solid tumours (reviewed in39) Parkin may account for theinverse relation between Parkinsonrsquos disease and cancer inci-dence40 Our findings of hypermethylation with increasing LDDscore suggest that PARK2 expression is reduced with increasingdisc degeneration but functional studies of intervertebral disc andother spine tissues are needed

Three further markers in the unadjusted meta-analysis hadplt5times10minus8 Marker rs17034687 is an intergenic marker onchromosome 3 Based on One Thousand Genomes (1KG)CEUdata it is not in LD (r2gt 03) with any known gene-basedmarkers Markers rs2187689 (supplementary figure S2) andrs7767277 are HLA region markers neither of which is includedin the 1KG pilot data Using data from HapMap V3 (release 2)rs2187689 and rs776277 are in perfect LD with each other andin LD (r2= 076) with an intronic marker in PSMB9 (prote-asome (prosome macropain) subunit β type 9 large

multifunctional peptidase 2) Proteasomes are distributedthroughout eukaroytic cells at high concentration and cleavepeptides in an ATPubiquitin-dependent process in a non-lysosomal pathway The gene is located in the class II region ofthe MHC Expression of the gene is induced by interferon γand this gene product replaces catalytic subunit 1 (proteasomeβ 6 subunit) in the immunoproteasome

While lumbar degeneration is not considered an inflamma-tory process and has not been reported to be auto-immune inaetiology there is evidence of pro-inflammatory cytokine acti-vation in herniated lumbar discs41 and anti-TNF has been usedsuccessfully to treat disc herniation-induced sciatica42 Of notethe COL11A2 gene lies 169 KB upstream from rs2187689 AnSNP (rs2076311) within this candidate gene has been shown tobe associated with MR determined disc signal intensity in acandidate gene study of Finnish male twins43 SNP rs2076311is not however in LD with our top hit rs2187689 (r2=0017)and so it seems unlikely that this collagen-encoding geneaccounts for our observed association Many published GWAstudies have identified SNPs in intergenic regions and the

Figure 4 Regional plot of association results and recombination rates for the PARK2 gene unadjusted results minuslog10 p values (y-axis) of the singlenucleotide polymorphisms (SNPs) are shown according to their chromosomal positions (x-axis) with lead SNP shown as a purple diamond Thecolour intensity of each symbol depicting an SNP reflects the extent of LD with the rs926849 coloured red (r2gt08) through to blue (r2lt02)Genetic recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build36 (NCBI) of the human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn toshow the relative positions and therefore the maps are not to physical scale

Figure 3 Forest plot of rs926849 in PARK2 unadjusted for covariates The contribution of the studies included in the meta-analysis is shown in thisfixed effects model The C allele is considered Heterogeneity I2=0 p(Q)=067 TE treatment effect seTE standard error treatment effect

1146 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

precise role of these regions is yet to be defined Long rangeenhancers for example could operate here and so an influenceon COL11A2 expression cannot be ruled out

Of suggestive significance is SNP rs4802666 (p=376times10minus6adjusted meta-analysis) which lies within the MYH14 genewhich encodes myosin heavy chain 14 non-muscle It isexpressed in cell lines derived from bone (httpwwwproteinatlasorg) and is implicated in autosomal dominanthearing impairment It is of interest in LDD because it lies onchromosome 19 under the linkage peak we have reported intwins for LDD44 and a peak reported by the Framinghamgroup for hand OA45 As there is a known relationship betweenthese two phenotypes this region on chromosome 19 forms ahighly plausible candidate region for OA It is not impossiblethat a muscle-expressed protein plays a role in LDD throughmechanisms similar to those proposed for OA considered bysome to be a multi-tissue syndrome rather than simply adisease of cartilage46

The main limitation of the study is one of obtaining anaccurate phenotype on individuals which is known to be animportant factor in the success of GWA47 There is at presentno agreed gold standard imaging method in the determinationof LDD although it is recognised that MRI offers the most sen-sitive widely available tool Even so MR is relatively expensiveand many of the largest spine cohorts in the world have plainradiographs which offer more limited phenotypic informationThe coding method applied to the imaging is also yet to be for-mally standardised our interest in the individual subtraits ofLDD led us to devise a coding method in which they wereseparated as reported previously7 In order to obtain sufficientsample size a number of cohorts contributed having differentimaging methods but traits were selected to enable comparisonacross the cohorts Thus study groups recoded their imagingwhere necessary to meet uniform requirements for inclusionWe included measures of disc height (coded 0ndash3) and anteriorosteophytes in RS1 RS3 GARP and TUK (also coded 0ndash3) andposterior osteophytes in FHS (coded 0ndash3) These subpheno-types were summed over the five discs and underwent inversenormal transformation to give a normal distribution A furtherlimitation is that four cohorts are population samples whileGARP is derived from OA-affected sibling pairs We includedGARP because it has made a contribution to similar analysesperformed for OA48 and with adjustment for relatedness pro-vides data comparable with other studies While the differingmethods of imaging provide different amounts of informationso the LDD variable has lower mean in those cohorts withradiographs the variance is comparable Where GARP samplesmade a contribution to the meta-analysis (a number of thesignificant SNPs did not include a contribution from GARPtable 2) the minor allele frequency was similar to those ofother groups The TUK group has a disproportionate numberof women for historical reasons The men were retainedhowever as they did not differ significantly from women inthe LDD variable or body mass index (data not shown) Thisstudy lacks a replication group A second sample of similar sizeto the first is considered important to show that the findingsof the first sample are true positives Unfortunately there areto our knowledge no other collections of Northern Europeanshaving spine imaging which together would approach oursample size There is considerable evidence in the literaturethat the genetic predisposition between Northern Europeansand Asians to OA is different49 and given the similaritiesbetween OA and LDD we felt replication should be made inNorthern Europeans We elected to include all the subjects in

a single powerful study rather than split the sample andreduce the chances of finding significant novel loci associatedwith LDD

In conclusion this is the first large-scale GWA study of LDDand we have identified several novel variants in the PARK2gene and in PSMB9 within MHC class 2 We have shown in asmall subset that methylation at one of the PARK2 promotersis associated with MRI determined LDD Both loci meritfurther investigation to shed light on the important clinicalendpoint of low back pain

Author affiliations1Department Twin Research and Genetic Epidemiology Kingrsquos College LondonLondon UK2Acclarogen Ltd St Johnrsquos Innovation Centre Cambridge UK3Department of Internal Medicine Erasmus MC Rotterdam The Netherlands4The Netherlands Genomics Initiative sponsored Netherlands Consortium for HealthyAging (NGI-NCHA) LeidenRotterdam The Netherlands5Department of Molecular Epidemiology Leiden University Medical Center LeidenRotterdam The Netherlands6VA Boston Healthcare System Harvard Medical School Spaulding RehabilitationHospital New England Baptist Hospital Boston USA7Department of Rheumatology University of Sydney and Royal North Shore HospitalSydney Australia8Department of Epidemiology Erasmus MC Rotterdam The Netherlands9Department of General Practice Erasmus MC Rotterdam The Netherlands10Department of Rheumatology and Department of Clinical Epidemiology LeidenUniversity Medical Center Leiden The Netherlands

Acknowledgements FMKW is a Wellcome Trust Intermediate Fellow The authorswould like to acknowledge the contribution of Maria Popham to the collection andanalysis of TwinsUK data The GWA study was funded by the NetherlandsOrganisation of Scientific Research NWO Investments (nr 1750102005011911-03-012) the Research Institute for Diseases in the Elderly (014-93-015 RIDE2)and the Netherlands Genomics Initiative (NGI)the Netherlands Consortium forHealthy Aging (NCHA) project nr 050-060-810 We thank Pascal Arp Mila JhamaiDr Michael Moorhouse Marijn Verkerk and Sander Bervoets for their help in creatingthe GWAS database The Rotterdam Study is funded by Erasmus Medical Center andErasmus University Rotterdam the Netherlands Organisation for the Health Researchand Development (ZonMw) the Research Institute for Diseases in the Elderly (RIDE)the Ministry of Education Culture and Science the Ministry for Health Welfare andSports the European Commission (DG XII) and the Municipality of Rotterdam Theauthors are very grateful to the participants and staff from the Rotterdam Study theparticipating general practitioners and the pharmacists We would like to thank DrTobias A Knoch Luc V de Zeeuw Anis Abuseiris and Rob de Graaf as well as theirinstitutions the Erasmus Computing Grid Rotterdam The Netherlands and especiallythe national German MediGRID and ServicesMediGRID part of the German D-Gridboth funded by the German Bundesministerium fuer Forschung und Technology undergrants 01 AK 803 A-H and 01 IG 07015 G for access to their grid resourcesThe GARP study was supported by the Leiden University Medical Centre and theDutch Arthritis Association Pfizer Inc Groton CT USA supported the inclusion ofthe GARP study The genotypic work was supported by the Netherlands Organisationof Scientific Research (MW 904-61-095 911-03-016 917 66344 and 911-03-012)Leiden University Medical Centre and the Centre of Medical System Biology and theNetherlands Consortium for Healthy Aging both in the framework of the NetherlandsGenomics Initiative (NGI) The research leading to these studies has received fundingfrom the European Unionrsquos Seventh Framework Programme (FP72007-2011) undergrant agreement ndeg 259679 Pradeep Suri was funded by K12HD01097 from theNational Institute of Health (US) and the Rehabilitation Medicine Scientist TrainingProgramme (RMSTP) The Framingham Study is supported by the National HeartLung and Blood Institutersquos Framingham Heart Study contract (No N01-HC-25195) forthe recruitment enrolment and examination of the Offspring and Third Generationcohorts and the imaging by CT scan

Contributors The authors listed fulfilled the criteria for authorship

Funding Fellowship to FMKW who led the project

Competing interests None

Ethics approval All bodies appropriate to the groups contributing data

Provenance and peer review Not commissioned externally peer reviewed

Open Access This is an Open Access article distributed in accordance with theCreative Commons Attribution Non Commercial (CC BY-NC 30) license whichpermits others to distribute remix adapt build upon this work non-commercially andlicense their derivative works on different terms provided the original work is

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1147

Clinical and epidemiological research

properly cited and the use is non-commercial See httpcreativecommonsorglicensesby-nc30

REFERENCES1 Powell MC Wilson M Szypryt P et al Prevalence of lumbar disc degeneration

observed by magnetic resonance in symptomless women Lancet 198621366ndash72 de Schepper EI Damen J van Meurs JB et al The association between lumbar

disc degeneration and low back pain the influence of age gender and individualradiographic features Spine (Phila Pa 1976) 201035531ndash6

3 MacGregor AJ Andrew T Sambrook PN et al Structural psychological andgenetic influences on low back and neck pain a study of adult female twinsArthritis Rheum 200451160ndash7

4 Phillips CJ Economic burden of chronic pain Expert Rev Pharmacoecon OutcomesRes 20066591ndash601

5 Roberts S Evans H Trivedi J et al Histology and pathology of the humanintervertebral disc J Bone Joint Surg Am 200688(Suppl 2)10ndash14

6 Battie MC Levalahti E Videman T et al Heritability of lumbar flexibility and therole of disc degeneration and body weight J Appl Physiol 2008104379ndash85

7 Sambrook PN MacGregor AJ Spector TD Genetic influences on cervical andlumbar disc degeneration a magnetic resonance imaging study in twins ArthritisRheum 199942366ndash72

8 Ryder JJ Garrison K Song F et al Genetic associations in peripheral jointosteoarthritis and spinal degenerative disease a systematic review Ann Rheum Dis200767584ndash91

9 Valdes AM Loughlin J Oene MV et al Sex and ethnic differences in theassociation of ASPN CALM1 COL2A1 COMP and FRZB with genetic susceptibilityto osteoarthritis of the knee Arthritis Rheum 200656137ndash46

10 Richards JB Kavvoura FK Rivadeneira F et al Collaborative meta-analysisassociations of 150 candidate genes with osteoporosis and osteoporotic fractureAnn Intern Med 2009151528ndash37

11 Manolio TA Collins FS Cox NJ et al Finding the missing heritability of complexdiseases Nature 2009461747ndash53

12 Hoffmann U Siebert U Bull-Stewart A et al Evidence for lower variability ofcoronary artery calcium mineral mass measurements by multi-detector computedtomography in a community-based cohortndashconsequences for progression studiesEur J Radiol 200657396ndash402

13 Riyazi N Meulenbelt I Kroon HM et al Evidence for familial aggregation of handhip and spine but not knee osteoarthritis in siblings with multiple joint involvementthe GARP study Ann Rheum Dis 200564438ndash43

14 Lane NE Nevitt MC Genant HK et al Reliability of new indices of radiographicosteoarthritis of the hand and hip and lumbar disc degeneration J Rheumatol1993201911ndash18

15 Hofman A Breteler MM van Duijn CM The Rotterdam Study 2010 objectives anddesign update Eur J Epidemiol 200924553ndash72

16 Bijkerk C Houwing-Duistermaat JJ Valkenburg HA et al Heritabilities of radiologicosteoarthritis in peripheral joints and of disc degeneration of the spine ArthritisRheum 1999421729ndash35

17 Spector TD Williams FM The UK Adult Twin Registry (TwinsUK) Twin Res HumGenet 20069899ndash906

18 Suchindran S Rivedal D Guyton JR et al Genome-wide association study of Lp-PLA(2) activity and mass in the Framingham Heart Study PLoS Genet 20106e1000928

19 Howie BN Donnelly P Marchini J A flexible and accurate genotype imputation method forthe next generation of genome-wide association studies PLoS Genet 20095e1000529

20 Marchini J Howie B Myers S et al A new multipoint method for genome-wideassociation studies by imputation of genotypes Nat Genet 200739906ndash13

21 Uh HW Wijk HJ Houwing-Duistermaat JJ Testing for genetic associationtaking into account phenotypic information of relatives BMC Proc 20093(Suppl 7)S123

22 Richards JB Rivadeneira F Inouye M et al Bone mineral density osteoporosisand osteoporotic fractures a genome-wide association study Lancet20083711505ndash12

23 Stata for Windows Version 10 edn College Station Texas USA Statacorp 200724 Pritchard JK Stephens M Donnelly P Inference of population structure using

multilocus genotype data Genetics 2000155945ndash59

25 Aulchenko YS Struchalin MV van Duijn CM ProbABEL package for genome-wideassociation analysis of imputed data BMC Bioinformatics 201011134

26 Li Y Willer CJ Ding J et al MaCH using sequence and genotype data to estimatehaplotypes and unobserved genotypes Genet Epidemiol 201034816ndash34

27 Bell JT Tsai PC Yang TP et al Epigenome-wide scans identify differentiallymethylated regions for age and age-related phenotypes in a healthy ageingpopulation PLoS Genet 20128e1002629

28 Bell JT Pai AA Pickrell JK et al DNA methylation patterns associate withgenetic and gene expression variation in HapMap cell lines Genome Biol 201112R10

29 Devlin B Roeder K Genomic control for association studies Biometrics199955997ndash1004

30 Liuke M Solovieva S Lamminen A et al Disc degeneration of the lumbar spine inrelation to overweight IntJObes (Lond) 200529903ndash8

31 Pye SR Reid DM Adams JE et al Influence of weight body mass index andlifestyle factors on radiographic features of lumbar disc degeneration Ann RheumDis 200766426ndash7

32 Symmons DP van Hemert AM Vandenbroucke JP et al A longitudinal study ofback pain and radiological changes in the lumbar spines of middle aged women IIRadiographic findings Ann Rheum Dis 199150162ndash6

33 Williams FM Popham M Livshits G et al A response to Videman et allsquochallenging the cumulative injury model positive effects of greater body mass ondisc degenerationrsquo Spine J 201010571ndash2

34 Yoshimura N Dennison E Wilman C et al Epidemiology of chronic discdegeneration and osteoarthritis of the lumbar spine in Britain and Japan acomparative study J Rheumatol 200027429ndash33

35 Battie MC Videman T Gill K et al 1991 Volvo Award in clinical sciences Smokingand lumbar intervertebral disc degeneration an MRI study of identical twins Spine1991161015ndash21

36 Videman T Battie MC The influence of occupation on lumbar degeneration Spine1999241164ndash8

37 Williams FM Sambrook PN Neck and back pain and intervertebral discdegeneration role of occupational factors Best Pract Res Clin Rheumatol20112569ndash79

38 Tanaka A Parkin-mediated selective mitochondrial autophagy mitophagy Parkinpurges damaged organelles from the vital mitochondrial network FEBS Lett20105841386ndash92

39 Kay DM Stevens CF Hamza TH et al A comprehensive analysis of deletionsmultiplications and copy number variations in PARK2 Neurology 2010751189ndash94

40 Gao X Ning Y Cancer and Parkinsons disease the odd couple Drugs Today (Barc)201147215ndash22

41 Kang JD Georgescu HI McIntyre-Larkin L et al Herniated lumbar intervertebraldiscs spontaneously produce matrix metalloproteinases nitric oxide interleukin-6and prostaglandin E2 Spine (Phila Pa 1976) 199621271ndash7

42 Korhonen T Karppinen J Paimela L et al The treatment of disc-herniation-inducedsciatica with infliximab one-year follow-up results of FIRST II a randomizedcontrolled trial Spine (Phila Pa 1976) 2006312759ndash66

43 Videman T Saarela J Kaprio J et al Associations of 25 structural degradativeand inflammatory candidate genes with lumbar disc desiccation bulging and heightnarrowing Arthritis Rheum 200960470ndash81

44 Williams FM Kato BS Livshits G et al Lumbar disc disease shows linkage tochromosome 19 overlapping with a QTL for hand OA Ann Rheum Dis200867117ndash19

45 Demissie S Cupples LA Myers R et al Genome scan for quantity of handosteoarthritis the Framingham Study Arthritis Rheum 200246946ndash52

46 Brandt KD Radin EL Dieppe PA et al Yet more evidence that osteoarthritis is nota cartilage disease Ann Rheum Dis 2006651261ndash4

47 Amos CI Successful design and conduct of genome-wide association studies HumMol Genet 200716R220ndashR225

48 Zhai G van Meurs JB Livshits G et al A genome-wide association studysuggests that a locus within the ataxin 2 binding protein 1 gene is associatedwith hand osteoarthritis the Treat-OA consortium J Med Genet 200946614ndash16

49 Valdes AM Spector TD Genetic epidemiology of hip and knee osteoarthritisNat Rev Rheumatol 2011723ndash32

1148 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

Supplementary Table 1 Genotyping and imputation methods by study

FHS GARP RS1 RS3 TUK

Genotyping

Platform Affymetrix Illumina Illumina Illumina Illumina

Chip 500K or 100K or 10 K Human660W

quad Human Hap550 Human Hap550

Human Hap300 or Human

Hap610Q

Call rate gt097 gt 095 gt 0975 gt 0975 gt090

MAF gt 005 gt 00025 gt 001 gt 001 gt001

HWE p-value gt 10-6

gt 10-6

gt 10-6

gt 10-6

gt 10-4

Imputation

Variants included for imputation 534982 462361 512349 514073 314075

Imputation software MACH vs1 IMPUTE MACH MACH IMPUTE vs 2

Genome build

HapMap Phase II

release 22

Hapmap Phase II

release 21

Hapmap Phase II

release 22

Hapmap Phase II

release 22

Hapmap Phase II

release 36

Total number of SNPs available 2543887 1830498 2543887 2543887 3044064

Legend to Supplementary Table 1

FHS represents Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study RS1 Rotterdam study cohort 1 RS3 Rotterdam study

cohort 3 TUK TwinsUK BMI body mass index MAF Minor Allele Frequency HWE Hardy-Weinberg equilibrium

Supplementary Table 2 Results of the GWA meta-analysis adjusted for age and sex showing those SNPs having plt10-5

MAF Imputation Quality

SNP Chr gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP

Eff

All BETA SE P

rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C -012 0023 950E-08

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 019 0037 367E-07

rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 019 0039 663E-07

rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 019 0039 664E-07

rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A -028 0058 730E-07

rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A -011 0022 742E-07

rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C -025 0051 135E-06

rs11136566 8 CSMD1 4610 039 039 036 039 046 099 098 097 096 9739 A -01 0021 224E-06

rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 016 0035 307E-06

rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 016 0035 307E-06

rs365962 16 NA 4608 044 044 048 047 042 099 099 096 069 9861 C -01 0020 310E-06

rs387953 16 NA 4683 046 045 048 046 043 1 099 1 067 9999 G -009 0020 337E-06

rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A -011 0023 357E-06

rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A -013 0027 376E-06

rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C -012 0026 468E-06

rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0045 477E-06

rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 016 0035 497E-06

rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 016 0035 593E-06

rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 016 0035 593E-06

rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A -01 0023 624E-06

rs7234567 18 NA 4682 026 026 022 022 029 1 1 1 089 9767 A -01 0023 634E-06

rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 016 0035 652E-06

rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C -01 0022 654E-06

rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 01 0022 679E-06

rs10890236 1 NA 4683 009 009 009 009 009 1 1 1 093 100 C 016 0035 707E-06

rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 016 0037 708E-06

rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0022 728E-06

rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 015 0034 821E-06

rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 01 0022 823E-06

rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 015 0035 873E-06

rs7205409 16 RAB40C 4590 042 043 045 043 047 099 099 096 044 9933 C 009 0021 878E-06

rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A -047 0107 970E-06

Legend to Supplementary Table 2

Studies contributing data are denoted RS1 Rotterdam study cohort 1 RS3 Rotterdam study cohort 3 TUK TwinsUK BMI body mass index FHS

Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study

SNP single nucleotide polymorphism Chr chromosome position SNP location in base pairs MAF minor allele frequency

R2 from MACH for RS1 RS3 FHS and GARP Information score from IMPUTE for TUK

Eff All effect allele beta effect size SE standard error of beta p p value

Supplementary Figure 1 Quantile-quantile plot of meta-analysis adjusted GWA results

Legend to Supplementary Figure 1

The plots show GWA meta-analysis quantile-quantile plot of observed against expected results adjusted for age and sex

Supplementary Figure 2 Manhatten plot for meta-analysis of adjusted GWA results

Legend to Supplementary Figure 2

Plot shows combined results for the 5 studies included in the meta-analysis adjusted for age and sex The blue and red horizontal lines mark the levels of

suggestive and likely significance respectively

Supplementary Figure 3 Regional plot of association results and recombination rates for the class II MHC region unadjusted for age and sex

Legend to Supplementary Figure 3

minuslog10 P values (y axis) of the SNPs are shown according to their chromosomal positions (x axis) with lead SNP shown as a purple diamond The colour

intensity of each symbol depicting a SNP reflects the extent of LD with the rs926849 coloured red (r2 gt 08) through to blue (r

2 lt 02) Genetic

recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build 36 (NCBI) of the

human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn to show the relative

positions and therefore the maps are not to physical scale

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 3: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

KCL for collation and analysis using the statistical packageSTATA (StataCorp)23 Strict quality control was applied314 075 SNPs were retained for analysis (987) 733 wereexcluded because their call rates were le90 and 725 SNPs hadminor allele frequency lt001 In TUK significant populationsubstructure was excluded using the STRUCTURE program24

GWA analysisAll analyses were performed on inverse normal transformedsummary LDD score as described above Each study performedGWA analysis for LDD scores using either MACH2QTL (httpwwwsphumicheducsgabecasisMACHindexhtml) (RS1and RS3) or SNPTEST (httpwwwstatsoxacuk~marchinisoftwaregwassnptesthtml) (GARP) which use genotypedosage value as continuous additive predictors of LDD score ina linear regression framework or ProbABEL25 using an additivegenetic model while accounting for relatedness between themembers of a family Analysis of imputed genotype dataaccounted for uncertainty in each genotype prediction by usingeither the dosage information from MACH26 or the genotypeprobabilities from IMPUTE19

Meta-analysis of the five study groupsGenotypes for 25ndash3 million autosomal SNPs were imputed sep-arately to increase coverage using HapMap V2 (httpwwwhapmaporg) as the reference panel In GARP and TUK imput-ation was performed with IMPUTE V219 and in the otherstudies with MACH26 The common reference panel led to thereporting of results for the positive strand for all cohorts In add-ition allele pairs were compared between cohorts and no detect-able strand-flips were found the minor allele frequency was alsocompared between datasets The distributions of β values of thecohorts were found to be similar and therefore suitable formeta-analysis All directly genotyped or imputed autosomalSNPs having information from more than one study group(n=2 552 511) were included in the meta-analysis Associationresults were combined using inverse variance weighted fixedeffects meta-analysis using PLINK V106 (httppngumghharvardedupurcellplink) Two meta-analyses were run thefirst was unadjusted the second was adjusted for age and sex asboth are known risk factors for LDD and each risk factor wascorrelated with LDD in each study group Heterogeneity of esti-mated effect was expressed using Q (weighted sum of squares)and I2 (ratio of true heterogeneity to total observed variation)SNPs were excluded from the meta-analysis if the cohort-specificimputation quality as assessed by r2 (MACH) or InformationScore (IMPUTE) metric was lt040 On this basis one markerwas excluded from the unadjusted association and one from theadjusted association

DNA methylation data and analysisWhole blood DNA methylation levels were obtained for 38individuals in the TUK cohort using the IlluminaHumanMethylation27 DNA Analysis BeadChip assay as previ-ously described27 The sample included four monozygotic twinpairs eight dizygotic twin pairs and 14 unrelated individualsAt each CpG site within an individual the methylation levelwas presented as β which represents the ratio of intensitysignal obtained from the methylated beads in the array overthe sum of methylated and unmethylated bead signalsFollowing quality control checks we obtained DNA methyla-tion at three CpG sites in the promoter region of the PARK2gene within 2kb of the transcription start site The threeprobes (cg15832436 cg21926612 and cg24816866) mappeduniquely to the human genome (hg18) within two mismatches(see Bell et al28) We fitted linear mixed effects models to assessassociation between DNA methylation levels at the three CpGsites in the PARK2 promoter and LDD We regressed the rawmethylation levels on fixed-effect terms including methylationchip and LDD and random-effect terms denoting family struc-ture and zygosity and compared the association of differen-tially methylated regions with a null model which excludedLDD from the fixed-effects terms We also repeated the associ-ation analyses by normalising the methylation values at eachCpG site to N(0 1)

RESULTSThe study samples for the meta-analysis included 4683 indivi-duals of European ancestries Table 1 shows sample size demo-graphic characteristics LDD and lumbar spine imaging methodfor each independent cohort The majority of participants werefemale subjects (670) and the samples had a mean age of577 years Across the cohorts the mean level of LDD variedfrom 0011 to 346 reflecting differences in imaging methodsHowever the variance of the LDD variables was broadlysimilar (range 0958ndash114) as were the distributions of the esti-mated genetic effect sizes (β) The genotyping and imputationmethods are shown in online supplementary table S1

Quantilendashquantile plots for the unadjusted LDD GWA meta-analysis are presented in figure 1 (see online supplementaryfigure S1 adjusted) Test statistic inflation post meta-analysisas measured by the genomic control statistic29 was low(λGC unadjusted=102 λGC adjusted=103) Results of theunadjusted and adjusted association analyses were broadlysimilar with the p values of the adjusted analysis somewhatattenuated A Manhattan plot for the unadjusted analysis isshown in figure 2 with numeric results in table 2 (unadjusted)and online supplementary table S2 (adjusted) for SNPs havingplt10minus5

Table 1 Characteristics of the study samples

FHS GARP RS1 RS3 TUK

N 330 192 2440 974 744Age (years) 543 (110) 603 (71) 657 (67) 547 (34) 536 (83)Women () 422 797 57 59 968BMI (kgm2) 281 (51) 2644 (48) 263 (34) 2712 (46) 249 (44)Lumbar spine imaging CT Radiograph Radiograph Radiograph MRILDD variable 249 (097) 002 (0958) 0006 (0978) 0011 (0965) 346 (114)

Values are mean (SD) unless specified otherwiseBMI body mass index FHS Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study LDD Lumbar disc degeneration RS1 Rotterdam study cohort 1RS3 Rotterdam study cohort 3 TUK TwinsUK

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1143

Clinical and epidemiological research

Four markers achieved genome-wide significance in theunadjusted GWAS of which three were on chromosome 6(rs926849 rs2187689 rs7767277) and an intergenic marker onchromosome 3 (rs17034687) The results of the meta-analysisadjusted for age and gender were broadly similar the strongestsignal was for SNP rs926849 This SNP lies on an intronicregion of the Parkinson protein 2 E3 ubiquitin protein ligase(PARK2) gene on chromosome 6 A Forest plot of the groupsrsquoresults and their meta-analysis is shown in figure 3 Data were

available from four study groups and the range of estimatedminor allele frequencies was 023ndash032 Imputation quality washigh for all four studies contributing this SNP (gt090 table 2)and the estimated between-study heterogeneity was acceptable(I2=0 p(Q)=067) The results of a four-study meta-analysiswhich excluded GARP show the marker to have p=95eminus8

(unadjusted) The minor or C allele of rs926849 was associatedwith a lower level of LDD implying that the minor allele is pro-tective Figure 4 shows a regional plot of both genotyped andimputed SNPs within 200 Kb of the PARK2 gene along withrecombination rates

Two of the other strongly associated SNPs are in perfectlinkage disequilibrium (LD) rs2187689 and rs7767277 onchromosome 6 Data were available for four studies and therange of estimated allele frequency was 006ndash009 Imputationquality was high for all four studies (gt090) Both SNPs are instrong LD (r2=076) with an intronic marker on the prote-asome subunit β type 9 large multifunctional peptidase 2 gene(PSMB9) that is located in the class II region of the majorhistocompatibility complex (MHC) Both genotyped andimputed SNPs within 400 Kb of rs2187689 along with recom-bination rates are shown in a regional plot in onlinesupplementary figure S3 None of these top SNPs is in LD withknown functional SNPs in either PARK2 or PSMB9

We tested for an association between LDD and DNA methy-lation variants at three CpG sites in the PARK2 promoterA significant association between DNA methylation at CpGsite cg15832436 and LDD (β=874times10minus4 SE=249times10minus4p=0006) was observed The pattern of hypermethylation withincreasing LDD levels was reflected at the remaining promoterCpG sites however these did not reach nominal significance(cg21926612 β=0003 p=014 cg24816866 β=676times10minus4p=039) We repeated the analyses using normalised methyla-tion levels and observed that the association betweencg15832436 and LDD remained nominally significant

DISCUSSIONGWA offers an unbiased scan of common genetic variants(minor allele frequency gt5) and thus may deliver novel var-iants in genes not hitherto suspected of playing a role in discdegeneration This work is among the first to report on agenome-wide meta-analysis being conducted for LDD LDD isan age-related process which occurs in all people to someextent and may be detected as early as the teenage years1 LDDis known to have genetic determinants7 16 and its expression isalso influenced by gender (women develop LDD later) bodymass index30ndash34 and smoking35 Occupational factors also playa small role in LDD36 37 LDD as determined by MRI has beenimplicated in the development of episodes of severe and disab-ling low back pain3 We undertook this large meta-analysis inorder to identify novel genetic variants associated with LDDand to shed light on the underlying pathology of discdegeneration

GWA data obtained using differing chip technology may bereadily compared using imputation with HapMap In total2 552 511 overlapping markers were available in each cohortWe identified four markers having significant association withthe LDD phenotype (plt5times10minus8) There was similarity in theresults obtained with and without adjustment for the covari-ates age and sex A total of 26 markers had plt10minus5 in bothmeta-analyses As expected results of the adjusted analyseshad slightly attenuated p values (see online supplementarytable S2) which likely reflect the confounding effect of age Inboth analyses there were multiple associations to the Human

Figure 1 Quantilendashquantile plot of observed versus expecteddistribution of p values for the genome-wide association (GWA)meta-analysis The plots show GWA meta-analysis quantilendashquantileplot of observed against expected results unadjusted for covariates

Figure 2 Manhattan plot for meta-analysis of unadjustedgenome-wide association results Plot shows combined results for thefive studies included in the meta-analysis unadjusted results The blueand red horizontal lines mark the levels of suggestive and likelysignificance respectively

1144 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

Leukocyte Antigen (HLA) region and to markers in PARK2(Parkinson protein 2 E3 ubiquitin protein ligase) Among themost significant findings (table 2) is SNP rs926849 that lies at6q252ndash27 within an intron in the PARK2 gene a large gene of13 Mb comprising 12 exons The SNP encodes a change of basefrom T to C and is reported to have a minor allele frequency of023ndash034 in dbSNP which is keeping with the findings in ourstudy groups (table 2 figures 3 and 4) Although this SNP hasnot been directly genotyped by any study group estimates

suggest imputation to be accurate for rs926849 (range 95ndash

99 table 2) PARK2 encodes a protein called parkin which isa component of a multiprotein E3 ubiquitin ligase complexthat mediates the targeting of unwanted proteins for proteaso-mal degradation This complex also controls the level of pro-teins involved in cell activities such as cell division and growthit may be a tumour suppressor protein Parkin is also involvedin mitophagy it translocates from the cytosol and promotesthe degradation of uncoupled mitochondria38 Parkin is widely

Table 2 Results of the genome-wide association meta-analysis (unadjusted showing those SNPs having plt10minus5)

MAF Imputation quality

SNP Chr Gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP Eff all β SE p Value

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 023 0038 182E-09rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 023 0041 272E-08rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 023 0041 281E-08rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C minus013 0024 325E-08rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 02 0037 558E-08rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 02 0037 559E-08rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 019 0036 936E-08rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 019 0036 939E-08rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 019 0036 104E-07rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 019 0037 146E-07rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 019 0037 147E-07rs10214886 6 NA 4479 01 009 007 011 NA 098 1 092 097 NA A 019 0036 232E-07rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 019 0037 322E-07rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A minus012 0024 361E-07rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A minus012 0023 368E-07rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 019 0038 458E-07rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C minus026 0054 947E-07rs7896691 10 PFKP 4652 01 01 009 013 007 099 1 096 067 9761 C 017 0035 202E-06rs6602024 10 PFKP 4673 01 01 009 013 007 099 1 098 067 9762 A 017 0035 220E-06rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C minus011 0023 356E-06rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A minus053 0114 359E-06rs17132175 10 PFKP 4659 01 01 009 013 007 098 099 097 063 9713 C 016 0036 373E-06rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 011 0023 375E-06rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C minus013 0028 399E-06rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A minus011 0025 432E-06rs2484990 10 NA 4429 001 001 0 002 NA 056 082 047 093 NA C 068 0148 439E-06rs1250307 10 NA 4415 001 001 0 002 NA 056 082 045 094 NA A 068 0148 439E-06rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 011 0023 443E-06rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A minus013 0029 455E-06rs2484992 10 NA 4440 001 001 0 002 NA 056 082 048 093 NA C 068 0149 481E-06rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A minus028 0061 517E-06rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0023 543E-06rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0047 564E-06rs11918654 3 ARL8B 4310 027 027 029 027 NA 097 098 09 099 NA C minus011 0024 710E-06rs4881085 10 PFKP 4490 01 01 01 013 NA 1 1 099 067 NA A 016 0036 744E-06rs2657195 8 NA 4316 022 021 018 021 NA 085 085 087 074 NA A minus013 0028 754E-06rs11754641 6 NA 4478 003 003 002 004 NA 093 095 091 1 NA C 029 0064 784E-06rs12805875 11 NA 4674 042 041 042 043 041 1 1 1 1 9812 A 009 0021 851E-06rs980238 8 NA 4491 03 029 031 029 NA 1 1 1 1 NA A minus01 0023 911E-06rs7103004 11 NA 4667 042 041 042 043 041 1 1 099 1 9709 C 009 0021 920E-06rs4554859 11 NA 4674 042 041 042 043 041 1 1 1 1 9812 G 009 0021 924E-06rs7118412 11 NA 4669 042 041 042 043 041 1 1 099 1 9709 A 009 0021 937E-06rs2017567 16 NA 4683 042 043 042 042 047 099 1 1 045 100 C 01 0021 938E-06rs710924 16 PIGQ 4683 042 043 042 042 047 1 1 1 045 9964 C 009 0021 993E-06

r2 from MACH for RS1 RS3 FHS and GARP information score from IMPUTE for TUKStudies contributing data are denotedSNP single nucleiotide polymorphism Chr chromosome N number of subjects studied MAF minor allele frequency RS1 Rotterdam study cohort 1 RS3 Rotterdam studycohort 3 TUK TwinsUK FHS Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study Eff all effect allele β effect size SE SE of β

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1145

Clinical and epidemiological research

expressed in solid organs as well as skeletal muscle (httpwwwproteinatlasorg) Mutations within PARK2 are associated withdiverse conditions including autosomal recessive juvenileParkinsonrsquos disease Alzheimer rsquos disease diabetes mellitus andseveral solid tumours (reviewed in39) Parkin may account for theinverse relation between Parkinsonrsquos disease and cancer inci-dence40 Our findings of hypermethylation with increasing LDDscore suggest that PARK2 expression is reduced with increasingdisc degeneration but functional studies of intervertebral disc andother spine tissues are needed

Three further markers in the unadjusted meta-analysis hadplt5times10minus8 Marker rs17034687 is an intergenic marker onchromosome 3 Based on One Thousand Genomes (1KG)CEUdata it is not in LD (r2gt 03) with any known gene-basedmarkers Markers rs2187689 (supplementary figure S2) andrs7767277 are HLA region markers neither of which is includedin the 1KG pilot data Using data from HapMap V3 (release 2)rs2187689 and rs776277 are in perfect LD with each other andin LD (r2= 076) with an intronic marker in PSMB9 (prote-asome (prosome macropain) subunit β type 9 large

multifunctional peptidase 2) Proteasomes are distributedthroughout eukaroytic cells at high concentration and cleavepeptides in an ATPubiquitin-dependent process in a non-lysosomal pathway The gene is located in the class II region ofthe MHC Expression of the gene is induced by interferon γand this gene product replaces catalytic subunit 1 (proteasomeβ 6 subunit) in the immunoproteasome

While lumbar degeneration is not considered an inflamma-tory process and has not been reported to be auto-immune inaetiology there is evidence of pro-inflammatory cytokine acti-vation in herniated lumbar discs41 and anti-TNF has been usedsuccessfully to treat disc herniation-induced sciatica42 Of notethe COL11A2 gene lies 169 KB upstream from rs2187689 AnSNP (rs2076311) within this candidate gene has been shown tobe associated with MR determined disc signal intensity in acandidate gene study of Finnish male twins43 SNP rs2076311is not however in LD with our top hit rs2187689 (r2=0017)and so it seems unlikely that this collagen-encoding geneaccounts for our observed association Many published GWAstudies have identified SNPs in intergenic regions and the

Figure 4 Regional plot of association results and recombination rates for the PARK2 gene unadjusted results minuslog10 p values (y-axis) of the singlenucleotide polymorphisms (SNPs) are shown according to their chromosomal positions (x-axis) with lead SNP shown as a purple diamond Thecolour intensity of each symbol depicting an SNP reflects the extent of LD with the rs926849 coloured red (r2gt08) through to blue (r2lt02)Genetic recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build36 (NCBI) of the human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn toshow the relative positions and therefore the maps are not to physical scale

Figure 3 Forest plot of rs926849 in PARK2 unadjusted for covariates The contribution of the studies included in the meta-analysis is shown in thisfixed effects model The C allele is considered Heterogeneity I2=0 p(Q)=067 TE treatment effect seTE standard error treatment effect

1146 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

precise role of these regions is yet to be defined Long rangeenhancers for example could operate here and so an influenceon COL11A2 expression cannot be ruled out

Of suggestive significance is SNP rs4802666 (p=376times10minus6adjusted meta-analysis) which lies within the MYH14 genewhich encodes myosin heavy chain 14 non-muscle It isexpressed in cell lines derived from bone (httpwwwproteinatlasorg) and is implicated in autosomal dominanthearing impairment It is of interest in LDD because it lies onchromosome 19 under the linkage peak we have reported intwins for LDD44 and a peak reported by the Framinghamgroup for hand OA45 As there is a known relationship betweenthese two phenotypes this region on chromosome 19 forms ahighly plausible candidate region for OA It is not impossiblethat a muscle-expressed protein plays a role in LDD throughmechanisms similar to those proposed for OA considered bysome to be a multi-tissue syndrome rather than simply adisease of cartilage46

The main limitation of the study is one of obtaining anaccurate phenotype on individuals which is known to be animportant factor in the success of GWA47 There is at presentno agreed gold standard imaging method in the determinationof LDD although it is recognised that MRI offers the most sen-sitive widely available tool Even so MR is relatively expensiveand many of the largest spine cohorts in the world have plainradiographs which offer more limited phenotypic informationThe coding method applied to the imaging is also yet to be for-mally standardised our interest in the individual subtraits ofLDD led us to devise a coding method in which they wereseparated as reported previously7 In order to obtain sufficientsample size a number of cohorts contributed having differentimaging methods but traits were selected to enable comparisonacross the cohorts Thus study groups recoded their imagingwhere necessary to meet uniform requirements for inclusionWe included measures of disc height (coded 0ndash3) and anteriorosteophytes in RS1 RS3 GARP and TUK (also coded 0ndash3) andposterior osteophytes in FHS (coded 0ndash3) These subpheno-types were summed over the five discs and underwent inversenormal transformation to give a normal distribution A furtherlimitation is that four cohorts are population samples whileGARP is derived from OA-affected sibling pairs We includedGARP because it has made a contribution to similar analysesperformed for OA48 and with adjustment for relatedness pro-vides data comparable with other studies While the differingmethods of imaging provide different amounts of informationso the LDD variable has lower mean in those cohorts withradiographs the variance is comparable Where GARP samplesmade a contribution to the meta-analysis (a number of thesignificant SNPs did not include a contribution from GARPtable 2) the minor allele frequency was similar to those ofother groups The TUK group has a disproportionate numberof women for historical reasons The men were retainedhowever as they did not differ significantly from women inthe LDD variable or body mass index (data not shown) Thisstudy lacks a replication group A second sample of similar sizeto the first is considered important to show that the findingsof the first sample are true positives Unfortunately there areto our knowledge no other collections of Northern Europeanshaving spine imaging which together would approach oursample size There is considerable evidence in the literaturethat the genetic predisposition between Northern Europeansand Asians to OA is different49 and given the similaritiesbetween OA and LDD we felt replication should be made inNorthern Europeans We elected to include all the subjects in

a single powerful study rather than split the sample andreduce the chances of finding significant novel loci associatedwith LDD

In conclusion this is the first large-scale GWA study of LDDand we have identified several novel variants in the PARK2gene and in PSMB9 within MHC class 2 We have shown in asmall subset that methylation at one of the PARK2 promotersis associated with MRI determined LDD Both loci meritfurther investigation to shed light on the important clinicalendpoint of low back pain

Author affiliations1Department Twin Research and Genetic Epidemiology Kingrsquos College LondonLondon UK2Acclarogen Ltd St Johnrsquos Innovation Centre Cambridge UK3Department of Internal Medicine Erasmus MC Rotterdam The Netherlands4The Netherlands Genomics Initiative sponsored Netherlands Consortium for HealthyAging (NGI-NCHA) LeidenRotterdam The Netherlands5Department of Molecular Epidemiology Leiden University Medical Center LeidenRotterdam The Netherlands6VA Boston Healthcare System Harvard Medical School Spaulding RehabilitationHospital New England Baptist Hospital Boston USA7Department of Rheumatology University of Sydney and Royal North Shore HospitalSydney Australia8Department of Epidemiology Erasmus MC Rotterdam The Netherlands9Department of General Practice Erasmus MC Rotterdam The Netherlands10Department of Rheumatology and Department of Clinical Epidemiology LeidenUniversity Medical Center Leiden The Netherlands

Acknowledgements FMKW is a Wellcome Trust Intermediate Fellow The authorswould like to acknowledge the contribution of Maria Popham to the collection andanalysis of TwinsUK data The GWA study was funded by the NetherlandsOrganisation of Scientific Research NWO Investments (nr 1750102005011911-03-012) the Research Institute for Diseases in the Elderly (014-93-015 RIDE2)and the Netherlands Genomics Initiative (NGI)the Netherlands Consortium forHealthy Aging (NCHA) project nr 050-060-810 We thank Pascal Arp Mila JhamaiDr Michael Moorhouse Marijn Verkerk and Sander Bervoets for their help in creatingthe GWAS database The Rotterdam Study is funded by Erasmus Medical Center andErasmus University Rotterdam the Netherlands Organisation for the Health Researchand Development (ZonMw) the Research Institute for Diseases in the Elderly (RIDE)the Ministry of Education Culture and Science the Ministry for Health Welfare andSports the European Commission (DG XII) and the Municipality of Rotterdam Theauthors are very grateful to the participants and staff from the Rotterdam Study theparticipating general practitioners and the pharmacists We would like to thank DrTobias A Knoch Luc V de Zeeuw Anis Abuseiris and Rob de Graaf as well as theirinstitutions the Erasmus Computing Grid Rotterdam The Netherlands and especiallythe national German MediGRID and ServicesMediGRID part of the German D-Gridboth funded by the German Bundesministerium fuer Forschung und Technology undergrants 01 AK 803 A-H and 01 IG 07015 G for access to their grid resourcesThe GARP study was supported by the Leiden University Medical Centre and theDutch Arthritis Association Pfizer Inc Groton CT USA supported the inclusion ofthe GARP study The genotypic work was supported by the Netherlands Organisationof Scientific Research (MW 904-61-095 911-03-016 917 66344 and 911-03-012)Leiden University Medical Centre and the Centre of Medical System Biology and theNetherlands Consortium for Healthy Aging both in the framework of the NetherlandsGenomics Initiative (NGI) The research leading to these studies has received fundingfrom the European Unionrsquos Seventh Framework Programme (FP72007-2011) undergrant agreement ndeg 259679 Pradeep Suri was funded by K12HD01097 from theNational Institute of Health (US) and the Rehabilitation Medicine Scientist TrainingProgramme (RMSTP) The Framingham Study is supported by the National HeartLung and Blood Institutersquos Framingham Heart Study contract (No N01-HC-25195) forthe recruitment enrolment and examination of the Offspring and Third Generationcohorts and the imaging by CT scan

Contributors The authors listed fulfilled the criteria for authorship

Funding Fellowship to FMKW who led the project

Competing interests None

Ethics approval All bodies appropriate to the groups contributing data

Provenance and peer review Not commissioned externally peer reviewed

Open Access This is an Open Access article distributed in accordance with theCreative Commons Attribution Non Commercial (CC BY-NC 30) license whichpermits others to distribute remix adapt build upon this work non-commercially andlicense their derivative works on different terms provided the original work is

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1147

Clinical and epidemiological research

properly cited and the use is non-commercial See httpcreativecommonsorglicensesby-nc30

REFERENCES1 Powell MC Wilson M Szypryt P et al Prevalence of lumbar disc degeneration

observed by magnetic resonance in symptomless women Lancet 198621366ndash72 de Schepper EI Damen J van Meurs JB et al The association between lumbar

disc degeneration and low back pain the influence of age gender and individualradiographic features Spine (Phila Pa 1976) 201035531ndash6

3 MacGregor AJ Andrew T Sambrook PN et al Structural psychological andgenetic influences on low back and neck pain a study of adult female twinsArthritis Rheum 200451160ndash7

4 Phillips CJ Economic burden of chronic pain Expert Rev Pharmacoecon OutcomesRes 20066591ndash601

5 Roberts S Evans H Trivedi J et al Histology and pathology of the humanintervertebral disc J Bone Joint Surg Am 200688(Suppl 2)10ndash14

6 Battie MC Levalahti E Videman T et al Heritability of lumbar flexibility and therole of disc degeneration and body weight J Appl Physiol 2008104379ndash85

7 Sambrook PN MacGregor AJ Spector TD Genetic influences on cervical andlumbar disc degeneration a magnetic resonance imaging study in twins ArthritisRheum 199942366ndash72

8 Ryder JJ Garrison K Song F et al Genetic associations in peripheral jointosteoarthritis and spinal degenerative disease a systematic review Ann Rheum Dis200767584ndash91

9 Valdes AM Loughlin J Oene MV et al Sex and ethnic differences in theassociation of ASPN CALM1 COL2A1 COMP and FRZB with genetic susceptibilityto osteoarthritis of the knee Arthritis Rheum 200656137ndash46

10 Richards JB Kavvoura FK Rivadeneira F et al Collaborative meta-analysisassociations of 150 candidate genes with osteoporosis and osteoporotic fractureAnn Intern Med 2009151528ndash37

11 Manolio TA Collins FS Cox NJ et al Finding the missing heritability of complexdiseases Nature 2009461747ndash53

12 Hoffmann U Siebert U Bull-Stewart A et al Evidence for lower variability ofcoronary artery calcium mineral mass measurements by multi-detector computedtomography in a community-based cohortndashconsequences for progression studiesEur J Radiol 200657396ndash402

13 Riyazi N Meulenbelt I Kroon HM et al Evidence for familial aggregation of handhip and spine but not knee osteoarthritis in siblings with multiple joint involvementthe GARP study Ann Rheum Dis 200564438ndash43

14 Lane NE Nevitt MC Genant HK et al Reliability of new indices of radiographicosteoarthritis of the hand and hip and lumbar disc degeneration J Rheumatol1993201911ndash18

15 Hofman A Breteler MM van Duijn CM The Rotterdam Study 2010 objectives anddesign update Eur J Epidemiol 200924553ndash72

16 Bijkerk C Houwing-Duistermaat JJ Valkenburg HA et al Heritabilities of radiologicosteoarthritis in peripheral joints and of disc degeneration of the spine ArthritisRheum 1999421729ndash35

17 Spector TD Williams FM The UK Adult Twin Registry (TwinsUK) Twin Res HumGenet 20069899ndash906

18 Suchindran S Rivedal D Guyton JR et al Genome-wide association study of Lp-PLA(2) activity and mass in the Framingham Heart Study PLoS Genet 20106e1000928

19 Howie BN Donnelly P Marchini J A flexible and accurate genotype imputation method forthe next generation of genome-wide association studies PLoS Genet 20095e1000529

20 Marchini J Howie B Myers S et al A new multipoint method for genome-wideassociation studies by imputation of genotypes Nat Genet 200739906ndash13

21 Uh HW Wijk HJ Houwing-Duistermaat JJ Testing for genetic associationtaking into account phenotypic information of relatives BMC Proc 20093(Suppl 7)S123

22 Richards JB Rivadeneira F Inouye M et al Bone mineral density osteoporosisand osteoporotic fractures a genome-wide association study Lancet20083711505ndash12

23 Stata for Windows Version 10 edn College Station Texas USA Statacorp 200724 Pritchard JK Stephens M Donnelly P Inference of population structure using

multilocus genotype data Genetics 2000155945ndash59

25 Aulchenko YS Struchalin MV van Duijn CM ProbABEL package for genome-wideassociation analysis of imputed data BMC Bioinformatics 201011134

26 Li Y Willer CJ Ding J et al MaCH using sequence and genotype data to estimatehaplotypes and unobserved genotypes Genet Epidemiol 201034816ndash34

27 Bell JT Tsai PC Yang TP et al Epigenome-wide scans identify differentiallymethylated regions for age and age-related phenotypes in a healthy ageingpopulation PLoS Genet 20128e1002629

28 Bell JT Pai AA Pickrell JK et al DNA methylation patterns associate withgenetic and gene expression variation in HapMap cell lines Genome Biol 201112R10

29 Devlin B Roeder K Genomic control for association studies Biometrics199955997ndash1004

30 Liuke M Solovieva S Lamminen A et al Disc degeneration of the lumbar spine inrelation to overweight IntJObes (Lond) 200529903ndash8

31 Pye SR Reid DM Adams JE et al Influence of weight body mass index andlifestyle factors on radiographic features of lumbar disc degeneration Ann RheumDis 200766426ndash7

32 Symmons DP van Hemert AM Vandenbroucke JP et al A longitudinal study ofback pain and radiological changes in the lumbar spines of middle aged women IIRadiographic findings Ann Rheum Dis 199150162ndash6

33 Williams FM Popham M Livshits G et al A response to Videman et allsquochallenging the cumulative injury model positive effects of greater body mass ondisc degenerationrsquo Spine J 201010571ndash2

34 Yoshimura N Dennison E Wilman C et al Epidemiology of chronic discdegeneration and osteoarthritis of the lumbar spine in Britain and Japan acomparative study J Rheumatol 200027429ndash33

35 Battie MC Videman T Gill K et al 1991 Volvo Award in clinical sciences Smokingand lumbar intervertebral disc degeneration an MRI study of identical twins Spine1991161015ndash21

36 Videman T Battie MC The influence of occupation on lumbar degeneration Spine1999241164ndash8

37 Williams FM Sambrook PN Neck and back pain and intervertebral discdegeneration role of occupational factors Best Pract Res Clin Rheumatol20112569ndash79

38 Tanaka A Parkin-mediated selective mitochondrial autophagy mitophagy Parkinpurges damaged organelles from the vital mitochondrial network FEBS Lett20105841386ndash92

39 Kay DM Stevens CF Hamza TH et al A comprehensive analysis of deletionsmultiplications and copy number variations in PARK2 Neurology 2010751189ndash94

40 Gao X Ning Y Cancer and Parkinsons disease the odd couple Drugs Today (Barc)201147215ndash22

41 Kang JD Georgescu HI McIntyre-Larkin L et al Herniated lumbar intervertebraldiscs spontaneously produce matrix metalloproteinases nitric oxide interleukin-6and prostaglandin E2 Spine (Phila Pa 1976) 199621271ndash7

42 Korhonen T Karppinen J Paimela L et al The treatment of disc-herniation-inducedsciatica with infliximab one-year follow-up results of FIRST II a randomizedcontrolled trial Spine (Phila Pa 1976) 2006312759ndash66

43 Videman T Saarela J Kaprio J et al Associations of 25 structural degradativeand inflammatory candidate genes with lumbar disc desiccation bulging and heightnarrowing Arthritis Rheum 200960470ndash81

44 Williams FM Kato BS Livshits G et al Lumbar disc disease shows linkage tochromosome 19 overlapping with a QTL for hand OA Ann Rheum Dis200867117ndash19

45 Demissie S Cupples LA Myers R et al Genome scan for quantity of handosteoarthritis the Framingham Study Arthritis Rheum 200246946ndash52

46 Brandt KD Radin EL Dieppe PA et al Yet more evidence that osteoarthritis is nota cartilage disease Ann Rheum Dis 2006651261ndash4

47 Amos CI Successful design and conduct of genome-wide association studies HumMol Genet 200716R220ndashR225

48 Zhai G van Meurs JB Livshits G et al A genome-wide association studysuggests that a locus within the ataxin 2 binding protein 1 gene is associatedwith hand osteoarthritis the Treat-OA consortium J Med Genet 200946614ndash16

49 Valdes AM Spector TD Genetic epidemiology of hip and knee osteoarthritisNat Rev Rheumatol 2011723ndash32

1148 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

Supplementary Table 1 Genotyping and imputation methods by study

FHS GARP RS1 RS3 TUK

Genotyping

Platform Affymetrix Illumina Illumina Illumina Illumina

Chip 500K or 100K or 10 K Human660W

quad Human Hap550 Human Hap550

Human Hap300 or Human

Hap610Q

Call rate gt097 gt 095 gt 0975 gt 0975 gt090

MAF gt 005 gt 00025 gt 001 gt 001 gt001

HWE p-value gt 10-6

gt 10-6

gt 10-6

gt 10-6

gt 10-4

Imputation

Variants included for imputation 534982 462361 512349 514073 314075

Imputation software MACH vs1 IMPUTE MACH MACH IMPUTE vs 2

Genome build

HapMap Phase II

release 22

Hapmap Phase II

release 21

Hapmap Phase II

release 22

Hapmap Phase II

release 22

Hapmap Phase II

release 36

Total number of SNPs available 2543887 1830498 2543887 2543887 3044064

Legend to Supplementary Table 1

FHS represents Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study RS1 Rotterdam study cohort 1 RS3 Rotterdam study

cohort 3 TUK TwinsUK BMI body mass index MAF Minor Allele Frequency HWE Hardy-Weinberg equilibrium

Supplementary Table 2 Results of the GWA meta-analysis adjusted for age and sex showing those SNPs having plt10-5

MAF Imputation Quality

SNP Chr gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP

Eff

All BETA SE P

rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C -012 0023 950E-08

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 019 0037 367E-07

rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 019 0039 663E-07

rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 019 0039 664E-07

rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A -028 0058 730E-07

rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A -011 0022 742E-07

rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C -025 0051 135E-06

rs11136566 8 CSMD1 4610 039 039 036 039 046 099 098 097 096 9739 A -01 0021 224E-06

rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 016 0035 307E-06

rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 016 0035 307E-06

rs365962 16 NA 4608 044 044 048 047 042 099 099 096 069 9861 C -01 0020 310E-06

rs387953 16 NA 4683 046 045 048 046 043 1 099 1 067 9999 G -009 0020 337E-06

rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A -011 0023 357E-06

rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A -013 0027 376E-06

rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C -012 0026 468E-06

rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0045 477E-06

rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 016 0035 497E-06

rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 016 0035 593E-06

rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 016 0035 593E-06

rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A -01 0023 624E-06

rs7234567 18 NA 4682 026 026 022 022 029 1 1 1 089 9767 A -01 0023 634E-06

rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 016 0035 652E-06

rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C -01 0022 654E-06

rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 01 0022 679E-06

rs10890236 1 NA 4683 009 009 009 009 009 1 1 1 093 100 C 016 0035 707E-06

rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 016 0037 708E-06

rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0022 728E-06

rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 015 0034 821E-06

rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 01 0022 823E-06

rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 015 0035 873E-06

rs7205409 16 RAB40C 4590 042 043 045 043 047 099 099 096 044 9933 C 009 0021 878E-06

rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A -047 0107 970E-06

Legend to Supplementary Table 2

Studies contributing data are denoted RS1 Rotterdam study cohort 1 RS3 Rotterdam study cohort 3 TUK TwinsUK BMI body mass index FHS

Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study

SNP single nucleotide polymorphism Chr chromosome position SNP location in base pairs MAF minor allele frequency

R2 from MACH for RS1 RS3 FHS and GARP Information score from IMPUTE for TUK

Eff All effect allele beta effect size SE standard error of beta p p value

Supplementary Figure 1 Quantile-quantile plot of meta-analysis adjusted GWA results

Legend to Supplementary Figure 1

The plots show GWA meta-analysis quantile-quantile plot of observed against expected results adjusted for age and sex

Supplementary Figure 2 Manhatten plot for meta-analysis of adjusted GWA results

Legend to Supplementary Figure 2

Plot shows combined results for the 5 studies included in the meta-analysis adjusted for age and sex The blue and red horizontal lines mark the levels of

suggestive and likely significance respectively

Supplementary Figure 3 Regional plot of association results and recombination rates for the class II MHC region unadjusted for age and sex

Legend to Supplementary Figure 3

minuslog10 P values (y axis) of the SNPs are shown according to their chromosomal positions (x axis) with lead SNP shown as a purple diamond The colour

intensity of each symbol depicting a SNP reflects the extent of LD with the rs926849 coloured red (r2 gt 08) through to blue (r

2 lt 02) Genetic

recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build 36 (NCBI) of the

human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn to show the relative

positions and therefore the maps are not to physical scale

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 4: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

Four markers achieved genome-wide significance in theunadjusted GWAS of which three were on chromosome 6(rs926849 rs2187689 rs7767277) and an intergenic marker onchromosome 3 (rs17034687) The results of the meta-analysisadjusted for age and gender were broadly similar the strongestsignal was for SNP rs926849 This SNP lies on an intronicregion of the Parkinson protein 2 E3 ubiquitin protein ligase(PARK2) gene on chromosome 6 A Forest plot of the groupsrsquoresults and their meta-analysis is shown in figure 3 Data were

available from four study groups and the range of estimatedminor allele frequencies was 023ndash032 Imputation quality washigh for all four studies contributing this SNP (gt090 table 2)and the estimated between-study heterogeneity was acceptable(I2=0 p(Q)=067) The results of a four-study meta-analysiswhich excluded GARP show the marker to have p=95eminus8

(unadjusted) The minor or C allele of rs926849 was associatedwith a lower level of LDD implying that the minor allele is pro-tective Figure 4 shows a regional plot of both genotyped andimputed SNPs within 200 Kb of the PARK2 gene along withrecombination rates

Two of the other strongly associated SNPs are in perfectlinkage disequilibrium (LD) rs2187689 and rs7767277 onchromosome 6 Data were available for four studies and therange of estimated allele frequency was 006ndash009 Imputationquality was high for all four studies (gt090) Both SNPs are instrong LD (r2=076) with an intronic marker on the prote-asome subunit β type 9 large multifunctional peptidase 2 gene(PSMB9) that is located in the class II region of the majorhistocompatibility complex (MHC) Both genotyped andimputed SNPs within 400 Kb of rs2187689 along with recom-bination rates are shown in a regional plot in onlinesupplementary figure S3 None of these top SNPs is in LD withknown functional SNPs in either PARK2 or PSMB9

We tested for an association between LDD and DNA methy-lation variants at three CpG sites in the PARK2 promoterA significant association between DNA methylation at CpGsite cg15832436 and LDD (β=874times10minus4 SE=249times10minus4p=0006) was observed The pattern of hypermethylation withincreasing LDD levels was reflected at the remaining promoterCpG sites however these did not reach nominal significance(cg21926612 β=0003 p=014 cg24816866 β=676times10minus4p=039) We repeated the analyses using normalised methyla-tion levels and observed that the association betweencg15832436 and LDD remained nominally significant

DISCUSSIONGWA offers an unbiased scan of common genetic variants(minor allele frequency gt5) and thus may deliver novel var-iants in genes not hitherto suspected of playing a role in discdegeneration This work is among the first to report on agenome-wide meta-analysis being conducted for LDD LDD isan age-related process which occurs in all people to someextent and may be detected as early as the teenage years1 LDDis known to have genetic determinants7 16 and its expression isalso influenced by gender (women develop LDD later) bodymass index30ndash34 and smoking35 Occupational factors also playa small role in LDD36 37 LDD as determined by MRI has beenimplicated in the development of episodes of severe and disab-ling low back pain3 We undertook this large meta-analysis inorder to identify novel genetic variants associated with LDDand to shed light on the underlying pathology of discdegeneration

GWA data obtained using differing chip technology may bereadily compared using imputation with HapMap In total2 552 511 overlapping markers were available in each cohortWe identified four markers having significant association withthe LDD phenotype (plt5times10minus8) There was similarity in theresults obtained with and without adjustment for the covari-ates age and sex A total of 26 markers had plt10minus5 in bothmeta-analyses As expected results of the adjusted analyseshad slightly attenuated p values (see online supplementarytable S2) which likely reflect the confounding effect of age Inboth analyses there were multiple associations to the Human

Figure 1 Quantilendashquantile plot of observed versus expecteddistribution of p values for the genome-wide association (GWA)meta-analysis The plots show GWA meta-analysis quantilendashquantileplot of observed against expected results unadjusted for covariates

Figure 2 Manhattan plot for meta-analysis of unadjustedgenome-wide association results Plot shows combined results for thefive studies included in the meta-analysis unadjusted results The blueand red horizontal lines mark the levels of suggestive and likelysignificance respectively

1144 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

Leukocyte Antigen (HLA) region and to markers in PARK2(Parkinson protein 2 E3 ubiquitin protein ligase) Among themost significant findings (table 2) is SNP rs926849 that lies at6q252ndash27 within an intron in the PARK2 gene a large gene of13 Mb comprising 12 exons The SNP encodes a change of basefrom T to C and is reported to have a minor allele frequency of023ndash034 in dbSNP which is keeping with the findings in ourstudy groups (table 2 figures 3 and 4) Although this SNP hasnot been directly genotyped by any study group estimates

suggest imputation to be accurate for rs926849 (range 95ndash

99 table 2) PARK2 encodes a protein called parkin which isa component of a multiprotein E3 ubiquitin ligase complexthat mediates the targeting of unwanted proteins for proteaso-mal degradation This complex also controls the level of pro-teins involved in cell activities such as cell division and growthit may be a tumour suppressor protein Parkin is also involvedin mitophagy it translocates from the cytosol and promotesthe degradation of uncoupled mitochondria38 Parkin is widely

Table 2 Results of the genome-wide association meta-analysis (unadjusted showing those SNPs having plt10minus5)

MAF Imputation quality

SNP Chr Gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP Eff all β SE p Value

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 023 0038 182E-09rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 023 0041 272E-08rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 023 0041 281E-08rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C minus013 0024 325E-08rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 02 0037 558E-08rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 02 0037 559E-08rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 019 0036 936E-08rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 019 0036 939E-08rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 019 0036 104E-07rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 019 0037 146E-07rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 019 0037 147E-07rs10214886 6 NA 4479 01 009 007 011 NA 098 1 092 097 NA A 019 0036 232E-07rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 019 0037 322E-07rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A minus012 0024 361E-07rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A minus012 0023 368E-07rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 019 0038 458E-07rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C minus026 0054 947E-07rs7896691 10 PFKP 4652 01 01 009 013 007 099 1 096 067 9761 C 017 0035 202E-06rs6602024 10 PFKP 4673 01 01 009 013 007 099 1 098 067 9762 A 017 0035 220E-06rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C minus011 0023 356E-06rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A minus053 0114 359E-06rs17132175 10 PFKP 4659 01 01 009 013 007 098 099 097 063 9713 C 016 0036 373E-06rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 011 0023 375E-06rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C minus013 0028 399E-06rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A minus011 0025 432E-06rs2484990 10 NA 4429 001 001 0 002 NA 056 082 047 093 NA C 068 0148 439E-06rs1250307 10 NA 4415 001 001 0 002 NA 056 082 045 094 NA A 068 0148 439E-06rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 011 0023 443E-06rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A minus013 0029 455E-06rs2484992 10 NA 4440 001 001 0 002 NA 056 082 048 093 NA C 068 0149 481E-06rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A minus028 0061 517E-06rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0023 543E-06rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0047 564E-06rs11918654 3 ARL8B 4310 027 027 029 027 NA 097 098 09 099 NA C minus011 0024 710E-06rs4881085 10 PFKP 4490 01 01 01 013 NA 1 1 099 067 NA A 016 0036 744E-06rs2657195 8 NA 4316 022 021 018 021 NA 085 085 087 074 NA A minus013 0028 754E-06rs11754641 6 NA 4478 003 003 002 004 NA 093 095 091 1 NA C 029 0064 784E-06rs12805875 11 NA 4674 042 041 042 043 041 1 1 1 1 9812 A 009 0021 851E-06rs980238 8 NA 4491 03 029 031 029 NA 1 1 1 1 NA A minus01 0023 911E-06rs7103004 11 NA 4667 042 041 042 043 041 1 1 099 1 9709 C 009 0021 920E-06rs4554859 11 NA 4674 042 041 042 043 041 1 1 1 1 9812 G 009 0021 924E-06rs7118412 11 NA 4669 042 041 042 043 041 1 1 099 1 9709 A 009 0021 937E-06rs2017567 16 NA 4683 042 043 042 042 047 099 1 1 045 100 C 01 0021 938E-06rs710924 16 PIGQ 4683 042 043 042 042 047 1 1 1 045 9964 C 009 0021 993E-06

r2 from MACH for RS1 RS3 FHS and GARP information score from IMPUTE for TUKStudies contributing data are denotedSNP single nucleiotide polymorphism Chr chromosome N number of subjects studied MAF minor allele frequency RS1 Rotterdam study cohort 1 RS3 Rotterdam studycohort 3 TUK TwinsUK FHS Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study Eff all effect allele β effect size SE SE of β

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1145

Clinical and epidemiological research

expressed in solid organs as well as skeletal muscle (httpwwwproteinatlasorg) Mutations within PARK2 are associated withdiverse conditions including autosomal recessive juvenileParkinsonrsquos disease Alzheimer rsquos disease diabetes mellitus andseveral solid tumours (reviewed in39) Parkin may account for theinverse relation between Parkinsonrsquos disease and cancer inci-dence40 Our findings of hypermethylation with increasing LDDscore suggest that PARK2 expression is reduced with increasingdisc degeneration but functional studies of intervertebral disc andother spine tissues are needed

Three further markers in the unadjusted meta-analysis hadplt5times10minus8 Marker rs17034687 is an intergenic marker onchromosome 3 Based on One Thousand Genomes (1KG)CEUdata it is not in LD (r2gt 03) with any known gene-basedmarkers Markers rs2187689 (supplementary figure S2) andrs7767277 are HLA region markers neither of which is includedin the 1KG pilot data Using data from HapMap V3 (release 2)rs2187689 and rs776277 are in perfect LD with each other andin LD (r2= 076) with an intronic marker in PSMB9 (prote-asome (prosome macropain) subunit β type 9 large

multifunctional peptidase 2) Proteasomes are distributedthroughout eukaroytic cells at high concentration and cleavepeptides in an ATPubiquitin-dependent process in a non-lysosomal pathway The gene is located in the class II region ofthe MHC Expression of the gene is induced by interferon γand this gene product replaces catalytic subunit 1 (proteasomeβ 6 subunit) in the immunoproteasome

While lumbar degeneration is not considered an inflamma-tory process and has not been reported to be auto-immune inaetiology there is evidence of pro-inflammatory cytokine acti-vation in herniated lumbar discs41 and anti-TNF has been usedsuccessfully to treat disc herniation-induced sciatica42 Of notethe COL11A2 gene lies 169 KB upstream from rs2187689 AnSNP (rs2076311) within this candidate gene has been shown tobe associated with MR determined disc signal intensity in acandidate gene study of Finnish male twins43 SNP rs2076311is not however in LD with our top hit rs2187689 (r2=0017)and so it seems unlikely that this collagen-encoding geneaccounts for our observed association Many published GWAstudies have identified SNPs in intergenic regions and the

Figure 4 Regional plot of association results and recombination rates for the PARK2 gene unadjusted results minuslog10 p values (y-axis) of the singlenucleotide polymorphisms (SNPs) are shown according to their chromosomal positions (x-axis) with lead SNP shown as a purple diamond Thecolour intensity of each symbol depicting an SNP reflects the extent of LD with the rs926849 coloured red (r2gt08) through to blue (r2lt02)Genetic recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build36 (NCBI) of the human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn toshow the relative positions and therefore the maps are not to physical scale

Figure 3 Forest plot of rs926849 in PARK2 unadjusted for covariates The contribution of the studies included in the meta-analysis is shown in thisfixed effects model The C allele is considered Heterogeneity I2=0 p(Q)=067 TE treatment effect seTE standard error treatment effect

1146 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

precise role of these regions is yet to be defined Long rangeenhancers for example could operate here and so an influenceon COL11A2 expression cannot be ruled out

Of suggestive significance is SNP rs4802666 (p=376times10minus6adjusted meta-analysis) which lies within the MYH14 genewhich encodes myosin heavy chain 14 non-muscle It isexpressed in cell lines derived from bone (httpwwwproteinatlasorg) and is implicated in autosomal dominanthearing impairment It is of interest in LDD because it lies onchromosome 19 under the linkage peak we have reported intwins for LDD44 and a peak reported by the Framinghamgroup for hand OA45 As there is a known relationship betweenthese two phenotypes this region on chromosome 19 forms ahighly plausible candidate region for OA It is not impossiblethat a muscle-expressed protein plays a role in LDD throughmechanisms similar to those proposed for OA considered bysome to be a multi-tissue syndrome rather than simply adisease of cartilage46

The main limitation of the study is one of obtaining anaccurate phenotype on individuals which is known to be animportant factor in the success of GWA47 There is at presentno agreed gold standard imaging method in the determinationof LDD although it is recognised that MRI offers the most sen-sitive widely available tool Even so MR is relatively expensiveand many of the largest spine cohorts in the world have plainradiographs which offer more limited phenotypic informationThe coding method applied to the imaging is also yet to be for-mally standardised our interest in the individual subtraits ofLDD led us to devise a coding method in which they wereseparated as reported previously7 In order to obtain sufficientsample size a number of cohorts contributed having differentimaging methods but traits were selected to enable comparisonacross the cohorts Thus study groups recoded their imagingwhere necessary to meet uniform requirements for inclusionWe included measures of disc height (coded 0ndash3) and anteriorosteophytes in RS1 RS3 GARP and TUK (also coded 0ndash3) andposterior osteophytes in FHS (coded 0ndash3) These subpheno-types were summed over the five discs and underwent inversenormal transformation to give a normal distribution A furtherlimitation is that four cohorts are population samples whileGARP is derived from OA-affected sibling pairs We includedGARP because it has made a contribution to similar analysesperformed for OA48 and with adjustment for relatedness pro-vides data comparable with other studies While the differingmethods of imaging provide different amounts of informationso the LDD variable has lower mean in those cohorts withradiographs the variance is comparable Where GARP samplesmade a contribution to the meta-analysis (a number of thesignificant SNPs did not include a contribution from GARPtable 2) the minor allele frequency was similar to those ofother groups The TUK group has a disproportionate numberof women for historical reasons The men were retainedhowever as they did not differ significantly from women inthe LDD variable or body mass index (data not shown) Thisstudy lacks a replication group A second sample of similar sizeto the first is considered important to show that the findingsof the first sample are true positives Unfortunately there areto our knowledge no other collections of Northern Europeanshaving spine imaging which together would approach oursample size There is considerable evidence in the literaturethat the genetic predisposition between Northern Europeansand Asians to OA is different49 and given the similaritiesbetween OA and LDD we felt replication should be made inNorthern Europeans We elected to include all the subjects in

a single powerful study rather than split the sample andreduce the chances of finding significant novel loci associatedwith LDD

In conclusion this is the first large-scale GWA study of LDDand we have identified several novel variants in the PARK2gene and in PSMB9 within MHC class 2 We have shown in asmall subset that methylation at one of the PARK2 promotersis associated with MRI determined LDD Both loci meritfurther investigation to shed light on the important clinicalendpoint of low back pain

Author affiliations1Department Twin Research and Genetic Epidemiology Kingrsquos College LondonLondon UK2Acclarogen Ltd St Johnrsquos Innovation Centre Cambridge UK3Department of Internal Medicine Erasmus MC Rotterdam The Netherlands4The Netherlands Genomics Initiative sponsored Netherlands Consortium for HealthyAging (NGI-NCHA) LeidenRotterdam The Netherlands5Department of Molecular Epidemiology Leiden University Medical Center LeidenRotterdam The Netherlands6VA Boston Healthcare System Harvard Medical School Spaulding RehabilitationHospital New England Baptist Hospital Boston USA7Department of Rheumatology University of Sydney and Royal North Shore HospitalSydney Australia8Department of Epidemiology Erasmus MC Rotterdam The Netherlands9Department of General Practice Erasmus MC Rotterdam The Netherlands10Department of Rheumatology and Department of Clinical Epidemiology LeidenUniversity Medical Center Leiden The Netherlands

Acknowledgements FMKW is a Wellcome Trust Intermediate Fellow The authorswould like to acknowledge the contribution of Maria Popham to the collection andanalysis of TwinsUK data The GWA study was funded by the NetherlandsOrganisation of Scientific Research NWO Investments (nr 1750102005011911-03-012) the Research Institute for Diseases in the Elderly (014-93-015 RIDE2)and the Netherlands Genomics Initiative (NGI)the Netherlands Consortium forHealthy Aging (NCHA) project nr 050-060-810 We thank Pascal Arp Mila JhamaiDr Michael Moorhouse Marijn Verkerk and Sander Bervoets for their help in creatingthe GWAS database The Rotterdam Study is funded by Erasmus Medical Center andErasmus University Rotterdam the Netherlands Organisation for the Health Researchand Development (ZonMw) the Research Institute for Diseases in the Elderly (RIDE)the Ministry of Education Culture and Science the Ministry for Health Welfare andSports the European Commission (DG XII) and the Municipality of Rotterdam Theauthors are very grateful to the participants and staff from the Rotterdam Study theparticipating general practitioners and the pharmacists We would like to thank DrTobias A Knoch Luc V de Zeeuw Anis Abuseiris and Rob de Graaf as well as theirinstitutions the Erasmus Computing Grid Rotterdam The Netherlands and especiallythe national German MediGRID and ServicesMediGRID part of the German D-Gridboth funded by the German Bundesministerium fuer Forschung und Technology undergrants 01 AK 803 A-H and 01 IG 07015 G for access to their grid resourcesThe GARP study was supported by the Leiden University Medical Centre and theDutch Arthritis Association Pfizer Inc Groton CT USA supported the inclusion ofthe GARP study The genotypic work was supported by the Netherlands Organisationof Scientific Research (MW 904-61-095 911-03-016 917 66344 and 911-03-012)Leiden University Medical Centre and the Centre of Medical System Biology and theNetherlands Consortium for Healthy Aging both in the framework of the NetherlandsGenomics Initiative (NGI) The research leading to these studies has received fundingfrom the European Unionrsquos Seventh Framework Programme (FP72007-2011) undergrant agreement ndeg 259679 Pradeep Suri was funded by K12HD01097 from theNational Institute of Health (US) and the Rehabilitation Medicine Scientist TrainingProgramme (RMSTP) The Framingham Study is supported by the National HeartLung and Blood Institutersquos Framingham Heart Study contract (No N01-HC-25195) forthe recruitment enrolment and examination of the Offspring and Third Generationcohorts and the imaging by CT scan

Contributors The authors listed fulfilled the criteria for authorship

Funding Fellowship to FMKW who led the project

Competing interests None

Ethics approval All bodies appropriate to the groups contributing data

Provenance and peer review Not commissioned externally peer reviewed

Open Access This is an Open Access article distributed in accordance with theCreative Commons Attribution Non Commercial (CC BY-NC 30) license whichpermits others to distribute remix adapt build upon this work non-commercially andlicense their derivative works on different terms provided the original work is

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1147

Clinical and epidemiological research

properly cited and the use is non-commercial See httpcreativecommonsorglicensesby-nc30

REFERENCES1 Powell MC Wilson M Szypryt P et al Prevalence of lumbar disc degeneration

observed by magnetic resonance in symptomless women Lancet 198621366ndash72 de Schepper EI Damen J van Meurs JB et al The association between lumbar

disc degeneration and low back pain the influence of age gender and individualradiographic features Spine (Phila Pa 1976) 201035531ndash6

3 MacGregor AJ Andrew T Sambrook PN et al Structural psychological andgenetic influences on low back and neck pain a study of adult female twinsArthritis Rheum 200451160ndash7

4 Phillips CJ Economic burden of chronic pain Expert Rev Pharmacoecon OutcomesRes 20066591ndash601

5 Roberts S Evans H Trivedi J et al Histology and pathology of the humanintervertebral disc J Bone Joint Surg Am 200688(Suppl 2)10ndash14

6 Battie MC Levalahti E Videman T et al Heritability of lumbar flexibility and therole of disc degeneration and body weight J Appl Physiol 2008104379ndash85

7 Sambrook PN MacGregor AJ Spector TD Genetic influences on cervical andlumbar disc degeneration a magnetic resonance imaging study in twins ArthritisRheum 199942366ndash72

8 Ryder JJ Garrison K Song F et al Genetic associations in peripheral jointosteoarthritis and spinal degenerative disease a systematic review Ann Rheum Dis200767584ndash91

9 Valdes AM Loughlin J Oene MV et al Sex and ethnic differences in theassociation of ASPN CALM1 COL2A1 COMP and FRZB with genetic susceptibilityto osteoarthritis of the knee Arthritis Rheum 200656137ndash46

10 Richards JB Kavvoura FK Rivadeneira F et al Collaborative meta-analysisassociations of 150 candidate genes with osteoporosis and osteoporotic fractureAnn Intern Med 2009151528ndash37

11 Manolio TA Collins FS Cox NJ et al Finding the missing heritability of complexdiseases Nature 2009461747ndash53

12 Hoffmann U Siebert U Bull-Stewart A et al Evidence for lower variability ofcoronary artery calcium mineral mass measurements by multi-detector computedtomography in a community-based cohortndashconsequences for progression studiesEur J Radiol 200657396ndash402

13 Riyazi N Meulenbelt I Kroon HM et al Evidence for familial aggregation of handhip and spine but not knee osteoarthritis in siblings with multiple joint involvementthe GARP study Ann Rheum Dis 200564438ndash43

14 Lane NE Nevitt MC Genant HK et al Reliability of new indices of radiographicosteoarthritis of the hand and hip and lumbar disc degeneration J Rheumatol1993201911ndash18

15 Hofman A Breteler MM van Duijn CM The Rotterdam Study 2010 objectives anddesign update Eur J Epidemiol 200924553ndash72

16 Bijkerk C Houwing-Duistermaat JJ Valkenburg HA et al Heritabilities of radiologicosteoarthritis in peripheral joints and of disc degeneration of the spine ArthritisRheum 1999421729ndash35

17 Spector TD Williams FM The UK Adult Twin Registry (TwinsUK) Twin Res HumGenet 20069899ndash906

18 Suchindran S Rivedal D Guyton JR et al Genome-wide association study of Lp-PLA(2) activity and mass in the Framingham Heart Study PLoS Genet 20106e1000928

19 Howie BN Donnelly P Marchini J A flexible and accurate genotype imputation method forthe next generation of genome-wide association studies PLoS Genet 20095e1000529

20 Marchini J Howie B Myers S et al A new multipoint method for genome-wideassociation studies by imputation of genotypes Nat Genet 200739906ndash13

21 Uh HW Wijk HJ Houwing-Duistermaat JJ Testing for genetic associationtaking into account phenotypic information of relatives BMC Proc 20093(Suppl 7)S123

22 Richards JB Rivadeneira F Inouye M et al Bone mineral density osteoporosisand osteoporotic fractures a genome-wide association study Lancet20083711505ndash12

23 Stata for Windows Version 10 edn College Station Texas USA Statacorp 200724 Pritchard JK Stephens M Donnelly P Inference of population structure using

multilocus genotype data Genetics 2000155945ndash59

25 Aulchenko YS Struchalin MV van Duijn CM ProbABEL package for genome-wideassociation analysis of imputed data BMC Bioinformatics 201011134

26 Li Y Willer CJ Ding J et al MaCH using sequence and genotype data to estimatehaplotypes and unobserved genotypes Genet Epidemiol 201034816ndash34

27 Bell JT Tsai PC Yang TP et al Epigenome-wide scans identify differentiallymethylated regions for age and age-related phenotypes in a healthy ageingpopulation PLoS Genet 20128e1002629

28 Bell JT Pai AA Pickrell JK et al DNA methylation patterns associate withgenetic and gene expression variation in HapMap cell lines Genome Biol 201112R10

29 Devlin B Roeder K Genomic control for association studies Biometrics199955997ndash1004

30 Liuke M Solovieva S Lamminen A et al Disc degeneration of the lumbar spine inrelation to overweight IntJObes (Lond) 200529903ndash8

31 Pye SR Reid DM Adams JE et al Influence of weight body mass index andlifestyle factors on radiographic features of lumbar disc degeneration Ann RheumDis 200766426ndash7

32 Symmons DP van Hemert AM Vandenbroucke JP et al A longitudinal study ofback pain and radiological changes in the lumbar spines of middle aged women IIRadiographic findings Ann Rheum Dis 199150162ndash6

33 Williams FM Popham M Livshits G et al A response to Videman et allsquochallenging the cumulative injury model positive effects of greater body mass ondisc degenerationrsquo Spine J 201010571ndash2

34 Yoshimura N Dennison E Wilman C et al Epidemiology of chronic discdegeneration and osteoarthritis of the lumbar spine in Britain and Japan acomparative study J Rheumatol 200027429ndash33

35 Battie MC Videman T Gill K et al 1991 Volvo Award in clinical sciences Smokingand lumbar intervertebral disc degeneration an MRI study of identical twins Spine1991161015ndash21

36 Videman T Battie MC The influence of occupation on lumbar degeneration Spine1999241164ndash8

37 Williams FM Sambrook PN Neck and back pain and intervertebral discdegeneration role of occupational factors Best Pract Res Clin Rheumatol20112569ndash79

38 Tanaka A Parkin-mediated selective mitochondrial autophagy mitophagy Parkinpurges damaged organelles from the vital mitochondrial network FEBS Lett20105841386ndash92

39 Kay DM Stevens CF Hamza TH et al A comprehensive analysis of deletionsmultiplications and copy number variations in PARK2 Neurology 2010751189ndash94

40 Gao X Ning Y Cancer and Parkinsons disease the odd couple Drugs Today (Barc)201147215ndash22

41 Kang JD Georgescu HI McIntyre-Larkin L et al Herniated lumbar intervertebraldiscs spontaneously produce matrix metalloproteinases nitric oxide interleukin-6and prostaglandin E2 Spine (Phila Pa 1976) 199621271ndash7

42 Korhonen T Karppinen J Paimela L et al The treatment of disc-herniation-inducedsciatica with infliximab one-year follow-up results of FIRST II a randomizedcontrolled trial Spine (Phila Pa 1976) 2006312759ndash66

43 Videman T Saarela J Kaprio J et al Associations of 25 structural degradativeand inflammatory candidate genes with lumbar disc desiccation bulging and heightnarrowing Arthritis Rheum 200960470ndash81

44 Williams FM Kato BS Livshits G et al Lumbar disc disease shows linkage tochromosome 19 overlapping with a QTL for hand OA Ann Rheum Dis200867117ndash19

45 Demissie S Cupples LA Myers R et al Genome scan for quantity of handosteoarthritis the Framingham Study Arthritis Rheum 200246946ndash52

46 Brandt KD Radin EL Dieppe PA et al Yet more evidence that osteoarthritis is nota cartilage disease Ann Rheum Dis 2006651261ndash4

47 Amos CI Successful design and conduct of genome-wide association studies HumMol Genet 200716R220ndashR225

48 Zhai G van Meurs JB Livshits G et al A genome-wide association studysuggests that a locus within the ataxin 2 binding protein 1 gene is associatedwith hand osteoarthritis the Treat-OA consortium J Med Genet 200946614ndash16

49 Valdes AM Spector TD Genetic epidemiology of hip and knee osteoarthritisNat Rev Rheumatol 2011723ndash32

1148 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

Supplementary Table 1 Genotyping and imputation methods by study

FHS GARP RS1 RS3 TUK

Genotyping

Platform Affymetrix Illumina Illumina Illumina Illumina

Chip 500K or 100K or 10 K Human660W

quad Human Hap550 Human Hap550

Human Hap300 or Human

Hap610Q

Call rate gt097 gt 095 gt 0975 gt 0975 gt090

MAF gt 005 gt 00025 gt 001 gt 001 gt001

HWE p-value gt 10-6

gt 10-6

gt 10-6

gt 10-6

gt 10-4

Imputation

Variants included for imputation 534982 462361 512349 514073 314075

Imputation software MACH vs1 IMPUTE MACH MACH IMPUTE vs 2

Genome build

HapMap Phase II

release 22

Hapmap Phase II

release 21

Hapmap Phase II

release 22

Hapmap Phase II

release 22

Hapmap Phase II

release 36

Total number of SNPs available 2543887 1830498 2543887 2543887 3044064

Legend to Supplementary Table 1

FHS represents Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study RS1 Rotterdam study cohort 1 RS3 Rotterdam study

cohort 3 TUK TwinsUK BMI body mass index MAF Minor Allele Frequency HWE Hardy-Weinberg equilibrium

Supplementary Table 2 Results of the GWA meta-analysis adjusted for age and sex showing those SNPs having plt10-5

MAF Imputation Quality

SNP Chr gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP

Eff

All BETA SE P

rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C -012 0023 950E-08

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 019 0037 367E-07

rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 019 0039 663E-07

rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 019 0039 664E-07

rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A -028 0058 730E-07

rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A -011 0022 742E-07

rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C -025 0051 135E-06

rs11136566 8 CSMD1 4610 039 039 036 039 046 099 098 097 096 9739 A -01 0021 224E-06

rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 016 0035 307E-06

rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 016 0035 307E-06

rs365962 16 NA 4608 044 044 048 047 042 099 099 096 069 9861 C -01 0020 310E-06

rs387953 16 NA 4683 046 045 048 046 043 1 099 1 067 9999 G -009 0020 337E-06

rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A -011 0023 357E-06

rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A -013 0027 376E-06

rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C -012 0026 468E-06

rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0045 477E-06

rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 016 0035 497E-06

rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 016 0035 593E-06

rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 016 0035 593E-06

rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A -01 0023 624E-06

rs7234567 18 NA 4682 026 026 022 022 029 1 1 1 089 9767 A -01 0023 634E-06

rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 016 0035 652E-06

rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C -01 0022 654E-06

rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 01 0022 679E-06

rs10890236 1 NA 4683 009 009 009 009 009 1 1 1 093 100 C 016 0035 707E-06

rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 016 0037 708E-06

rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0022 728E-06

rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 015 0034 821E-06

rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 01 0022 823E-06

rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 015 0035 873E-06

rs7205409 16 RAB40C 4590 042 043 045 043 047 099 099 096 044 9933 C 009 0021 878E-06

rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A -047 0107 970E-06

Legend to Supplementary Table 2

Studies contributing data are denoted RS1 Rotterdam study cohort 1 RS3 Rotterdam study cohort 3 TUK TwinsUK BMI body mass index FHS

Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study

SNP single nucleotide polymorphism Chr chromosome position SNP location in base pairs MAF minor allele frequency

R2 from MACH for RS1 RS3 FHS and GARP Information score from IMPUTE for TUK

Eff All effect allele beta effect size SE standard error of beta p p value

Supplementary Figure 1 Quantile-quantile plot of meta-analysis adjusted GWA results

Legend to Supplementary Figure 1

The plots show GWA meta-analysis quantile-quantile plot of observed against expected results adjusted for age and sex

Supplementary Figure 2 Manhatten plot for meta-analysis of adjusted GWA results

Legend to Supplementary Figure 2

Plot shows combined results for the 5 studies included in the meta-analysis adjusted for age and sex The blue and red horizontal lines mark the levels of

suggestive and likely significance respectively

Supplementary Figure 3 Regional plot of association results and recombination rates for the class II MHC region unadjusted for age and sex

Legend to Supplementary Figure 3

minuslog10 P values (y axis) of the SNPs are shown according to their chromosomal positions (x axis) with lead SNP shown as a purple diamond The colour

intensity of each symbol depicting a SNP reflects the extent of LD with the rs926849 coloured red (r2 gt 08) through to blue (r

2 lt 02) Genetic

recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build 36 (NCBI) of the

human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn to show the relative

positions and therefore the maps are not to physical scale

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 5: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

Leukocyte Antigen (HLA) region and to markers in PARK2(Parkinson protein 2 E3 ubiquitin protein ligase) Among themost significant findings (table 2) is SNP rs926849 that lies at6q252ndash27 within an intron in the PARK2 gene a large gene of13 Mb comprising 12 exons The SNP encodes a change of basefrom T to C and is reported to have a minor allele frequency of023ndash034 in dbSNP which is keeping with the findings in ourstudy groups (table 2 figures 3 and 4) Although this SNP hasnot been directly genotyped by any study group estimates

suggest imputation to be accurate for rs926849 (range 95ndash

99 table 2) PARK2 encodes a protein called parkin which isa component of a multiprotein E3 ubiquitin ligase complexthat mediates the targeting of unwanted proteins for proteaso-mal degradation This complex also controls the level of pro-teins involved in cell activities such as cell division and growthit may be a tumour suppressor protein Parkin is also involvedin mitophagy it translocates from the cytosol and promotesthe degradation of uncoupled mitochondria38 Parkin is widely

Table 2 Results of the genome-wide association meta-analysis (unadjusted showing those SNPs having plt10minus5)

MAF Imputation quality

SNP Chr Gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP Eff all β SE p Value

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 023 0038 182E-09rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 023 0041 272E-08rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 023 0041 281E-08rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C minus013 0024 325E-08rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 02 0037 558E-08rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 02 0037 559E-08rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 019 0036 936E-08rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 019 0036 939E-08rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 019 0036 104E-07rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 019 0037 146E-07rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 019 0037 147E-07rs10214886 6 NA 4479 01 009 007 011 NA 098 1 092 097 NA A 019 0036 232E-07rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 019 0037 322E-07rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A minus012 0024 361E-07rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A minus012 0023 368E-07rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 019 0038 458E-07rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C minus026 0054 947E-07rs7896691 10 PFKP 4652 01 01 009 013 007 099 1 096 067 9761 C 017 0035 202E-06rs6602024 10 PFKP 4673 01 01 009 013 007 099 1 098 067 9762 A 017 0035 220E-06rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C minus011 0023 356E-06rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A minus053 0114 359E-06rs17132175 10 PFKP 4659 01 01 009 013 007 098 099 097 063 9713 C 016 0036 373E-06rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 011 0023 375E-06rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C minus013 0028 399E-06rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A minus011 0025 432E-06rs2484990 10 NA 4429 001 001 0 002 NA 056 082 047 093 NA C 068 0148 439E-06rs1250307 10 NA 4415 001 001 0 002 NA 056 082 045 094 NA A 068 0148 439E-06rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 011 0023 443E-06rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A minus013 0029 455E-06rs2484992 10 NA 4440 001 001 0 002 NA 056 082 048 093 NA C 068 0149 481E-06rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A minus028 0061 517E-06rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0023 543E-06rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0047 564E-06rs11918654 3 ARL8B 4310 027 027 029 027 NA 097 098 09 099 NA C minus011 0024 710E-06rs4881085 10 PFKP 4490 01 01 01 013 NA 1 1 099 067 NA A 016 0036 744E-06rs2657195 8 NA 4316 022 021 018 021 NA 085 085 087 074 NA A minus013 0028 754E-06rs11754641 6 NA 4478 003 003 002 004 NA 093 095 091 1 NA C 029 0064 784E-06rs12805875 11 NA 4674 042 041 042 043 041 1 1 1 1 9812 A 009 0021 851E-06rs980238 8 NA 4491 03 029 031 029 NA 1 1 1 1 NA A minus01 0023 911E-06rs7103004 11 NA 4667 042 041 042 043 041 1 1 099 1 9709 C 009 0021 920E-06rs4554859 11 NA 4674 042 041 042 043 041 1 1 1 1 9812 G 009 0021 924E-06rs7118412 11 NA 4669 042 041 042 043 041 1 1 099 1 9709 A 009 0021 937E-06rs2017567 16 NA 4683 042 043 042 042 047 099 1 1 045 100 C 01 0021 938E-06rs710924 16 PIGQ 4683 042 043 042 042 047 1 1 1 045 9964 C 009 0021 993E-06

r2 from MACH for RS1 RS3 FHS and GARP information score from IMPUTE for TUKStudies contributing data are denotedSNP single nucleiotide polymorphism Chr chromosome N number of subjects studied MAF minor allele frequency RS1 Rotterdam study cohort 1 RS3 Rotterdam studycohort 3 TUK TwinsUK FHS Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study Eff all effect allele β effect size SE SE of β

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1145

Clinical and epidemiological research

expressed in solid organs as well as skeletal muscle (httpwwwproteinatlasorg) Mutations within PARK2 are associated withdiverse conditions including autosomal recessive juvenileParkinsonrsquos disease Alzheimer rsquos disease diabetes mellitus andseveral solid tumours (reviewed in39) Parkin may account for theinverse relation between Parkinsonrsquos disease and cancer inci-dence40 Our findings of hypermethylation with increasing LDDscore suggest that PARK2 expression is reduced with increasingdisc degeneration but functional studies of intervertebral disc andother spine tissues are needed

Three further markers in the unadjusted meta-analysis hadplt5times10minus8 Marker rs17034687 is an intergenic marker onchromosome 3 Based on One Thousand Genomes (1KG)CEUdata it is not in LD (r2gt 03) with any known gene-basedmarkers Markers rs2187689 (supplementary figure S2) andrs7767277 are HLA region markers neither of which is includedin the 1KG pilot data Using data from HapMap V3 (release 2)rs2187689 and rs776277 are in perfect LD with each other andin LD (r2= 076) with an intronic marker in PSMB9 (prote-asome (prosome macropain) subunit β type 9 large

multifunctional peptidase 2) Proteasomes are distributedthroughout eukaroytic cells at high concentration and cleavepeptides in an ATPubiquitin-dependent process in a non-lysosomal pathway The gene is located in the class II region ofthe MHC Expression of the gene is induced by interferon γand this gene product replaces catalytic subunit 1 (proteasomeβ 6 subunit) in the immunoproteasome

While lumbar degeneration is not considered an inflamma-tory process and has not been reported to be auto-immune inaetiology there is evidence of pro-inflammatory cytokine acti-vation in herniated lumbar discs41 and anti-TNF has been usedsuccessfully to treat disc herniation-induced sciatica42 Of notethe COL11A2 gene lies 169 KB upstream from rs2187689 AnSNP (rs2076311) within this candidate gene has been shown tobe associated with MR determined disc signal intensity in acandidate gene study of Finnish male twins43 SNP rs2076311is not however in LD with our top hit rs2187689 (r2=0017)and so it seems unlikely that this collagen-encoding geneaccounts for our observed association Many published GWAstudies have identified SNPs in intergenic regions and the

Figure 4 Regional plot of association results and recombination rates for the PARK2 gene unadjusted results minuslog10 p values (y-axis) of the singlenucleotide polymorphisms (SNPs) are shown according to their chromosomal positions (x-axis) with lead SNP shown as a purple diamond Thecolour intensity of each symbol depicting an SNP reflects the extent of LD with the rs926849 coloured red (r2gt08) through to blue (r2lt02)Genetic recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build36 (NCBI) of the human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn toshow the relative positions and therefore the maps are not to physical scale

Figure 3 Forest plot of rs926849 in PARK2 unadjusted for covariates The contribution of the studies included in the meta-analysis is shown in thisfixed effects model The C allele is considered Heterogeneity I2=0 p(Q)=067 TE treatment effect seTE standard error treatment effect

1146 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

precise role of these regions is yet to be defined Long rangeenhancers for example could operate here and so an influenceon COL11A2 expression cannot be ruled out

Of suggestive significance is SNP rs4802666 (p=376times10minus6adjusted meta-analysis) which lies within the MYH14 genewhich encodes myosin heavy chain 14 non-muscle It isexpressed in cell lines derived from bone (httpwwwproteinatlasorg) and is implicated in autosomal dominanthearing impairment It is of interest in LDD because it lies onchromosome 19 under the linkage peak we have reported intwins for LDD44 and a peak reported by the Framinghamgroup for hand OA45 As there is a known relationship betweenthese two phenotypes this region on chromosome 19 forms ahighly plausible candidate region for OA It is not impossiblethat a muscle-expressed protein plays a role in LDD throughmechanisms similar to those proposed for OA considered bysome to be a multi-tissue syndrome rather than simply adisease of cartilage46

The main limitation of the study is one of obtaining anaccurate phenotype on individuals which is known to be animportant factor in the success of GWA47 There is at presentno agreed gold standard imaging method in the determinationof LDD although it is recognised that MRI offers the most sen-sitive widely available tool Even so MR is relatively expensiveand many of the largest spine cohorts in the world have plainradiographs which offer more limited phenotypic informationThe coding method applied to the imaging is also yet to be for-mally standardised our interest in the individual subtraits ofLDD led us to devise a coding method in which they wereseparated as reported previously7 In order to obtain sufficientsample size a number of cohorts contributed having differentimaging methods but traits were selected to enable comparisonacross the cohorts Thus study groups recoded their imagingwhere necessary to meet uniform requirements for inclusionWe included measures of disc height (coded 0ndash3) and anteriorosteophytes in RS1 RS3 GARP and TUK (also coded 0ndash3) andposterior osteophytes in FHS (coded 0ndash3) These subpheno-types were summed over the five discs and underwent inversenormal transformation to give a normal distribution A furtherlimitation is that four cohorts are population samples whileGARP is derived from OA-affected sibling pairs We includedGARP because it has made a contribution to similar analysesperformed for OA48 and with adjustment for relatedness pro-vides data comparable with other studies While the differingmethods of imaging provide different amounts of informationso the LDD variable has lower mean in those cohorts withradiographs the variance is comparable Where GARP samplesmade a contribution to the meta-analysis (a number of thesignificant SNPs did not include a contribution from GARPtable 2) the minor allele frequency was similar to those ofother groups The TUK group has a disproportionate numberof women for historical reasons The men were retainedhowever as they did not differ significantly from women inthe LDD variable or body mass index (data not shown) Thisstudy lacks a replication group A second sample of similar sizeto the first is considered important to show that the findingsof the first sample are true positives Unfortunately there areto our knowledge no other collections of Northern Europeanshaving spine imaging which together would approach oursample size There is considerable evidence in the literaturethat the genetic predisposition between Northern Europeansand Asians to OA is different49 and given the similaritiesbetween OA and LDD we felt replication should be made inNorthern Europeans We elected to include all the subjects in

a single powerful study rather than split the sample andreduce the chances of finding significant novel loci associatedwith LDD

In conclusion this is the first large-scale GWA study of LDDand we have identified several novel variants in the PARK2gene and in PSMB9 within MHC class 2 We have shown in asmall subset that methylation at one of the PARK2 promotersis associated with MRI determined LDD Both loci meritfurther investigation to shed light on the important clinicalendpoint of low back pain

Author affiliations1Department Twin Research and Genetic Epidemiology Kingrsquos College LondonLondon UK2Acclarogen Ltd St Johnrsquos Innovation Centre Cambridge UK3Department of Internal Medicine Erasmus MC Rotterdam The Netherlands4The Netherlands Genomics Initiative sponsored Netherlands Consortium for HealthyAging (NGI-NCHA) LeidenRotterdam The Netherlands5Department of Molecular Epidemiology Leiden University Medical Center LeidenRotterdam The Netherlands6VA Boston Healthcare System Harvard Medical School Spaulding RehabilitationHospital New England Baptist Hospital Boston USA7Department of Rheumatology University of Sydney and Royal North Shore HospitalSydney Australia8Department of Epidemiology Erasmus MC Rotterdam The Netherlands9Department of General Practice Erasmus MC Rotterdam The Netherlands10Department of Rheumatology and Department of Clinical Epidemiology LeidenUniversity Medical Center Leiden The Netherlands

Acknowledgements FMKW is a Wellcome Trust Intermediate Fellow The authorswould like to acknowledge the contribution of Maria Popham to the collection andanalysis of TwinsUK data The GWA study was funded by the NetherlandsOrganisation of Scientific Research NWO Investments (nr 1750102005011911-03-012) the Research Institute for Diseases in the Elderly (014-93-015 RIDE2)and the Netherlands Genomics Initiative (NGI)the Netherlands Consortium forHealthy Aging (NCHA) project nr 050-060-810 We thank Pascal Arp Mila JhamaiDr Michael Moorhouse Marijn Verkerk and Sander Bervoets for their help in creatingthe GWAS database The Rotterdam Study is funded by Erasmus Medical Center andErasmus University Rotterdam the Netherlands Organisation for the Health Researchand Development (ZonMw) the Research Institute for Diseases in the Elderly (RIDE)the Ministry of Education Culture and Science the Ministry for Health Welfare andSports the European Commission (DG XII) and the Municipality of Rotterdam Theauthors are very grateful to the participants and staff from the Rotterdam Study theparticipating general practitioners and the pharmacists We would like to thank DrTobias A Knoch Luc V de Zeeuw Anis Abuseiris and Rob de Graaf as well as theirinstitutions the Erasmus Computing Grid Rotterdam The Netherlands and especiallythe national German MediGRID and ServicesMediGRID part of the German D-Gridboth funded by the German Bundesministerium fuer Forschung und Technology undergrants 01 AK 803 A-H and 01 IG 07015 G for access to their grid resourcesThe GARP study was supported by the Leiden University Medical Centre and theDutch Arthritis Association Pfizer Inc Groton CT USA supported the inclusion ofthe GARP study The genotypic work was supported by the Netherlands Organisationof Scientific Research (MW 904-61-095 911-03-016 917 66344 and 911-03-012)Leiden University Medical Centre and the Centre of Medical System Biology and theNetherlands Consortium for Healthy Aging both in the framework of the NetherlandsGenomics Initiative (NGI) The research leading to these studies has received fundingfrom the European Unionrsquos Seventh Framework Programme (FP72007-2011) undergrant agreement ndeg 259679 Pradeep Suri was funded by K12HD01097 from theNational Institute of Health (US) and the Rehabilitation Medicine Scientist TrainingProgramme (RMSTP) The Framingham Study is supported by the National HeartLung and Blood Institutersquos Framingham Heart Study contract (No N01-HC-25195) forthe recruitment enrolment and examination of the Offspring and Third Generationcohorts and the imaging by CT scan

Contributors The authors listed fulfilled the criteria for authorship

Funding Fellowship to FMKW who led the project

Competing interests None

Ethics approval All bodies appropriate to the groups contributing data

Provenance and peer review Not commissioned externally peer reviewed

Open Access This is an Open Access article distributed in accordance with theCreative Commons Attribution Non Commercial (CC BY-NC 30) license whichpermits others to distribute remix adapt build upon this work non-commercially andlicense their derivative works on different terms provided the original work is

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1147

Clinical and epidemiological research

properly cited and the use is non-commercial See httpcreativecommonsorglicensesby-nc30

REFERENCES1 Powell MC Wilson M Szypryt P et al Prevalence of lumbar disc degeneration

observed by magnetic resonance in symptomless women Lancet 198621366ndash72 de Schepper EI Damen J van Meurs JB et al The association between lumbar

disc degeneration and low back pain the influence of age gender and individualradiographic features Spine (Phila Pa 1976) 201035531ndash6

3 MacGregor AJ Andrew T Sambrook PN et al Structural psychological andgenetic influences on low back and neck pain a study of adult female twinsArthritis Rheum 200451160ndash7

4 Phillips CJ Economic burden of chronic pain Expert Rev Pharmacoecon OutcomesRes 20066591ndash601

5 Roberts S Evans H Trivedi J et al Histology and pathology of the humanintervertebral disc J Bone Joint Surg Am 200688(Suppl 2)10ndash14

6 Battie MC Levalahti E Videman T et al Heritability of lumbar flexibility and therole of disc degeneration and body weight J Appl Physiol 2008104379ndash85

7 Sambrook PN MacGregor AJ Spector TD Genetic influences on cervical andlumbar disc degeneration a magnetic resonance imaging study in twins ArthritisRheum 199942366ndash72

8 Ryder JJ Garrison K Song F et al Genetic associations in peripheral jointosteoarthritis and spinal degenerative disease a systematic review Ann Rheum Dis200767584ndash91

9 Valdes AM Loughlin J Oene MV et al Sex and ethnic differences in theassociation of ASPN CALM1 COL2A1 COMP and FRZB with genetic susceptibilityto osteoarthritis of the knee Arthritis Rheum 200656137ndash46

10 Richards JB Kavvoura FK Rivadeneira F et al Collaborative meta-analysisassociations of 150 candidate genes with osteoporosis and osteoporotic fractureAnn Intern Med 2009151528ndash37

11 Manolio TA Collins FS Cox NJ et al Finding the missing heritability of complexdiseases Nature 2009461747ndash53

12 Hoffmann U Siebert U Bull-Stewart A et al Evidence for lower variability ofcoronary artery calcium mineral mass measurements by multi-detector computedtomography in a community-based cohortndashconsequences for progression studiesEur J Radiol 200657396ndash402

13 Riyazi N Meulenbelt I Kroon HM et al Evidence for familial aggregation of handhip and spine but not knee osteoarthritis in siblings with multiple joint involvementthe GARP study Ann Rheum Dis 200564438ndash43

14 Lane NE Nevitt MC Genant HK et al Reliability of new indices of radiographicosteoarthritis of the hand and hip and lumbar disc degeneration J Rheumatol1993201911ndash18

15 Hofman A Breteler MM van Duijn CM The Rotterdam Study 2010 objectives anddesign update Eur J Epidemiol 200924553ndash72

16 Bijkerk C Houwing-Duistermaat JJ Valkenburg HA et al Heritabilities of radiologicosteoarthritis in peripheral joints and of disc degeneration of the spine ArthritisRheum 1999421729ndash35

17 Spector TD Williams FM The UK Adult Twin Registry (TwinsUK) Twin Res HumGenet 20069899ndash906

18 Suchindran S Rivedal D Guyton JR et al Genome-wide association study of Lp-PLA(2) activity and mass in the Framingham Heart Study PLoS Genet 20106e1000928

19 Howie BN Donnelly P Marchini J A flexible and accurate genotype imputation method forthe next generation of genome-wide association studies PLoS Genet 20095e1000529

20 Marchini J Howie B Myers S et al A new multipoint method for genome-wideassociation studies by imputation of genotypes Nat Genet 200739906ndash13

21 Uh HW Wijk HJ Houwing-Duistermaat JJ Testing for genetic associationtaking into account phenotypic information of relatives BMC Proc 20093(Suppl 7)S123

22 Richards JB Rivadeneira F Inouye M et al Bone mineral density osteoporosisand osteoporotic fractures a genome-wide association study Lancet20083711505ndash12

23 Stata for Windows Version 10 edn College Station Texas USA Statacorp 200724 Pritchard JK Stephens M Donnelly P Inference of population structure using

multilocus genotype data Genetics 2000155945ndash59

25 Aulchenko YS Struchalin MV van Duijn CM ProbABEL package for genome-wideassociation analysis of imputed data BMC Bioinformatics 201011134

26 Li Y Willer CJ Ding J et al MaCH using sequence and genotype data to estimatehaplotypes and unobserved genotypes Genet Epidemiol 201034816ndash34

27 Bell JT Tsai PC Yang TP et al Epigenome-wide scans identify differentiallymethylated regions for age and age-related phenotypes in a healthy ageingpopulation PLoS Genet 20128e1002629

28 Bell JT Pai AA Pickrell JK et al DNA methylation patterns associate withgenetic and gene expression variation in HapMap cell lines Genome Biol 201112R10

29 Devlin B Roeder K Genomic control for association studies Biometrics199955997ndash1004

30 Liuke M Solovieva S Lamminen A et al Disc degeneration of the lumbar spine inrelation to overweight IntJObes (Lond) 200529903ndash8

31 Pye SR Reid DM Adams JE et al Influence of weight body mass index andlifestyle factors on radiographic features of lumbar disc degeneration Ann RheumDis 200766426ndash7

32 Symmons DP van Hemert AM Vandenbroucke JP et al A longitudinal study ofback pain and radiological changes in the lumbar spines of middle aged women IIRadiographic findings Ann Rheum Dis 199150162ndash6

33 Williams FM Popham M Livshits G et al A response to Videman et allsquochallenging the cumulative injury model positive effects of greater body mass ondisc degenerationrsquo Spine J 201010571ndash2

34 Yoshimura N Dennison E Wilman C et al Epidemiology of chronic discdegeneration and osteoarthritis of the lumbar spine in Britain and Japan acomparative study J Rheumatol 200027429ndash33

35 Battie MC Videman T Gill K et al 1991 Volvo Award in clinical sciences Smokingand lumbar intervertebral disc degeneration an MRI study of identical twins Spine1991161015ndash21

36 Videman T Battie MC The influence of occupation on lumbar degeneration Spine1999241164ndash8

37 Williams FM Sambrook PN Neck and back pain and intervertebral discdegeneration role of occupational factors Best Pract Res Clin Rheumatol20112569ndash79

38 Tanaka A Parkin-mediated selective mitochondrial autophagy mitophagy Parkinpurges damaged organelles from the vital mitochondrial network FEBS Lett20105841386ndash92

39 Kay DM Stevens CF Hamza TH et al A comprehensive analysis of deletionsmultiplications and copy number variations in PARK2 Neurology 2010751189ndash94

40 Gao X Ning Y Cancer and Parkinsons disease the odd couple Drugs Today (Barc)201147215ndash22

41 Kang JD Georgescu HI McIntyre-Larkin L et al Herniated lumbar intervertebraldiscs spontaneously produce matrix metalloproteinases nitric oxide interleukin-6and prostaglandin E2 Spine (Phila Pa 1976) 199621271ndash7

42 Korhonen T Karppinen J Paimela L et al The treatment of disc-herniation-inducedsciatica with infliximab one-year follow-up results of FIRST II a randomizedcontrolled trial Spine (Phila Pa 1976) 2006312759ndash66

43 Videman T Saarela J Kaprio J et al Associations of 25 structural degradativeand inflammatory candidate genes with lumbar disc desiccation bulging and heightnarrowing Arthritis Rheum 200960470ndash81

44 Williams FM Kato BS Livshits G et al Lumbar disc disease shows linkage tochromosome 19 overlapping with a QTL for hand OA Ann Rheum Dis200867117ndash19

45 Demissie S Cupples LA Myers R et al Genome scan for quantity of handosteoarthritis the Framingham Study Arthritis Rheum 200246946ndash52

46 Brandt KD Radin EL Dieppe PA et al Yet more evidence that osteoarthritis is nota cartilage disease Ann Rheum Dis 2006651261ndash4

47 Amos CI Successful design and conduct of genome-wide association studies HumMol Genet 200716R220ndashR225

48 Zhai G van Meurs JB Livshits G et al A genome-wide association studysuggests that a locus within the ataxin 2 binding protein 1 gene is associatedwith hand osteoarthritis the Treat-OA consortium J Med Genet 200946614ndash16

49 Valdes AM Spector TD Genetic epidemiology of hip and knee osteoarthritisNat Rev Rheumatol 2011723ndash32

1148 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

Supplementary Table 1 Genotyping and imputation methods by study

FHS GARP RS1 RS3 TUK

Genotyping

Platform Affymetrix Illumina Illumina Illumina Illumina

Chip 500K or 100K or 10 K Human660W

quad Human Hap550 Human Hap550

Human Hap300 or Human

Hap610Q

Call rate gt097 gt 095 gt 0975 gt 0975 gt090

MAF gt 005 gt 00025 gt 001 gt 001 gt001

HWE p-value gt 10-6

gt 10-6

gt 10-6

gt 10-6

gt 10-4

Imputation

Variants included for imputation 534982 462361 512349 514073 314075

Imputation software MACH vs1 IMPUTE MACH MACH IMPUTE vs 2

Genome build

HapMap Phase II

release 22

Hapmap Phase II

release 21

Hapmap Phase II

release 22

Hapmap Phase II

release 22

Hapmap Phase II

release 36

Total number of SNPs available 2543887 1830498 2543887 2543887 3044064

Legend to Supplementary Table 1

FHS represents Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study RS1 Rotterdam study cohort 1 RS3 Rotterdam study

cohort 3 TUK TwinsUK BMI body mass index MAF Minor Allele Frequency HWE Hardy-Weinberg equilibrium

Supplementary Table 2 Results of the GWA meta-analysis adjusted for age and sex showing those SNPs having plt10-5

MAF Imputation Quality

SNP Chr gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP

Eff

All BETA SE P

rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C -012 0023 950E-08

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 019 0037 367E-07

rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 019 0039 663E-07

rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 019 0039 664E-07

rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A -028 0058 730E-07

rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A -011 0022 742E-07

rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C -025 0051 135E-06

rs11136566 8 CSMD1 4610 039 039 036 039 046 099 098 097 096 9739 A -01 0021 224E-06

rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 016 0035 307E-06

rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 016 0035 307E-06

rs365962 16 NA 4608 044 044 048 047 042 099 099 096 069 9861 C -01 0020 310E-06

rs387953 16 NA 4683 046 045 048 046 043 1 099 1 067 9999 G -009 0020 337E-06

rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A -011 0023 357E-06

rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A -013 0027 376E-06

rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C -012 0026 468E-06

rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0045 477E-06

rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 016 0035 497E-06

rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 016 0035 593E-06

rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 016 0035 593E-06

rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A -01 0023 624E-06

rs7234567 18 NA 4682 026 026 022 022 029 1 1 1 089 9767 A -01 0023 634E-06

rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 016 0035 652E-06

rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C -01 0022 654E-06

rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 01 0022 679E-06

rs10890236 1 NA 4683 009 009 009 009 009 1 1 1 093 100 C 016 0035 707E-06

rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 016 0037 708E-06

rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0022 728E-06

rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 015 0034 821E-06

rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 01 0022 823E-06

rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 015 0035 873E-06

rs7205409 16 RAB40C 4590 042 043 045 043 047 099 099 096 044 9933 C 009 0021 878E-06

rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A -047 0107 970E-06

Legend to Supplementary Table 2

Studies contributing data are denoted RS1 Rotterdam study cohort 1 RS3 Rotterdam study cohort 3 TUK TwinsUK BMI body mass index FHS

Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study

SNP single nucleotide polymorphism Chr chromosome position SNP location in base pairs MAF minor allele frequency

R2 from MACH for RS1 RS3 FHS and GARP Information score from IMPUTE for TUK

Eff All effect allele beta effect size SE standard error of beta p p value

Supplementary Figure 1 Quantile-quantile plot of meta-analysis adjusted GWA results

Legend to Supplementary Figure 1

The plots show GWA meta-analysis quantile-quantile plot of observed against expected results adjusted for age and sex

Supplementary Figure 2 Manhatten plot for meta-analysis of adjusted GWA results

Legend to Supplementary Figure 2

Plot shows combined results for the 5 studies included in the meta-analysis adjusted for age and sex The blue and red horizontal lines mark the levels of

suggestive and likely significance respectively

Supplementary Figure 3 Regional plot of association results and recombination rates for the class II MHC region unadjusted for age and sex

Legend to Supplementary Figure 3

minuslog10 P values (y axis) of the SNPs are shown according to their chromosomal positions (x axis) with lead SNP shown as a purple diamond The colour

intensity of each symbol depicting a SNP reflects the extent of LD with the rs926849 coloured red (r2 gt 08) through to blue (r

2 lt 02) Genetic

recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build 36 (NCBI) of the

human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn to show the relative

positions and therefore the maps are not to physical scale

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 6: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

expressed in solid organs as well as skeletal muscle (httpwwwproteinatlasorg) Mutations within PARK2 are associated withdiverse conditions including autosomal recessive juvenileParkinsonrsquos disease Alzheimer rsquos disease diabetes mellitus andseveral solid tumours (reviewed in39) Parkin may account for theinverse relation between Parkinsonrsquos disease and cancer inci-dence40 Our findings of hypermethylation with increasing LDDscore suggest that PARK2 expression is reduced with increasingdisc degeneration but functional studies of intervertebral disc andother spine tissues are needed

Three further markers in the unadjusted meta-analysis hadplt5times10minus8 Marker rs17034687 is an intergenic marker onchromosome 3 Based on One Thousand Genomes (1KG)CEUdata it is not in LD (r2gt 03) with any known gene-basedmarkers Markers rs2187689 (supplementary figure S2) andrs7767277 are HLA region markers neither of which is includedin the 1KG pilot data Using data from HapMap V3 (release 2)rs2187689 and rs776277 are in perfect LD with each other andin LD (r2= 076) with an intronic marker in PSMB9 (prote-asome (prosome macropain) subunit β type 9 large

multifunctional peptidase 2) Proteasomes are distributedthroughout eukaroytic cells at high concentration and cleavepeptides in an ATPubiquitin-dependent process in a non-lysosomal pathway The gene is located in the class II region ofthe MHC Expression of the gene is induced by interferon γand this gene product replaces catalytic subunit 1 (proteasomeβ 6 subunit) in the immunoproteasome

While lumbar degeneration is not considered an inflamma-tory process and has not been reported to be auto-immune inaetiology there is evidence of pro-inflammatory cytokine acti-vation in herniated lumbar discs41 and anti-TNF has been usedsuccessfully to treat disc herniation-induced sciatica42 Of notethe COL11A2 gene lies 169 KB upstream from rs2187689 AnSNP (rs2076311) within this candidate gene has been shown tobe associated with MR determined disc signal intensity in acandidate gene study of Finnish male twins43 SNP rs2076311is not however in LD with our top hit rs2187689 (r2=0017)and so it seems unlikely that this collagen-encoding geneaccounts for our observed association Many published GWAstudies have identified SNPs in intergenic regions and the

Figure 4 Regional plot of association results and recombination rates for the PARK2 gene unadjusted results minuslog10 p values (y-axis) of the singlenucleotide polymorphisms (SNPs) are shown according to their chromosomal positions (x-axis) with lead SNP shown as a purple diamond Thecolour intensity of each symbol depicting an SNP reflects the extent of LD with the rs926849 coloured red (r2gt08) through to blue (r2lt02)Genetic recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build36 (NCBI) of the human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn toshow the relative positions and therefore the maps are not to physical scale

Figure 3 Forest plot of rs926849 in PARK2 unadjusted for covariates The contribution of the studies included in the meta-analysis is shown in thisfixed effects model The C allele is considered Heterogeneity I2=0 p(Q)=067 TE treatment effect seTE standard error treatment effect

1146 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

precise role of these regions is yet to be defined Long rangeenhancers for example could operate here and so an influenceon COL11A2 expression cannot be ruled out

Of suggestive significance is SNP rs4802666 (p=376times10minus6adjusted meta-analysis) which lies within the MYH14 genewhich encodes myosin heavy chain 14 non-muscle It isexpressed in cell lines derived from bone (httpwwwproteinatlasorg) and is implicated in autosomal dominanthearing impairment It is of interest in LDD because it lies onchromosome 19 under the linkage peak we have reported intwins for LDD44 and a peak reported by the Framinghamgroup for hand OA45 As there is a known relationship betweenthese two phenotypes this region on chromosome 19 forms ahighly plausible candidate region for OA It is not impossiblethat a muscle-expressed protein plays a role in LDD throughmechanisms similar to those proposed for OA considered bysome to be a multi-tissue syndrome rather than simply adisease of cartilage46

The main limitation of the study is one of obtaining anaccurate phenotype on individuals which is known to be animportant factor in the success of GWA47 There is at presentno agreed gold standard imaging method in the determinationof LDD although it is recognised that MRI offers the most sen-sitive widely available tool Even so MR is relatively expensiveand many of the largest spine cohorts in the world have plainradiographs which offer more limited phenotypic informationThe coding method applied to the imaging is also yet to be for-mally standardised our interest in the individual subtraits ofLDD led us to devise a coding method in which they wereseparated as reported previously7 In order to obtain sufficientsample size a number of cohorts contributed having differentimaging methods but traits were selected to enable comparisonacross the cohorts Thus study groups recoded their imagingwhere necessary to meet uniform requirements for inclusionWe included measures of disc height (coded 0ndash3) and anteriorosteophytes in RS1 RS3 GARP and TUK (also coded 0ndash3) andposterior osteophytes in FHS (coded 0ndash3) These subpheno-types were summed over the five discs and underwent inversenormal transformation to give a normal distribution A furtherlimitation is that four cohorts are population samples whileGARP is derived from OA-affected sibling pairs We includedGARP because it has made a contribution to similar analysesperformed for OA48 and with adjustment for relatedness pro-vides data comparable with other studies While the differingmethods of imaging provide different amounts of informationso the LDD variable has lower mean in those cohorts withradiographs the variance is comparable Where GARP samplesmade a contribution to the meta-analysis (a number of thesignificant SNPs did not include a contribution from GARPtable 2) the minor allele frequency was similar to those ofother groups The TUK group has a disproportionate numberof women for historical reasons The men were retainedhowever as they did not differ significantly from women inthe LDD variable or body mass index (data not shown) Thisstudy lacks a replication group A second sample of similar sizeto the first is considered important to show that the findingsof the first sample are true positives Unfortunately there areto our knowledge no other collections of Northern Europeanshaving spine imaging which together would approach oursample size There is considerable evidence in the literaturethat the genetic predisposition between Northern Europeansand Asians to OA is different49 and given the similaritiesbetween OA and LDD we felt replication should be made inNorthern Europeans We elected to include all the subjects in

a single powerful study rather than split the sample andreduce the chances of finding significant novel loci associatedwith LDD

In conclusion this is the first large-scale GWA study of LDDand we have identified several novel variants in the PARK2gene and in PSMB9 within MHC class 2 We have shown in asmall subset that methylation at one of the PARK2 promotersis associated with MRI determined LDD Both loci meritfurther investigation to shed light on the important clinicalendpoint of low back pain

Author affiliations1Department Twin Research and Genetic Epidemiology Kingrsquos College LondonLondon UK2Acclarogen Ltd St Johnrsquos Innovation Centre Cambridge UK3Department of Internal Medicine Erasmus MC Rotterdam The Netherlands4The Netherlands Genomics Initiative sponsored Netherlands Consortium for HealthyAging (NGI-NCHA) LeidenRotterdam The Netherlands5Department of Molecular Epidemiology Leiden University Medical Center LeidenRotterdam The Netherlands6VA Boston Healthcare System Harvard Medical School Spaulding RehabilitationHospital New England Baptist Hospital Boston USA7Department of Rheumatology University of Sydney and Royal North Shore HospitalSydney Australia8Department of Epidemiology Erasmus MC Rotterdam The Netherlands9Department of General Practice Erasmus MC Rotterdam The Netherlands10Department of Rheumatology and Department of Clinical Epidemiology LeidenUniversity Medical Center Leiden The Netherlands

Acknowledgements FMKW is a Wellcome Trust Intermediate Fellow The authorswould like to acknowledge the contribution of Maria Popham to the collection andanalysis of TwinsUK data The GWA study was funded by the NetherlandsOrganisation of Scientific Research NWO Investments (nr 1750102005011911-03-012) the Research Institute for Diseases in the Elderly (014-93-015 RIDE2)and the Netherlands Genomics Initiative (NGI)the Netherlands Consortium forHealthy Aging (NCHA) project nr 050-060-810 We thank Pascal Arp Mila JhamaiDr Michael Moorhouse Marijn Verkerk and Sander Bervoets for their help in creatingthe GWAS database The Rotterdam Study is funded by Erasmus Medical Center andErasmus University Rotterdam the Netherlands Organisation for the Health Researchand Development (ZonMw) the Research Institute for Diseases in the Elderly (RIDE)the Ministry of Education Culture and Science the Ministry for Health Welfare andSports the European Commission (DG XII) and the Municipality of Rotterdam Theauthors are very grateful to the participants and staff from the Rotterdam Study theparticipating general practitioners and the pharmacists We would like to thank DrTobias A Knoch Luc V de Zeeuw Anis Abuseiris and Rob de Graaf as well as theirinstitutions the Erasmus Computing Grid Rotterdam The Netherlands and especiallythe national German MediGRID and ServicesMediGRID part of the German D-Gridboth funded by the German Bundesministerium fuer Forschung und Technology undergrants 01 AK 803 A-H and 01 IG 07015 G for access to their grid resourcesThe GARP study was supported by the Leiden University Medical Centre and theDutch Arthritis Association Pfizer Inc Groton CT USA supported the inclusion ofthe GARP study The genotypic work was supported by the Netherlands Organisationof Scientific Research (MW 904-61-095 911-03-016 917 66344 and 911-03-012)Leiden University Medical Centre and the Centre of Medical System Biology and theNetherlands Consortium for Healthy Aging both in the framework of the NetherlandsGenomics Initiative (NGI) The research leading to these studies has received fundingfrom the European Unionrsquos Seventh Framework Programme (FP72007-2011) undergrant agreement ndeg 259679 Pradeep Suri was funded by K12HD01097 from theNational Institute of Health (US) and the Rehabilitation Medicine Scientist TrainingProgramme (RMSTP) The Framingham Study is supported by the National HeartLung and Blood Institutersquos Framingham Heart Study contract (No N01-HC-25195) forthe recruitment enrolment and examination of the Offspring and Third Generationcohorts and the imaging by CT scan

Contributors The authors listed fulfilled the criteria for authorship

Funding Fellowship to FMKW who led the project

Competing interests None

Ethics approval All bodies appropriate to the groups contributing data

Provenance and peer review Not commissioned externally peer reviewed

Open Access This is an Open Access article distributed in accordance with theCreative Commons Attribution Non Commercial (CC BY-NC 30) license whichpermits others to distribute remix adapt build upon this work non-commercially andlicense their derivative works on different terms provided the original work is

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1147

Clinical and epidemiological research

properly cited and the use is non-commercial See httpcreativecommonsorglicensesby-nc30

REFERENCES1 Powell MC Wilson M Szypryt P et al Prevalence of lumbar disc degeneration

observed by magnetic resonance in symptomless women Lancet 198621366ndash72 de Schepper EI Damen J van Meurs JB et al The association between lumbar

disc degeneration and low back pain the influence of age gender and individualradiographic features Spine (Phila Pa 1976) 201035531ndash6

3 MacGregor AJ Andrew T Sambrook PN et al Structural psychological andgenetic influences on low back and neck pain a study of adult female twinsArthritis Rheum 200451160ndash7

4 Phillips CJ Economic burden of chronic pain Expert Rev Pharmacoecon OutcomesRes 20066591ndash601

5 Roberts S Evans H Trivedi J et al Histology and pathology of the humanintervertebral disc J Bone Joint Surg Am 200688(Suppl 2)10ndash14

6 Battie MC Levalahti E Videman T et al Heritability of lumbar flexibility and therole of disc degeneration and body weight J Appl Physiol 2008104379ndash85

7 Sambrook PN MacGregor AJ Spector TD Genetic influences on cervical andlumbar disc degeneration a magnetic resonance imaging study in twins ArthritisRheum 199942366ndash72

8 Ryder JJ Garrison K Song F et al Genetic associations in peripheral jointosteoarthritis and spinal degenerative disease a systematic review Ann Rheum Dis200767584ndash91

9 Valdes AM Loughlin J Oene MV et al Sex and ethnic differences in theassociation of ASPN CALM1 COL2A1 COMP and FRZB with genetic susceptibilityto osteoarthritis of the knee Arthritis Rheum 200656137ndash46

10 Richards JB Kavvoura FK Rivadeneira F et al Collaborative meta-analysisassociations of 150 candidate genes with osteoporosis and osteoporotic fractureAnn Intern Med 2009151528ndash37

11 Manolio TA Collins FS Cox NJ et al Finding the missing heritability of complexdiseases Nature 2009461747ndash53

12 Hoffmann U Siebert U Bull-Stewart A et al Evidence for lower variability ofcoronary artery calcium mineral mass measurements by multi-detector computedtomography in a community-based cohortndashconsequences for progression studiesEur J Radiol 200657396ndash402

13 Riyazi N Meulenbelt I Kroon HM et al Evidence for familial aggregation of handhip and spine but not knee osteoarthritis in siblings with multiple joint involvementthe GARP study Ann Rheum Dis 200564438ndash43

14 Lane NE Nevitt MC Genant HK et al Reliability of new indices of radiographicosteoarthritis of the hand and hip and lumbar disc degeneration J Rheumatol1993201911ndash18

15 Hofman A Breteler MM van Duijn CM The Rotterdam Study 2010 objectives anddesign update Eur J Epidemiol 200924553ndash72

16 Bijkerk C Houwing-Duistermaat JJ Valkenburg HA et al Heritabilities of radiologicosteoarthritis in peripheral joints and of disc degeneration of the spine ArthritisRheum 1999421729ndash35

17 Spector TD Williams FM The UK Adult Twin Registry (TwinsUK) Twin Res HumGenet 20069899ndash906

18 Suchindran S Rivedal D Guyton JR et al Genome-wide association study of Lp-PLA(2) activity and mass in the Framingham Heart Study PLoS Genet 20106e1000928

19 Howie BN Donnelly P Marchini J A flexible and accurate genotype imputation method forthe next generation of genome-wide association studies PLoS Genet 20095e1000529

20 Marchini J Howie B Myers S et al A new multipoint method for genome-wideassociation studies by imputation of genotypes Nat Genet 200739906ndash13

21 Uh HW Wijk HJ Houwing-Duistermaat JJ Testing for genetic associationtaking into account phenotypic information of relatives BMC Proc 20093(Suppl 7)S123

22 Richards JB Rivadeneira F Inouye M et al Bone mineral density osteoporosisand osteoporotic fractures a genome-wide association study Lancet20083711505ndash12

23 Stata for Windows Version 10 edn College Station Texas USA Statacorp 200724 Pritchard JK Stephens M Donnelly P Inference of population structure using

multilocus genotype data Genetics 2000155945ndash59

25 Aulchenko YS Struchalin MV van Duijn CM ProbABEL package for genome-wideassociation analysis of imputed data BMC Bioinformatics 201011134

26 Li Y Willer CJ Ding J et al MaCH using sequence and genotype data to estimatehaplotypes and unobserved genotypes Genet Epidemiol 201034816ndash34

27 Bell JT Tsai PC Yang TP et al Epigenome-wide scans identify differentiallymethylated regions for age and age-related phenotypes in a healthy ageingpopulation PLoS Genet 20128e1002629

28 Bell JT Pai AA Pickrell JK et al DNA methylation patterns associate withgenetic and gene expression variation in HapMap cell lines Genome Biol 201112R10

29 Devlin B Roeder K Genomic control for association studies Biometrics199955997ndash1004

30 Liuke M Solovieva S Lamminen A et al Disc degeneration of the lumbar spine inrelation to overweight IntJObes (Lond) 200529903ndash8

31 Pye SR Reid DM Adams JE et al Influence of weight body mass index andlifestyle factors on radiographic features of lumbar disc degeneration Ann RheumDis 200766426ndash7

32 Symmons DP van Hemert AM Vandenbroucke JP et al A longitudinal study ofback pain and radiological changes in the lumbar spines of middle aged women IIRadiographic findings Ann Rheum Dis 199150162ndash6

33 Williams FM Popham M Livshits G et al A response to Videman et allsquochallenging the cumulative injury model positive effects of greater body mass ondisc degenerationrsquo Spine J 201010571ndash2

34 Yoshimura N Dennison E Wilman C et al Epidemiology of chronic discdegeneration and osteoarthritis of the lumbar spine in Britain and Japan acomparative study J Rheumatol 200027429ndash33

35 Battie MC Videman T Gill K et al 1991 Volvo Award in clinical sciences Smokingand lumbar intervertebral disc degeneration an MRI study of identical twins Spine1991161015ndash21

36 Videman T Battie MC The influence of occupation on lumbar degeneration Spine1999241164ndash8

37 Williams FM Sambrook PN Neck and back pain and intervertebral discdegeneration role of occupational factors Best Pract Res Clin Rheumatol20112569ndash79

38 Tanaka A Parkin-mediated selective mitochondrial autophagy mitophagy Parkinpurges damaged organelles from the vital mitochondrial network FEBS Lett20105841386ndash92

39 Kay DM Stevens CF Hamza TH et al A comprehensive analysis of deletionsmultiplications and copy number variations in PARK2 Neurology 2010751189ndash94

40 Gao X Ning Y Cancer and Parkinsons disease the odd couple Drugs Today (Barc)201147215ndash22

41 Kang JD Georgescu HI McIntyre-Larkin L et al Herniated lumbar intervertebraldiscs spontaneously produce matrix metalloproteinases nitric oxide interleukin-6and prostaglandin E2 Spine (Phila Pa 1976) 199621271ndash7

42 Korhonen T Karppinen J Paimela L et al The treatment of disc-herniation-inducedsciatica with infliximab one-year follow-up results of FIRST II a randomizedcontrolled trial Spine (Phila Pa 1976) 2006312759ndash66

43 Videman T Saarela J Kaprio J et al Associations of 25 structural degradativeand inflammatory candidate genes with lumbar disc desiccation bulging and heightnarrowing Arthritis Rheum 200960470ndash81

44 Williams FM Kato BS Livshits G et al Lumbar disc disease shows linkage tochromosome 19 overlapping with a QTL for hand OA Ann Rheum Dis200867117ndash19

45 Demissie S Cupples LA Myers R et al Genome scan for quantity of handosteoarthritis the Framingham Study Arthritis Rheum 200246946ndash52

46 Brandt KD Radin EL Dieppe PA et al Yet more evidence that osteoarthritis is nota cartilage disease Ann Rheum Dis 2006651261ndash4

47 Amos CI Successful design and conduct of genome-wide association studies HumMol Genet 200716R220ndashR225

48 Zhai G van Meurs JB Livshits G et al A genome-wide association studysuggests that a locus within the ataxin 2 binding protein 1 gene is associatedwith hand osteoarthritis the Treat-OA consortium J Med Genet 200946614ndash16

49 Valdes AM Spector TD Genetic epidemiology of hip and knee osteoarthritisNat Rev Rheumatol 2011723ndash32

1148 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

Supplementary Table 1 Genotyping and imputation methods by study

FHS GARP RS1 RS3 TUK

Genotyping

Platform Affymetrix Illumina Illumina Illumina Illumina

Chip 500K or 100K or 10 K Human660W

quad Human Hap550 Human Hap550

Human Hap300 or Human

Hap610Q

Call rate gt097 gt 095 gt 0975 gt 0975 gt090

MAF gt 005 gt 00025 gt 001 gt 001 gt001

HWE p-value gt 10-6

gt 10-6

gt 10-6

gt 10-6

gt 10-4

Imputation

Variants included for imputation 534982 462361 512349 514073 314075

Imputation software MACH vs1 IMPUTE MACH MACH IMPUTE vs 2

Genome build

HapMap Phase II

release 22

Hapmap Phase II

release 21

Hapmap Phase II

release 22

Hapmap Phase II

release 22

Hapmap Phase II

release 36

Total number of SNPs available 2543887 1830498 2543887 2543887 3044064

Legend to Supplementary Table 1

FHS represents Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study RS1 Rotterdam study cohort 1 RS3 Rotterdam study

cohort 3 TUK TwinsUK BMI body mass index MAF Minor Allele Frequency HWE Hardy-Weinberg equilibrium

Supplementary Table 2 Results of the GWA meta-analysis adjusted for age and sex showing those SNPs having plt10-5

MAF Imputation Quality

SNP Chr gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP

Eff

All BETA SE P

rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C -012 0023 950E-08

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 019 0037 367E-07

rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 019 0039 663E-07

rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 019 0039 664E-07

rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A -028 0058 730E-07

rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A -011 0022 742E-07

rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C -025 0051 135E-06

rs11136566 8 CSMD1 4610 039 039 036 039 046 099 098 097 096 9739 A -01 0021 224E-06

rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 016 0035 307E-06

rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 016 0035 307E-06

rs365962 16 NA 4608 044 044 048 047 042 099 099 096 069 9861 C -01 0020 310E-06

rs387953 16 NA 4683 046 045 048 046 043 1 099 1 067 9999 G -009 0020 337E-06

rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A -011 0023 357E-06

rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A -013 0027 376E-06

rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C -012 0026 468E-06

rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0045 477E-06

rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 016 0035 497E-06

rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 016 0035 593E-06

rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 016 0035 593E-06

rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A -01 0023 624E-06

rs7234567 18 NA 4682 026 026 022 022 029 1 1 1 089 9767 A -01 0023 634E-06

rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 016 0035 652E-06

rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C -01 0022 654E-06

rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 01 0022 679E-06

rs10890236 1 NA 4683 009 009 009 009 009 1 1 1 093 100 C 016 0035 707E-06

rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 016 0037 708E-06

rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0022 728E-06

rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 015 0034 821E-06

rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 01 0022 823E-06

rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 015 0035 873E-06

rs7205409 16 RAB40C 4590 042 043 045 043 047 099 099 096 044 9933 C 009 0021 878E-06

rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A -047 0107 970E-06

Legend to Supplementary Table 2

Studies contributing data are denoted RS1 Rotterdam study cohort 1 RS3 Rotterdam study cohort 3 TUK TwinsUK BMI body mass index FHS

Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study

SNP single nucleotide polymorphism Chr chromosome position SNP location in base pairs MAF minor allele frequency

R2 from MACH for RS1 RS3 FHS and GARP Information score from IMPUTE for TUK

Eff All effect allele beta effect size SE standard error of beta p p value

Supplementary Figure 1 Quantile-quantile plot of meta-analysis adjusted GWA results

Legend to Supplementary Figure 1

The plots show GWA meta-analysis quantile-quantile plot of observed against expected results adjusted for age and sex

Supplementary Figure 2 Manhatten plot for meta-analysis of adjusted GWA results

Legend to Supplementary Figure 2

Plot shows combined results for the 5 studies included in the meta-analysis adjusted for age and sex The blue and red horizontal lines mark the levels of

suggestive and likely significance respectively

Supplementary Figure 3 Regional plot of association results and recombination rates for the class II MHC region unadjusted for age and sex

Legend to Supplementary Figure 3

minuslog10 P values (y axis) of the SNPs are shown according to their chromosomal positions (x axis) with lead SNP shown as a purple diamond The colour

intensity of each symbol depicting a SNP reflects the extent of LD with the rs926849 coloured red (r2 gt 08) through to blue (r

2 lt 02) Genetic

recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build 36 (NCBI) of the

human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn to show the relative

positions and therefore the maps are not to physical scale

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 7: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

precise role of these regions is yet to be defined Long rangeenhancers for example could operate here and so an influenceon COL11A2 expression cannot be ruled out

Of suggestive significance is SNP rs4802666 (p=376times10minus6adjusted meta-analysis) which lies within the MYH14 genewhich encodes myosin heavy chain 14 non-muscle It isexpressed in cell lines derived from bone (httpwwwproteinatlasorg) and is implicated in autosomal dominanthearing impairment It is of interest in LDD because it lies onchromosome 19 under the linkage peak we have reported intwins for LDD44 and a peak reported by the Framinghamgroup for hand OA45 As there is a known relationship betweenthese two phenotypes this region on chromosome 19 forms ahighly plausible candidate region for OA It is not impossiblethat a muscle-expressed protein plays a role in LDD throughmechanisms similar to those proposed for OA considered bysome to be a multi-tissue syndrome rather than simply adisease of cartilage46

The main limitation of the study is one of obtaining anaccurate phenotype on individuals which is known to be animportant factor in the success of GWA47 There is at presentno agreed gold standard imaging method in the determinationof LDD although it is recognised that MRI offers the most sen-sitive widely available tool Even so MR is relatively expensiveand many of the largest spine cohorts in the world have plainradiographs which offer more limited phenotypic informationThe coding method applied to the imaging is also yet to be for-mally standardised our interest in the individual subtraits ofLDD led us to devise a coding method in which they wereseparated as reported previously7 In order to obtain sufficientsample size a number of cohorts contributed having differentimaging methods but traits were selected to enable comparisonacross the cohorts Thus study groups recoded their imagingwhere necessary to meet uniform requirements for inclusionWe included measures of disc height (coded 0ndash3) and anteriorosteophytes in RS1 RS3 GARP and TUK (also coded 0ndash3) andposterior osteophytes in FHS (coded 0ndash3) These subpheno-types were summed over the five discs and underwent inversenormal transformation to give a normal distribution A furtherlimitation is that four cohorts are population samples whileGARP is derived from OA-affected sibling pairs We includedGARP because it has made a contribution to similar analysesperformed for OA48 and with adjustment for relatedness pro-vides data comparable with other studies While the differingmethods of imaging provide different amounts of informationso the LDD variable has lower mean in those cohorts withradiographs the variance is comparable Where GARP samplesmade a contribution to the meta-analysis (a number of thesignificant SNPs did not include a contribution from GARPtable 2) the minor allele frequency was similar to those ofother groups The TUK group has a disproportionate numberof women for historical reasons The men were retainedhowever as they did not differ significantly from women inthe LDD variable or body mass index (data not shown) Thisstudy lacks a replication group A second sample of similar sizeto the first is considered important to show that the findingsof the first sample are true positives Unfortunately there areto our knowledge no other collections of Northern Europeanshaving spine imaging which together would approach oursample size There is considerable evidence in the literaturethat the genetic predisposition between Northern Europeansand Asians to OA is different49 and given the similaritiesbetween OA and LDD we felt replication should be made inNorthern Europeans We elected to include all the subjects in

a single powerful study rather than split the sample andreduce the chances of finding significant novel loci associatedwith LDD

In conclusion this is the first large-scale GWA study of LDDand we have identified several novel variants in the PARK2gene and in PSMB9 within MHC class 2 We have shown in asmall subset that methylation at one of the PARK2 promotersis associated with MRI determined LDD Both loci meritfurther investigation to shed light on the important clinicalendpoint of low back pain

Author affiliations1Department Twin Research and Genetic Epidemiology Kingrsquos College LondonLondon UK2Acclarogen Ltd St Johnrsquos Innovation Centre Cambridge UK3Department of Internal Medicine Erasmus MC Rotterdam The Netherlands4The Netherlands Genomics Initiative sponsored Netherlands Consortium for HealthyAging (NGI-NCHA) LeidenRotterdam The Netherlands5Department of Molecular Epidemiology Leiden University Medical Center LeidenRotterdam The Netherlands6VA Boston Healthcare System Harvard Medical School Spaulding RehabilitationHospital New England Baptist Hospital Boston USA7Department of Rheumatology University of Sydney and Royal North Shore HospitalSydney Australia8Department of Epidemiology Erasmus MC Rotterdam The Netherlands9Department of General Practice Erasmus MC Rotterdam The Netherlands10Department of Rheumatology and Department of Clinical Epidemiology LeidenUniversity Medical Center Leiden The Netherlands

Acknowledgements FMKW is a Wellcome Trust Intermediate Fellow The authorswould like to acknowledge the contribution of Maria Popham to the collection andanalysis of TwinsUK data The GWA study was funded by the NetherlandsOrganisation of Scientific Research NWO Investments (nr 1750102005011911-03-012) the Research Institute for Diseases in the Elderly (014-93-015 RIDE2)and the Netherlands Genomics Initiative (NGI)the Netherlands Consortium forHealthy Aging (NCHA) project nr 050-060-810 We thank Pascal Arp Mila JhamaiDr Michael Moorhouse Marijn Verkerk and Sander Bervoets for their help in creatingthe GWAS database The Rotterdam Study is funded by Erasmus Medical Center andErasmus University Rotterdam the Netherlands Organisation for the Health Researchand Development (ZonMw) the Research Institute for Diseases in the Elderly (RIDE)the Ministry of Education Culture and Science the Ministry for Health Welfare andSports the European Commission (DG XII) and the Municipality of Rotterdam Theauthors are very grateful to the participants and staff from the Rotterdam Study theparticipating general practitioners and the pharmacists We would like to thank DrTobias A Knoch Luc V de Zeeuw Anis Abuseiris and Rob de Graaf as well as theirinstitutions the Erasmus Computing Grid Rotterdam The Netherlands and especiallythe national German MediGRID and ServicesMediGRID part of the German D-Gridboth funded by the German Bundesministerium fuer Forschung und Technology undergrants 01 AK 803 A-H and 01 IG 07015 G for access to their grid resourcesThe GARP study was supported by the Leiden University Medical Centre and theDutch Arthritis Association Pfizer Inc Groton CT USA supported the inclusion ofthe GARP study The genotypic work was supported by the Netherlands Organisationof Scientific Research (MW 904-61-095 911-03-016 917 66344 and 911-03-012)Leiden University Medical Centre and the Centre of Medical System Biology and theNetherlands Consortium for Healthy Aging both in the framework of the NetherlandsGenomics Initiative (NGI) The research leading to these studies has received fundingfrom the European Unionrsquos Seventh Framework Programme (FP72007-2011) undergrant agreement ndeg 259679 Pradeep Suri was funded by K12HD01097 from theNational Institute of Health (US) and the Rehabilitation Medicine Scientist TrainingProgramme (RMSTP) The Framingham Study is supported by the National HeartLung and Blood Institutersquos Framingham Heart Study contract (No N01-HC-25195) forthe recruitment enrolment and examination of the Offspring and Third Generationcohorts and the imaging by CT scan

Contributors The authors listed fulfilled the criteria for authorship

Funding Fellowship to FMKW who led the project

Competing interests None

Ethics approval All bodies appropriate to the groups contributing data

Provenance and peer review Not commissioned externally peer reviewed

Open Access This is an Open Access article distributed in accordance with theCreative Commons Attribution Non Commercial (CC BY-NC 30) license whichpermits others to distribute remix adapt build upon this work non-commercially andlicense their derivative works on different terms provided the original work is

Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551 1147

Clinical and epidemiological research

properly cited and the use is non-commercial See httpcreativecommonsorglicensesby-nc30

REFERENCES1 Powell MC Wilson M Szypryt P et al Prevalence of lumbar disc degeneration

observed by magnetic resonance in symptomless women Lancet 198621366ndash72 de Schepper EI Damen J van Meurs JB et al The association between lumbar

disc degeneration and low back pain the influence of age gender and individualradiographic features Spine (Phila Pa 1976) 201035531ndash6

3 MacGregor AJ Andrew T Sambrook PN et al Structural psychological andgenetic influences on low back and neck pain a study of adult female twinsArthritis Rheum 200451160ndash7

4 Phillips CJ Economic burden of chronic pain Expert Rev Pharmacoecon OutcomesRes 20066591ndash601

5 Roberts S Evans H Trivedi J et al Histology and pathology of the humanintervertebral disc J Bone Joint Surg Am 200688(Suppl 2)10ndash14

6 Battie MC Levalahti E Videman T et al Heritability of lumbar flexibility and therole of disc degeneration and body weight J Appl Physiol 2008104379ndash85

7 Sambrook PN MacGregor AJ Spector TD Genetic influences on cervical andlumbar disc degeneration a magnetic resonance imaging study in twins ArthritisRheum 199942366ndash72

8 Ryder JJ Garrison K Song F et al Genetic associations in peripheral jointosteoarthritis and spinal degenerative disease a systematic review Ann Rheum Dis200767584ndash91

9 Valdes AM Loughlin J Oene MV et al Sex and ethnic differences in theassociation of ASPN CALM1 COL2A1 COMP and FRZB with genetic susceptibilityto osteoarthritis of the knee Arthritis Rheum 200656137ndash46

10 Richards JB Kavvoura FK Rivadeneira F et al Collaborative meta-analysisassociations of 150 candidate genes with osteoporosis and osteoporotic fractureAnn Intern Med 2009151528ndash37

11 Manolio TA Collins FS Cox NJ et al Finding the missing heritability of complexdiseases Nature 2009461747ndash53

12 Hoffmann U Siebert U Bull-Stewart A et al Evidence for lower variability ofcoronary artery calcium mineral mass measurements by multi-detector computedtomography in a community-based cohortndashconsequences for progression studiesEur J Radiol 200657396ndash402

13 Riyazi N Meulenbelt I Kroon HM et al Evidence for familial aggregation of handhip and spine but not knee osteoarthritis in siblings with multiple joint involvementthe GARP study Ann Rheum Dis 200564438ndash43

14 Lane NE Nevitt MC Genant HK et al Reliability of new indices of radiographicosteoarthritis of the hand and hip and lumbar disc degeneration J Rheumatol1993201911ndash18

15 Hofman A Breteler MM van Duijn CM The Rotterdam Study 2010 objectives anddesign update Eur J Epidemiol 200924553ndash72

16 Bijkerk C Houwing-Duistermaat JJ Valkenburg HA et al Heritabilities of radiologicosteoarthritis in peripheral joints and of disc degeneration of the spine ArthritisRheum 1999421729ndash35

17 Spector TD Williams FM The UK Adult Twin Registry (TwinsUK) Twin Res HumGenet 20069899ndash906

18 Suchindran S Rivedal D Guyton JR et al Genome-wide association study of Lp-PLA(2) activity and mass in the Framingham Heart Study PLoS Genet 20106e1000928

19 Howie BN Donnelly P Marchini J A flexible and accurate genotype imputation method forthe next generation of genome-wide association studies PLoS Genet 20095e1000529

20 Marchini J Howie B Myers S et al A new multipoint method for genome-wideassociation studies by imputation of genotypes Nat Genet 200739906ndash13

21 Uh HW Wijk HJ Houwing-Duistermaat JJ Testing for genetic associationtaking into account phenotypic information of relatives BMC Proc 20093(Suppl 7)S123

22 Richards JB Rivadeneira F Inouye M et al Bone mineral density osteoporosisand osteoporotic fractures a genome-wide association study Lancet20083711505ndash12

23 Stata for Windows Version 10 edn College Station Texas USA Statacorp 200724 Pritchard JK Stephens M Donnelly P Inference of population structure using

multilocus genotype data Genetics 2000155945ndash59

25 Aulchenko YS Struchalin MV van Duijn CM ProbABEL package for genome-wideassociation analysis of imputed data BMC Bioinformatics 201011134

26 Li Y Willer CJ Ding J et al MaCH using sequence and genotype data to estimatehaplotypes and unobserved genotypes Genet Epidemiol 201034816ndash34

27 Bell JT Tsai PC Yang TP et al Epigenome-wide scans identify differentiallymethylated regions for age and age-related phenotypes in a healthy ageingpopulation PLoS Genet 20128e1002629

28 Bell JT Pai AA Pickrell JK et al DNA methylation patterns associate withgenetic and gene expression variation in HapMap cell lines Genome Biol 201112R10

29 Devlin B Roeder K Genomic control for association studies Biometrics199955997ndash1004

30 Liuke M Solovieva S Lamminen A et al Disc degeneration of the lumbar spine inrelation to overweight IntJObes (Lond) 200529903ndash8

31 Pye SR Reid DM Adams JE et al Influence of weight body mass index andlifestyle factors on radiographic features of lumbar disc degeneration Ann RheumDis 200766426ndash7

32 Symmons DP van Hemert AM Vandenbroucke JP et al A longitudinal study ofback pain and radiological changes in the lumbar spines of middle aged women IIRadiographic findings Ann Rheum Dis 199150162ndash6

33 Williams FM Popham M Livshits G et al A response to Videman et allsquochallenging the cumulative injury model positive effects of greater body mass ondisc degenerationrsquo Spine J 201010571ndash2

34 Yoshimura N Dennison E Wilman C et al Epidemiology of chronic discdegeneration and osteoarthritis of the lumbar spine in Britain and Japan acomparative study J Rheumatol 200027429ndash33

35 Battie MC Videman T Gill K et al 1991 Volvo Award in clinical sciences Smokingand lumbar intervertebral disc degeneration an MRI study of identical twins Spine1991161015ndash21

36 Videman T Battie MC The influence of occupation on lumbar degeneration Spine1999241164ndash8

37 Williams FM Sambrook PN Neck and back pain and intervertebral discdegeneration role of occupational factors Best Pract Res Clin Rheumatol20112569ndash79

38 Tanaka A Parkin-mediated selective mitochondrial autophagy mitophagy Parkinpurges damaged organelles from the vital mitochondrial network FEBS Lett20105841386ndash92

39 Kay DM Stevens CF Hamza TH et al A comprehensive analysis of deletionsmultiplications and copy number variations in PARK2 Neurology 2010751189ndash94

40 Gao X Ning Y Cancer and Parkinsons disease the odd couple Drugs Today (Barc)201147215ndash22

41 Kang JD Georgescu HI McIntyre-Larkin L et al Herniated lumbar intervertebraldiscs spontaneously produce matrix metalloproteinases nitric oxide interleukin-6and prostaglandin E2 Spine (Phila Pa 1976) 199621271ndash7

42 Korhonen T Karppinen J Paimela L et al The treatment of disc-herniation-inducedsciatica with infliximab one-year follow-up results of FIRST II a randomizedcontrolled trial Spine (Phila Pa 1976) 2006312759ndash66

43 Videman T Saarela J Kaprio J et al Associations of 25 structural degradativeand inflammatory candidate genes with lumbar disc desiccation bulging and heightnarrowing Arthritis Rheum 200960470ndash81

44 Williams FM Kato BS Livshits G et al Lumbar disc disease shows linkage tochromosome 19 overlapping with a QTL for hand OA Ann Rheum Dis200867117ndash19

45 Demissie S Cupples LA Myers R et al Genome scan for quantity of handosteoarthritis the Framingham Study Arthritis Rheum 200246946ndash52

46 Brandt KD Radin EL Dieppe PA et al Yet more evidence that osteoarthritis is nota cartilage disease Ann Rheum Dis 2006651261ndash4

47 Amos CI Successful design and conduct of genome-wide association studies HumMol Genet 200716R220ndashR225

48 Zhai G van Meurs JB Livshits G et al A genome-wide association studysuggests that a locus within the ataxin 2 binding protein 1 gene is associatedwith hand osteoarthritis the Treat-OA consortium J Med Genet 200946614ndash16

49 Valdes AM Spector TD Genetic epidemiology of hip and knee osteoarthritisNat Rev Rheumatol 2011723ndash32

1148 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

Supplementary Table 1 Genotyping and imputation methods by study

FHS GARP RS1 RS3 TUK

Genotyping

Platform Affymetrix Illumina Illumina Illumina Illumina

Chip 500K or 100K or 10 K Human660W

quad Human Hap550 Human Hap550

Human Hap300 or Human

Hap610Q

Call rate gt097 gt 095 gt 0975 gt 0975 gt090

MAF gt 005 gt 00025 gt 001 gt 001 gt001

HWE p-value gt 10-6

gt 10-6

gt 10-6

gt 10-6

gt 10-4

Imputation

Variants included for imputation 534982 462361 512349 514073 314075

Imputation software MACH vs1 IMPUTE MACH MACH IMPUTE vs 2

Genome build

HapMap Phase II

release 22

Hapmap Phase II

release 21

Hapmap Phase II

release 22

Hapmap Phase II

release 22

Hapmap Phase II

release 36

Total number of SNPs available 2543887 1830498 2543887 2543887 3044064

Legend to Supplementary Table 1

FHS represents Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study RS1 Rotterdam study cohort 1 RS3 Rotterdam study

cohort 3 TUK TwinsUK BMI body mass index MAF Minor Allele Frequency HWE Hardy-Weinberg equilibrium

Supplementary Table 2 Results of the GWA meta-analysis adjusted for age and sex showing those SNPs having plt10-5

MAF Imputation Quality

SNP Chr gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP

Eff

All BETA SE P

rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C -012 0023 950E-08

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 019 0037 367E-07

rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 019 0039 663E-07

rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 019 0039 664E-07

rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A -028 0058 730E-07

rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A -011 0022 742E-07

rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C -025 0051 135E-06

rs11136566 8 CSMD1 4610 039 039 036 039 046 099 098 097 096 9739 A -01 0021 224E-06

rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 016 0035 307E-06

rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 016 0035 307E-06

rs365962 16 NA 4608 044 044 048 047 042 099 099 096 069 9861 C -01 0020 310E-06

rs387953 16 NA 4683 046 045 048 046 043 1 099 1 067 9999 G -009 0020 337E-06

rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A -011 0023 357E-06

rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A -013 0027 376E-06

rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C -012 0026 468E-06

rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0045 477E-06

rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 016 0035 497E-06

rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 016 0035 593E-06

rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 016 0035 593E-06

rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A -01 0023 624E-06

rs7234567 18 NA 4682 026 026 022 022 029 1 1 1 089 9767 A -01 0023 634E-06

rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 016 0035 652E-06

rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C -01 0022 654E-06

rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 01 0022 679E-06

rs10890236 1 NA 4683 009 009 009 009 009 1 1 1 093 100 C 016 0035 707E-06

rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 016 0037 708E-06

rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0022 728E-06

rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 015 0034 821E-06

rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 01 0022 823E-06

rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 015 0035 873E-06

rs7205409 16 RAB40C 4590 042 043 045 043 047 099 099 096 044 9933 C 009 0021 878E-06

rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A -047 0107 970E-06

Legend to Supplementary Table 2

Studies contributing data are denoted RS1 Rotterdam study cohort 1 RS3 Rotterdam study cohort 3 TUK TwinsUK BMI body mass index FHS

Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study

SNP single nucleotide polymorphism Chr chromosome position SNP location in base pairs MAF minor allele frequency

R2 from MACH for RS1 RS3 FHS and GARP Information score from IMPUTE for TUK

Eff All effect allele beta effect size SE standard error of beta p p value

Supplementary Figure 1 Quantile-quantile plot of meta-analysis adjusted GWA results

Legend to Supplementary Figure 1

The plots show GWA meta-analysis quantile-quantile plot of observed against expected results adjusted for age and sex

Supplementary Figure 2 Manhatten plot for meta-analysis of adjusted GWA results

Legend to Supplementary Figure 2

Plot shows combined results for the 5 studies included in the meta-analysis adjusted for age and sex The blue and red horizontal lines mark the levels of

suggestive and likely significance respectively

Supplementary Figure 3 Regional plot of association results and recombination rates for the class II MHC region unadjusted for age and sex

Legend to Supplementary Figure 3

minuslog10 P values (y axis) of the SNPs are shown according to their chromosomal positions (x axis) with lead SNP shown as a purple diamond The colour

intensity of each symbol depicting a SNP reflects the extent of LD with the rs926849 coloured red (r2 gt 08) through to blue (r

2 lt 02) Genetic

recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build 36 (NCBI) of the

human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn to show the relative

positions and therefore the maps are not to physical scale

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 8: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

properly cited and the use is non-commercial See httpcreativecommonsorglicensesby-nc30

REFERENCES1 Powell MC Wilson M Szypryt P et al Prevalence of lumbar disc degeneration

observed by magnetic resonance in symptomless women Lancet 198621366ndash72 de Schepper EI Damen J van Meurs JB et al The association between lumbar

disc degeneration and low back pain the influence of age gender and individualradiographic features Spine (Phila Pa 1976) 201035531ndash6

3 MacGregor AJ Andrew T Sambrook PN et al Structural psychological andgenetic influences on low back and neck pain a study of adult female twinsArthritis Rheum 200451160ndash7

4 Phillips CJ Economic burden of chronic pain Expert Rev Pharmacoecon OutcomesRes 20066591ndash601

5 Roberts S Evans H Trivedi J et al Histology and pathology of the humanintervertebral disc J Bone Joint Surg Am 200688(Suppl 2)10ndash14

6 Battie MC Levalahti E Videman T et al Heritability of lumbar flexibility and therole of disc degeneration and body weight J Appl Physiol 2008104379ndash85

7 Sambrook PN MacGregor AJ Spector TD Genetic influences on cervical andlumbar disc degeneration a magnetic resonance imaging study in twins ArthritisRheum 199942366ndash72

8 Ryder JJ Garrison K Song F et al Genetic associations in peripheral jointosteoarthritis and spinal degenerative disease a systematic review Ann Rheum Dis200767584ndash91

9 Valdes AM Loughlin J Oene MV et al Sex and ethnic differences in theassociation of ASPN CALM1 COL2A1 COMP and FRZB with genetic susceptibilityto osteoarthritis of the knee Arthritis Rheum 200656137ndash46

10 Richards JB Kavvoura FK Rivadeneira F et al Collaborative meta-analysisassociations of 150 candidate genes with osteoporosis and osteoporotic fractureAnn Intern Med 2009151528ndash37

11 Manolio TA Collins FS Cox NJ et al Finding the missing heritability of complexdiseases Nature 2009461747ndash53

12 Hoffmann U Siebert U Bull-Stewart A et al Evidence for lower variability ofcoronary artery calcium mineral mass measurements by multi-detector computedtomography in a community-based cohortndashconsequences for progression studiesEur J Radiol 200657396ndash402

13 Riyazi N Meulenbelt I Kroon HM et al Evidence for familial aggregation of handhip and spine but not knee osteoarthritis in siblings with multiple joint involvementthe GARP study Ann Rheum Dis 200564438ndash43

14 Lane NE Nevitt MC Genant HK et al Reliability of new indices of radiographicosteoarthritis of the hand and hip and lumbar disc degeneration J Rheumatol1993201911ndash18

15 Hofman A Breteler MM van Duijn CM The Rotterdam Study 2010 objectives anddesign update Eur J Epidemiol 200924553ndash72

16 Bijkerk C Houwing-Duistermaat JJ Valkenburg HA et al Heritabilities of radiologicosteoarthritis in peripheral joints and of disc degeneration of the spine ArthritisRheum 1999421729ndash35

17 Spector TD Williams FM The UK Adult Twin Registry (TwinsUK) Twin Res HumGenet 20069899ndash906

18 Suchindran S Rivedal D Guyton JR et al Genome-wide association study of Lp-PLA(2) activity and mass in the Framingham Heart Study PLoS Genet 20106e1000928

19 Howie BN Donnelly P Marchini J A flexible and accurate genotype imputation method forthe next generation of genome-wide association studies PLoS Genet 20095e1000529

20 Marchini J Howie B Myers S et al A new multipoint method for genome-wideassociation studies by imputation of genotypes Nat Genet 200739906ndash13

21 Uh HW Wijk HJ Houwing-Duistermaat JJ Testing for genetic associationtaking into account phenotypic information of relatives BMC Proc 20093(Suppl 7)S123

22 Richards JB Rivadeneira F Inouye M et al Bone mineral density osteoporosisand osteoporotic fractures a genome-wide association study Lancet20083711505ndash12

23 Stata for Windows Version 10 edn College Station Texas USA Statacorp 200724 Pritchard JK Stephens M Donnelly P Inference of population structure using

multilocus genotype data Genetics 2000155945ndash59

25 Aulchenko YS Struchalin MV van Duijn CM ProbABEL package for genome-wideassociation analysis of imputed data BMC Bioinformatics 201011134

26 Li Y Willer CJ Ding J et al MaCH using sequence and genotype data to estimatehaplotypes and unobserved genotypes Genet Epidemiol 201034816ndash34

27 Bell JT Tsai PC Yang TP et al Epigenome-wide scans identify differentiallymethylated regions for age and age-related phenotypes in a healthy ageingpopulation PLoS Genet 20128e1002629

28 Bell JT Pai AA Pickrell JK et al DNA methylation patterns associate withgenetic and gene expression variation in HapMap cell lines Genome Biol 201112R10

29 Devlin B Roeder K Genomic control for association studies Biometrics199955997ndash1004

30 Liuke M Solovieva S Lamminen A et al Disc degeneration of the lumbar spine inrelation to overweight IntJObes (Lond) 200529903ndash8

31 Pye SR Reid DM Adams JE et al Influence of weight body mass index andlifestyle factors on radiographic features of lumbar disc degeneration Ann RheumDis 200766426ndash7

32 Symmons DP van Hemert AM Vandenbroucke JP et al A longitudinal study ofback pain and radiological changes in the lumbar spines of middle aged women IIRadiographic findings Ann Rheum Dis 199150162ndash6

33 Williams FM Popham M Livshits G et al A response to Videman et allsquochallenging the cumulative injury model positive effects of greater body mass ondisc degenerationrsquo Spine J 201010571ndash2

34 Yoshimura N Dennison E Wilman C et al Epidemiology of chronic discdegeneration and osteoarthritis of the lumbar spine in Britain and Japan acomparative study J Rheumatol 200027429ndash33

35 Battie MC Videman T Gill K et al 1991 Volvo Award in clinical sciences Smokingand lumbar intervertebral disc degeneration an MRI study of identical twins Spine1991161015ndash21

36 Videman T Battie MC The influence of occupation on lumbar degeneration Spine1999241164ndash8

37 Williams FM Sambrook PN Neck and back pain and intervertebral discdegeneration role of occupational factors Best Pract Res Clin Rheumatol20112569ndash79

38 Tanaka A Parkin-mediated selective mitochondrial autophagy mitophagy Parkinpurges damaged organelles from the vital mitochondrial network FEBS Lett20105841386ndash92

39 Kay DM Stevens CF Hamza TH et al A comprehensive analysis of deletionsmultiplications and copy number variations in PARK2 Neurology 2010751189ndash94

40 Gao X Ning Y Cancer and Parkinsons disease the odd couple Drugs Today (Barc)201147215ndash22

41 Kang JD Georgescu HI McIntyre-Larkin L et al Herniated lumbar intervertebraldiscs spontaneously produce matrix metalloproteinases nitric oxide interleukin-6and prostaglandin E2 Spine (Phila Pa 1976) 199621271ndash7

42 Korhonen T Karppinen J Paimela L et al The treatment of disc-herniation-inducedsciatica with infliximab one-year follow-up results of FIRST II a randomizedcontrolled trial Spine (Phila Pa 1976) 2006312759ndash66

43 Videman T Saarela J Kaprio J et al Associations of 25 structural degradativeand inflammatory candidate genes with lumbar disc desiccation bulging and heightnarrowing Arthritis Rheum 200960470ndash81

44 Williams FM Kato BS Livshits G et al Lumbar disc disease shows linkage tochromosome 19 overlapping with a QTL for hand OA Ann Rheum Dis200867117ndash19

45 Demissie S Cupples LA Myers R et al Genome scan for quantity of handosteoarthritis the Framingham Study Arthritis Rheum 200246946ndash52

46 Brandt KD Radin EL Dieppe PA et al Yet more evidence that osteoarthritis is nota cartilage disease Ann Rheum Dis 2006651261ndash4

47 Amos CI Successful design and conduct of genome-wide association studies HumMol Genet 200716R220ndashR225

48 Zhai G van Meurs JB Livshits G et al A genome-wide association studysuggests that a locus within the ataxin 2 binding protein 1 gene is associatedwith hand osteoarthritis the Treat-OA consortium J Med Genet 200946614ndash16

49 Valdes AM Spector TD Genetic epidemiology of hip and knee osteoarthritisNat Rev Rheumatol 2011723ndash32

1148 Ann Rheum Dis 2013721141ndash1148 doi101136annrheumdis-2012-201551

Clinical and epidemiological research

Supplementary Table 1 Genotyping and imputation methods by study

FHS GARP RS1 RS3 TUK

Genotyping

Platform Affymetrix Illumina Illumina Illumina Illumina

Chip 500K or 100K or 10 K Human660W

quad Human Hap550 Human Hap550

Human Hap300 or Human

Hap610Q

Call rate gt097 gt 095 gt 0975 gt 0975 gt090

MAF gt 005 gt 00025 gt 001 gt 001 gt001

HWE p-value gt 10-6

gt 10-6

gt 10-6

gt 10-6

gt 10-4

Imputation

Variants included for imputation 534982 462361 512349 514073 314075

Imputation software MACH vs1 IMPUTE MACH MACH IMPUTE vs 2

Genome build

HapMap Phase II

release 22

Hapmap Phase II

release 21

Hapmap Phase II

release 22

Hapmap Phase II

release 22

Hapmap Phase II

release 36

Total number of SNPs available 2543887 1830498 2543887 2543887 3044064

Legend to Supplementary Table 1

FHS represents Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study RS1 Rotterdam study cohort 1 RS3 Rotterdam study

cohort 3 TUK TwinsUK BMI body mass index MAF Minor Allele Frequency HWE Hardy-Weinberg equilibrium

Supplementary Table 2 Results of the GWA meta-analysis adjusted for age and sex showing those SNPs having plt10-5

MAF Imputation Quality

SNP Chr gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP

Eff

All BETA SE P

rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C -012 0023 950E-08

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 019 0037 367E-07

rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 019 0039 663E-07

rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 019 0039 664E-07

rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A -028 0058 730E-07

rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A -011 0022 742E-07

rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C -025 0051 135E-06

rs11136566 8 CSMD1 4610 039 039 036 039 046 099 098 097 096 9739 A -01 0021 224E-06

rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 016 0035 307E-06

rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 016 0035 307E-06

rs365962 16 NA 4608 044 044 048 047 042 099 099 096 069 9861 C -01 0020 310E-06

rs387953 16 NA 4683 046 045 048 046 043 1 099 1 067 9999 G -009 0020 337E-06

rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A -011 0023 357E-06

rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A -013 0027 376E-06

rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C -012 0026 468E-06

rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0045 477E-06

rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 016 0035 497E-06

rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 016 0035 593E-06

rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 016 0035 593E-06

rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A -01 0023 624E-06

rs7234567 18 NA 4682 026 026 022 022 029 1 1 1 089 9767 A -01 0023 634E-06

rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 016 0035 652E-06

rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C -01 0022 654E-06

rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 01 0022 679E-06

rs10890236 1 NA 4683 009 009 009 009 009 1 1 1 093 100 C 016 0035 707E-06

rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 016 0037 708E-06

rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0022 728E-06

rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 015 0034 821E-06

rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 01 0022 823E-06

rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 015 0035 873E-06

rs7205409 16 RAB40C 4590 042 043 045 043 047 099 099 096 044 9933 C 009 0021 878E-06

rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A -047 0107 970E-06

Legend to Supplementary Table 2

Studies contributing data are denoted RS1 Rotterdam study cohort 1 RS3 Rotterdam study cohort 3 TUK TwinsUK BMI body mass index FHS

Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study

SNP single nucleotide polymorphism Chr chromosome position SNP location in base pairs MAF minor allele frequency

R2 from MACH for RS1 RS3 FHS and GARP Information score from IMPUTE for TUK

Eff All effect allele beta effect size SE standard error of beta p p value

Supplementary Figure 1 Quantile-quantile plot of meta-analysis adjusted GWA results

Legend to Supplementary Figure 1

The plots show GWA meta-analysis quantile-quantile plot of observed against expected results adjusted for age and sex

Supplementary Figure 2 Manhatten plot for meta-analysis of adjusted GWA results

Legend to Supplementary Figure 2

Plot shows combined results for the 5 studies included in the meta-analysis adjusted for age and sex The blue and red horizontal lines mark the levels of

suggestive and likely significance respectively

Supplementary Figure 3 Regional plot of association results and recombination rates for the class II MHC region unadjusted for age and sex

Legend to Supplementary Figure 3

minuslog10 P values (y axis) of the SNPs are shown according to their chromosomal positions (x axis) with lead SNP shown as a purple diamond The colour

intensity of each symbol depicting a SNP reflects the extent of LD with the rs926849 coloured red (r2 gt 08) through to blue (r

2 lt 02) Genetic

recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build 36 (NCBI) of the

human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn to show the relative

positions and therefore the maps are not to physical scale

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 9: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

Supplementary Table 1 Genotyping and imputation methods by study

FHS GARP RS1 RS3 TUK

Genotyping

Platform Affymetrix Illumina Illumina Illumina Illumina

Chip 500K or 100K or 10 K Human660W

quad Human Hap550 Human Hap550

Human Hap300 or Human

Hap610Q

Call rate gt097 gt 095 gt 0975 gt 0975 gt090

MAF gt 005 gt 00025 gt 001 gt 001 gt001

HWE p-value gt 10-6

gt 10-6

gt 10-6

gt 10-6

gt 10-4

Imputation

Variants included for imputation 534982 462361 512349 514073 314075

Imputation software MACH vs1 IMPUTE MACH MACH IMPUTE vs 2

Genome build

HapMap Phase II

release 22

Hapmap Phase II

release 21

Hapmap Phase II

release 22

Hapmap Phase II

release 22

Hapmap Phase II

release 36

Total number of SNPs available 2543887 1830498 2543887 2543887 3044064

Legend to Supplementary Table 1

FHS represents Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study RS1 Rotterdam study cohort 1 RS3 Rotterdam study

cohort 3 TUK TwinsUK BMI body mass index MAF Minor Allele Frequency HWE Hardy-Weinberg equilibrium

Supplementary Table 2 Results of the GWA meta-analysis adjusted for age and sex showing those SNPs having plt10-5

MAF Imputation Quality

SNP Chr gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP

Eff

All BETA SE P

rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C -012 0023 950E-08

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 019 0037 367E-07

rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 019 0039 663E-07

rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 019 0039 664E-07

rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A -028 0058 730E-07

rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A -011 0022 742E-07

rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C -025 0051 135E-06

rs11136566 8 CSMD1 4610 039 039 036 039 046 099 098 097 096 9739 A -01 0021 224E-06

rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 016 0035 307E-06

rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 016 0035 307E-06

rs365962 16 NA 4608 044 044 048 047 042 099 099 096 069 9861 C -01 0020 310E-06

rs387953 16 NA 4683 046 045 048 046 043 1 099 1 067 9999 G -009 0020 337E-06

rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A -011 0023 357E-06

rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A -013 0027 376E-06

rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C -012 0026 468E-06

rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0045 477E-06

rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 016 0035 497E-06

rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 016 0035 593E-06

rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 016 0035 593E-06

rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A -01 0023 624E-06

rs7234567 18 NA 4682 026 026 022 022 029 1 1 1 089 9767 A -01 0023 634E-06

rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 016 0035 652E-06

rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C -01 0022 654E-06

rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 01 0022 679E-06

rs10890236 1 NA 4683 009 009 009 009 009 1 1 1 093 100 C 016 0035 707E-06

rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 016 0037 708E-06

rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0022 728E-06

rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 015 0034 821E-06

rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 01 0022 823E-06

rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 015 0035 873E-06

rs7205409 16 RAB40C 4590 042 043 045 043 047 099 099 096 044 9933 C 009 0021 878E-06

rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A -047 0107 970E-06

Legend to Supplementary Table 2

Studies contributing data are denoted RS1 Rotterdam study cohort 1 RS3 Rotterdam study cohort 3 TUK TwinsUK BMI body mass index FHS

Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study

SNP single nucleotide polymorphism Chr chromosome position SNP location in base pairs MAF minor allele frequency

R2 from MACH for RS1 RS3 FHS and GARP Information score from IMPUTE for TUK

Eff All effect allele beta effect size SE standard error of beta p p value

Supplementary Figure 1 Quantile-quantile plot of meta-analysis adjusted GWA results

Legend to Supplementary Figure 1

The plots show GWA meta-analysis quantile-quantile plot of observed against expected results adjusted for age and sex

Supplementary Figure 2 Manhatten plot for meta-analysis of adjusted GWA results

Legend to Supplementary Figure 2

Plot shows combined results for the 5 studies included in the meta-analysis adjusted for age and sex The blue and red horizontal lines mark the levels of

suggestive and likely significance respectively

Supplementary Figure 3 Regional plot of association results and recombination rates for the class II MHC region unadjusted for age and sex

Legend to Supplementary Figure 3

minuslog10 P values (y axis) of the SNPs are shown according to their chromosomal positions (x axis) with lead SNP shown as a purple diamond The colour

intensity of each symbol depicting a SNP reflects the extent of LD with the rs926849 coloured red (r2 gt 08) through to blue (r

2 lt 02) Genetic

recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build 36 (NCBI) of the

human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn to show the relative

positions and therefore the maps are not to physical scale

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 10: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

Supplementary Table 2 Results of the GWA meta-analysis adjusted for age and sex showing those SNPs having plt10-5

MAF Imputation Quality

SNP Chr gene N RS1 RS3 TUK FHS GARP RS1 RS3 TUK FHS GARP

Eff

All BETA SE P

rs926849 6 PARK2 3939 031 032 NA 031 023 098 099 NA 09 9504 C -012 0023 950E-08

rs17034687 3 NA 4429 009 01 005 008 NA 092 093 088 082 NA C 019 0037 367E-07

rs2187689 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA C 019 0039 663E-07

rs7767277 6 NA 4457 008 007 006 009 NA 098 099 095 093 NA A 019 0039 664E-07

rs9488238 6 NA 4464 004 003 002 003 NA 087 088 087 097 NA A -028 0058 730E-07

rs3019449 6 PARK2 4636 032 032 031 031 028 098 098 098 099 9718 A -011 0022 742E-07

rs9301951 13 GPC6 4397 004 004 003 005 NA 096 097 074 078 NA C -025 0051 135E-06

rs11136566 8 CSMD1 4610 039 039 036 039 046 099 098 097 096 9739 A -01 0021 224E-06

rs7744666 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA C 016 0035 307E-06

rs11969002 6 NA 4466 01 009 006 01 NA 099 1 097 096 NA A 016 0035 307E-06

rs365962 16 NA 4608 044 044 048 047 042 099 099 096 069 9861 C -01 0020 310E-06

rs387953 16 NA 4683 046 045 048 046 043 1 099 1 067 9999 G -009 0020 337E-06

rs737631 6 PARK2 4303 028 029 023 028 NA 092 091 089 083 NA A -011 0023 357E-06

rs4802666 19 MYH14 4094 027 027 018 027 NA 07 074 076 064 NA A -013 0027 376E-06

rs1154053 8 NA 4639 017 017 018 016 02 099 1 096 089 9617 C -012 0026 468E-06

rs1205863 6 NA 3939 006 006 NA 006 007 099 1 NA 099 9823 G 021 0045 477E-06

rs3749982 6 NA 4458 01 009 006 01 NA 099 1 096 096 NA A 016 0035 497E-06

rs6457690 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA A 016 0035 593E-06

rs1029296 6 NA 4464 01 009 007 011 NA 098 1 096 097 NA C 016 0035 593E-06

rs4875102 8 NA 4608 026 026 027 025 029 099 099 095 091 9765 A -01 0023 624E-06

rs7234567 18 NA 4682 026 026 022 022 029 1 1 1 089 9767 A -01 0023 634E-06

rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 016 0035 652E-06

rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C -01 0022 654E-06

rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 01 0022 679E-06

rs10890236 1 NA 4683 009 009 009 009 009 1 1 1 093 100 C 016 0035 707E-06

rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 016 0037 708E-06

rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0022 728E-06

rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 015 0034 821E-06

rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 01 0022 823E-06

rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 015 0035 873E-06

rs7205409 16 RAB40C 4590 042 043 045 043 047 099 099 096 044 9933 C 009 0021 878E-06

rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A -047 0107 970E-06

Legend to Supplementary Table 2

Studies contributing data are denoted RS1 Rotterdam study cohort 1 RS3 Rotterdam study cohort 3 TUK TwinsUK BMI body mass index FHS

Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study

SNP single nucleotide polymorphism Chr chromosome position SNP location in base pairs MAF minor allele frequency

R2 from MACH for RS1 RS3 FHS and GARP Information score from IMPUTE for TUK

Eff All effect allele beta effect size SE standard error of beta p p value

Supplementary Figure 1 Quantile-quantile plot of meta-analysis adjusted GWA results

Legend to Supplementary Figure 1

The plots show GWA meta-analysis quantile-quantile plot of observed against expected results adjusted for age and sex

Supplementary Figure 2 Manhatten plot for meta-analysis of adjusted GWA results

Legend to Supplementary Figure 2

Plot shows combined results for the 5 studies included in the meta-analysis adjusted for age and sex The blue and red horizontal lines mark the levels of

suggestive and likely significance respectively

Supplementary Figure 3 Regional plot of association results and recombination rates for the class II MHC region unadjusted for age and sex

Legend to Supplementary Figure 3

minuslog10 P values (y axis) of the SNPs are shown according to their chromosomal positions (x axis) with lead SNP shown as a purple diamond The colour

intensity of each symbol depicting a SNP reflects the extent of LD with the rs926849 coloured red (r2 gt 08) through to blue (r

2 lt 02) Genetic

recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build 36 (NCBI) of the

human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn to show the relative

positions and therefore the maps are not to physical scale

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 11: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

rs6936004 6 NA 4462 01 009 007 011 NA 098 1 096 097 NA C 016 0035 652E-06

rs1884158 6 PARK2 4491 031 032 031 031 NA 098 099 099 091 NA C -01 0022 654E-06

rs1981483 16 PIGQ 3939 042 043 NA 042 047 099 098 NA 044 9964 A 01 0022 679E-06

rs10890236 1 NA 4683 009 009 009 009 009 1 1 1 093 100 C 016 0035 707E-06

rs1029295 6 NA 3747 01 009 NA 011 NA 098 1 NA 097 NA C 016 0037 708E-06

rs763014 16 RAB40C 3939 042 043 NA 043 047 098 099 NA 042 994 C 01 0022 728E-06

rs10046257 6 NA 4461 01 009 008 011 NA 098 1 096 097 NA A 015 0034 821E-06

rs7204439 16 RAB40C 3939 042 043 NA 043 047 097 098 NA 043 9737 C 01 0022 823E-06

rs9469300 6 NA 4482 01 009 007 01 NA 099 1 092 096 NA A 015 0035 873E-06

rs7205409 16 RAB40C 4590 042 043 045 043 047 099 099 096 044 9933 C 009 0021 878E-06

rs10998466 10 NA 4446 001 001 001 002 NA 075 097 073 051 NA A -047 0107 970E-06

Legend to Supplementary Table 2

Studies contributing data are denoted RS1 Rotterdam study cohort 1 RS3 Rotterdam study cohort 3 TUK TwinsUK BMI body mass index FHS

Framingham Heart Study GARP Genetics of OsteoArthrosis and Progression study

SNP single nucleotide polymorphism Chr chromosome position SNP location in base pairs MAF minor allele frequency

R2 from MACH for RS1 RS3 FHS and GARP Information score from IMPUTE for TUK

Eff All effect allele beta effect size SE standard error of beta p p value

Supplementary Figure 1 Quantile-quantile plot of meta-analysis adjusted GWA results

Legend to Supplementary Figure 1

The plots show GWA meta-analysis quantile-quantile plot of observed against expected results adjusted for age and sex

Supplementary Figure 2 Manhatten plot for meta-analysis of adjusted GWA results

Legend to Supplementary Figure 2

Plot shows combined results for the 5 studies included in the meta-analysis adjusted for age and sex The blue and red horizontal lines mark the levels of

suggestive and likely significance respectively

Supplementary Figure 3 Regional plot of association results and recombination rates for the class II MHC region unadjusted for age and sex

Legend to Supplementary Figure 3

minuslog10 P values (y axis) of the SNPs are shown according to their chromosomal positions (x axis) with lead SNP shown as a purple diamond The colour

intensity of each symbol depicting a SNP reflects the extent of LD with the rs926849 coloured red (r2 gt 08) through to blue (r

2 lt 02) Genetic

recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build 36 (NCBI) of the

human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn to show the relative

positions and therefore the maps are not to physical scale

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 12: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

Supplementary Figure 1 Quantile-quantile plot of meta-analysis adjusted GWA results

Legend to Supplementary Figure 1

The plots show GWA meta-analysis quantile-quantile plot of observed against expected results adjusted for age and sex

Supplementary Figure 2 Manhatten plot for meta-analysis of adjusted GWA results

Legend to Supplementary Figure 2

Plot shows combined results for the 5 studies included in the meta-analysis adjusted for age and sex The blue and red horizontal lines mark the levels of

suggestive and likely significance respectively

Supplementary Figure 3 Regional plot of association results and recombination rates for the class II MHC region unadjusted for age and sex

Legend to Supplementary Figure 3

minuslog10 P values (y axis) of the SNPs are shown according to their chromosomal positions (x axis) with lead SNP shown as a purple diamond The colour

intensity of each symbol depicting a SNP reflects the extent of LD with the rs926849 coloured red (r2 gt 08) through to blue (r

2 lt 02) Genetic

recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build 36 (NCBI) of the

human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn to show the relative

positions and therefore the maps are not to physical scale

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 13: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

Supplementary Figure 2 Manhatten plot for meta-analysis of adjusted GWA results

Legend to Supplementary Figure 2

Plot shows combined results for the 5 studies included in the meta-analysis adjusted for age and sex The blue and red horizontal lines mark the levels of

suggestive and likely significance respectively

Supplementary Figure 3 Regional plot of association results and recombination rates for the class II MHC region unadjusted for age and sex

Legend to Supplementary Figure 3

minuslog10 P values (y axis) of the SNPs are shown according to their chromosomal positions (x axis) with lead SNP shown as a purple diamond The colour

intensity of each symbol depicting a SNP reflects the extent of LD with the rs926849 coloured red (r2 gt 08) through to blue (r

2 lt 02) Genetic

recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build 36 (NCBI) of the

human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn to show the relative

positions and therefore the maps are not to physical scale

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 14: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

Supplementary Figure 3 Regional plot of association results and recombination rates for the class II MHC region unadjusted for age and sex

Legend to Supplementary Figure 3

minuslog10 P values (y axis) of the SNPs are shown according to their chromosomal positions (x axis) with lead SNP shown as a purple diamond The colour

intensity of each symbol depicting a SNP reflects the extent of LD with the rs926849 coloured red (r2 gt 08) through to blue (r

2 lt 02) Genetic

recombination rates (cMMb) estimated using HapMap CEU samples are shown with a light blue line Physical positions are based on build 36 (NCBI) of the

human genome Also shown are the relative positions of genes mapping to the region of association Genes have been redrawn to show the relative

positions and therefore the maps are not to physical scale

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 15: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable

13

  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp
Page 16: OPEN ACCESS degeneration in northern Europeans: a meta ... · Methods We have developed a continuous trait based on disc space narrowing and osteophytes growth which is measurable
  • 2013 Williams Ann Rheum Dis
  • 2013 Williams AnnRheumDis Supp