ARTHRITIS & RHEUMATISMVol. 46, No. 6, June 2002, pp 1519–1527DOI 10.1002/art.10260© 2002, American College of Rheumatology

Association of the Interleukin-1 Gene Cluster onChromosome 2q13 With Knee Osteoarthritis

John Loughlin, Barbara Dowling, Zehra Mustafa, and Kay Chapman

Objective. To investigate whether theinterleukin-1 (IL-1) ligand gene cluster at 2q13 encodesfor genetic susceptibility to primary osteoarthritis (OA).

Methods. Seven single-nucleotide polymorphisms(SNPs) and a variable-number tandem repeat (VNTR)polymorphism from within the IL-1 ligand genes IL1A,IL1B, and IL1RN were genotyped in a cohort of 557 OAcases and 557 age-matched controls.

Results. None of the variants demonstrated asso-ciation in the unstratified data set. However, when caseswere stratified according to sex and site of disease (hipor knee), 4 SNPs showed marginal evidence for associ-ation (P < 0.1) in knee cases (n � 136) and male kneecases (n � 58). For 2 of these SNPs, evidence forassociation was enhanced when probands from 60 knee-only affected sibling pair families were genotyped andcombined with the original knee cases (P < 0.05).Further analysis revealed that the associated alleles at 2of these SNPs were markers for the same haplotype, thefrequency of which was significantly elevated when kneecases and knee probands were combined (P � 0.01, oddsratio [OR] 1.4) and when male knee cases and maleknee probands were combined (P � 0.009, OR 1.7).Furthermore, linkage analysis of 2q revealed suggestiveevidence for linkage to the IL-1 gene clusters in affectedsibling pairs concordant for knee OA but no evidencefor linkage in affected sibling pairs concordant for hipOA.

Conclusion. The IL-1 ligand cluster encodes forsusceptibility to knee OA but not to hip OA, highlight-ing the genetic heterogeneity of this common, complexdisease.

Primary osteoarthritis (OA) is a common, multi-factorial disease with a major genetic component that istransmitted in a complex manner (1,2). Genome-widelinkage scans have highlighted up to 7 chromosomalregions that may harbor OA susceptibility genes (3).Chromosome 2q was positive in several scans, suggestingthat this chromosome is likely to harbor 1 or moresusceptibility gene. In a Finnish study of affected siblingpairs, a region of linkage stretching from 2q12 to 2q21was reported for OA of the distal interphalangeal joint(4), and our previous study of affected sibling pairs inthe UK demonstrated a broader region of linkage,stretching from 2q12 to 2q31 (5). Both of these linkagesencompass the interleukin-1 receptor (IL-1R) and li-gand gene clusters at 2q12 and 2q13, respectively.

Although OA is not an autoimmune arthritis,several cytokines are involved in cartilage metabolismand are synthesized by synovial cells and cartilage chon-drocytes (6,7). IL-1 is the main catabolic cytokine of theOA joint and can stimulate synthesis of a number ofproteinases, which can result in the breakdown of carti-lage extracellular matrix proteins. IL-1R antagonist (IL-1Ra) competes with IL-1 for binding to the IL-1 recep-tors and can act as an inhibitor of cartilage loss. Whenthe catabolic and anabolic activities of the cytokines arebalanced, cartilage integrity is maintained. If there is animbalance favoring catabolism, however, cartilage de-struction can proceed, resulting in OA. It is thereforereasonable to propose that a proportion of the geneticsusceptibility to OA may be encoded for by variation inthe activity of interleukins, and that for chromosome 2qthis susceptibility could reside within the IL-1 geneclusters.

The 2 IL-1 genes (IL1A and IL1B) and the geneencoding IL-1Ra (IL1RN) are located on chromosome2q13 within a 430-kb genomic fragment (8). SeveralDNA variants within these genes have been reported,including single-nucleotide polymorphisms (SNPs), mi-crosatellite repeats, and variable-number tandem re-

Supported by grants from the Arthritis Research Campaignand The Wellcome Trust.

John Loughlin, PhD, Barbara Dowling, BSc, Zehra Mustafa,BSc, Kay Chapman, PhD: University of Oxford, Oxford, UK.

Address correspondence and reprint requests to John Lough-lin, PhD, University of Oxford, Institute of Molecular Medicine,Oxford OX3 9DS, UK. E-mail: john.loughlin@ndcls.ox.ac.uk.

Submitted for publication September 7, 2001; accepted inrevised form January 11, 2002.

1519

peats (VNTRs) within IL1RN. We genotyped 7 intra-genic SNPs and the IL1RN VNTR in a case–controlcohort of 557 cases with OA and 557 age-matchedcontrols. When we stratified the cases according to siteof OA (hip or knee) and sex, 3 SNPs showed marginalassociation (P � 0.1) in a knee stratum, and 1 showedmarginal association in a knee stratum and in a hipstratum. When we genotyped the probands from 60knee-only affected sibling pair families, evidence forassociation increased for 2 of these SNPs. Furtheranalysis of the 2 SNPs revealed that the associatedalleles at each SNP were markers for the same haplo-type, the frequency of which was significantly elevatedwhen knee cases and knee probands were combined(P � 0.01) and when male knee cases and male kneeprobands were combined (P � 0.009). We conclude thatthe IL-1 gene cluster harbors susceptibility for knee OAbut not for hip OA, illustrating the high degree ofheterogeneity (both clinical and genetic) involved in thiscommon, multifactorial arthritis.

PATIENTS AND METHODS

Case–control cohort. An association analysis was per-formed in a case–control cohort. All cases of OA werediagnosed at the Nuffield Orthopaedic Centre in Oxford.Cases had undergone total joint replacement (TJR) of the hip(THR) and/or knee (TKR) because of primary OA. Cases whounderwent TJR secondary to other factors, such as fracture orrheumatoid arthritis, were excluded. In each case, the primarystatus was supported by clinical, radiologic, operative, andhistologic findings and has been described in detail elsewhere(9). A total of 557 cases (342 women, 215 men) were studied(Table 1). Three hundred ninety of these cases had undergoneTHR only (unilateral or bilateral), and 136 had undergoneTKR only (unilateral or bilateral). These cases were classifiedas hip-only and knee-only, respectively.

The average age of cases at the time of recruitmentinto the study and at the time of TJR is shown in Table 1. Forcases who had undergone �1 TJR, the age at first TJR was

used for calculating the average age at the time of TJR.Spouses of cases served as controls; these individuals had notundergone joint replacement surgery and had not received anytreatment for OA. If a spouse had any evidence of symptom-atic OA at the time of entry into the study, then that spouseand his or her affected partner (the case) were excluded. In thisway, we identified a cohort of OA-affected individuals and anage-matched control group in which the frequency of clinicalOA was very low. All cases and controls were Caucasian andwere from the UK. Ethical approval for the study was obtainedfrom the Central Oxford Research Ethics Committee, andinformed consent was obtained from all subjects.

Variant typing. We genotyped the 557 cases and 557controls for 8 common polymorphic variants located in theIL-1 ligand gene cluster (Figure 1). Seven of the variants areSNPs, and the eighth is a VNTR polymorphism. All 8 variantshave been extensively described in the scientific literature. The7 SNPs were genotyped by polymerase chain reaction (PCR)–restriction enzyme analysis, with the digestion fragments sep-arated by electrophoresis through 3% agarose. The VNTR wasPCR amplified, and the alleles were separated by electrophor-esis through 3% agarose. Details regarding the primers, PCRconditions, and restriction enzymes used for discrimination ofalleles for each variant are listed in Table 2. For the IL1A �889SNP, the 35th nucleotide of the reverse primer was changedfrom the A that occurs in the published sequence (10) to a C.This change creates an Nco I site in individuals with a C allele.For the IL1B �3954 SNP, the 56th nucleotide of the forwardprimer was changed from the C that occurs in the publishedsequence (11) to a G, thus creating an invariant Taq I site thatwas used to monitor the digestion of PCR product. For theIL1B 5810 SNP, the 39th nucleotide of the reverse primer waschanged from the G that occurs in the published sequence (11)

Figure 1. The 8 polymorphic variants genotyped in the 3 interleukin-1(IL-1) ligand genes. VNTR � variable-number tandem repeats.

Table 1. Characteristics of the 557 cases and 557 controls*

Cases Controls

Female(n � 342)

Male(n � 215)

Female(n � 215)

Male(n � 342)

No. (%) who underwent THR only 244 (71.3) 146 (67.9) – –No. (%) who underwent TKR only 78 (22.8) 58 (27.0) – –No. (%) who underwent THR and TKR 20 (5.9) 11 (5.1) – –Average age at recruitment into study,

years (range)72 (56–90) 74 (61–88) 71 (59–88) 74 (61–89)

Average age at joint surgery, years (range) 64 (47–85) 67 (49–83) – –

* THR � total hip replacement, unilateral or bilateral; TKR � total knee replacement, unilateral orbilateral.

1520 LOUGHLIN ET AL

to a C. This change creates a Bst UI site in individuals with aG allele.

All PCRs were performed under standard conditionsin a 15-�l reaction volume containing 50 ng of genomic DNA.PCRs were performed in 96-well microtiter plates. The 7 SNPswere amplified using AmpliTaq Gold and KCl buffer (AppliedBiosystems, Foster City, CA), and the IL1RN VNTR wasamplified using ReddyMix PCR master mix with (NH4)2SO4buffer (Advanced Biotechnologies, Leatherhead, UK). Re-striction enzymes were purchased from New England Biolabs(Beverly, MA).

Association study. Allele distributions in cases andcontrols were compared using standard chi-square analysis-of-contingency tables. Haplotype frequencies in pairs of SNPswere estimated using the EH� program (http://www.iop.kcl.ac.uk/IoP/Departments/PsychMed/GEpiBST/software.stm)(12). The differences in haplotype frequencies as estimated byEH� were then compared between cases and controls, againusing standard chi-square analysis-of-contingency tables. Oddsratios (ORs) were calculated with 95% confidence intervals(95% CIs). For each stratification analysis, female cases werecompared with female controls, and male cases were comparedwith male controls.

Linkage study. Six polymorphic microsatellite markersthat encompassed the 2 IL-1 gene clusters (Table 3) weregenotyped in a separate cohort of 576 OA families. All of thesefamilies were Caucasian, were from the UK, and were unre-lated to any subject in the case–control cohort described above(i.e., the 2 cohorts were independent). Each of the 576 familiescontains at least 1 pair of siblings concordant for primary OA(affected sibling pair). As was true in the case–control cohort,OA in this cohort was defined as the need for TJR because ofOA joint damage; individuals who had undergone joint re-

placement secondary to other factors were excluded. Recruit-ment of these families has been described previously (9).

Of the 576 OA families, 382 contained affected siblingpairs who had undergone THR only (unilateral or bilateral),and 60 contained affected sibling pairs who had undergoneTKR only (unilateral or bilateral). The remaining familiescontained sibling pairs in which 1 sibling had undergone THRand the other sibling had undergone TKR. The microsatellitemarkers were PCR amplified, with 1 of the primers fluores-cently labeled. The amplification products were electropho-resed through 5% Long Ranger acrylamide (Flowgen, Leices-tershire, UK) using an ABI model 377 DNA sequencer(Applied Biosystems). Alleles were sized using GeneScan andGenotyper software (Applied Biosystems). Allele frequencieswere calculated from the input data using GAS (http://users.ox.ac.uk/�ayoung/gas.html), with subsequent multipoint

Table 2. Genotyping of the 8 polymorphic variants*

Gene VariantLocationin gene Primer sequences, forward/reverse

PCR

Enzyme Ref.Size,bp

Temp.,°C

Mg��,mM

IL1A �889 (T–C) Promoter 5�-GTT-CTA-CCA-CCT-GAA-CTA-GGC-3� 116 58 2.5 Nco I; T � Nco I (�), 105�-AAA-GGA-AGG-CAT-GGA-TTT-TTA-CAT-

ATG-AGC-CTT-CAA-TG-3�C � Nco I (�) 20

IL1B �3954 (C–T) Exon 5 5�-TTC-ATC-CCT-ACT-GGT-GTT-GTC-ATC-AGA-CTT-TGA-CCG-TAT-ATG-CTC-AGG-TGT-CCT-CGA-AGA-AAT-C-3�

227 60 2.0 Taq I; C � Taq I (�),T � Taq I (�)

1121

5�-AGG-TGG-AGA-GCT-TTC-AGT-TC-3�IL1B 5810 (G–A) Intron 4 5�-CTA-AGT-AGC-TCT-GTT-GCT-CG-3� 220 58 2.0 Bst UI; G � Bst UI (�), 11

5�-TTA-GGT-ATA-AAA-TCA-GAA-GGG-CAG-GCC-TCG-TGA-GGC-GAC-G-3�

A � Bst UI (�) 22

IL1B �31 (T–C) Promoter 5�-CCA-CCA-ATA-CTC-TTT-TCC-CC-3� 190 56 2.5 Alu I; T � Alu I (�), 115�-TGA-AGA-TTG-GCT-GAA-GAG-AAT-C-3� C � Alu I (�) 23

IL1B �511 (C–T) Promoter 5�-TGG-CAT-TGA-TCT-GGT-TCA-TC-3� 305 58 2.5 Ava I; C � Ava I (�), 115�-GTT-TAG-GAA-TCT-TCC-CAC-TT-3� T � Ava I (�) 24

IL1RN VNTR Intron 2 5�-CCC-CTC-AGC-AAC-ACT-CC-3� �200 58 1.5 235�-GGT-CAG-AAG-GGC-AGA-GA-3� 25

IL1RN 9589 (A–T) Intron 3 5�-TTG-TGG-GGA-CCA-GGG-GAG-AT-3� 379 59 2.0 Ssp I; A � Ssp I (�), 225�-AGC-CTG-GCA-CTC-TGC-TGA-AT-3� T � Ssp I (�)

IL1RN 11100 (T–C) Exon 4 5�-AGG-GAG-GCA-GCA-CAG-GAC-TT-3� 329 62 1.8 Msp A1I; T � Msp A1I (�), 225�-AGT-CCC-TGC-AGT-CCT-TGC-CA-3� C � Msp A1I (�)

* PCR � polymerase chain reaction; Temp. � annealing temperature; VNTR � variable-number tandem repeats.

Table 3. Characteristics of the 6 microsatellite markers*

MarkerDistance from

2p telomere, cM†Cytogenetic

band‡

D2S139 106.0 2p12D2S2333 107.7 2p11.2D2S2216 115.3 2p11.2D2S160 127.4 2q13D2S363 129.5 2q14.1D2S347 135.7 2q14.3

* Markers were genotyped in the linkage analysis of the 2q regionharboring the 2 interleukin-1 gene clusters.† Determined using GeneMap’99 (http://www.ncbi.nlm.nih.gov/genemap99).‡ Determined using Ensembl (http://www.ensembl.org).

IL-1 GENE CLUSTER AND SUSCEPTIBILITY TO KNEE OA 1521

linkage analysis performed using GeneHunter-Plus (http://linkage.rockefeller.edu/soft/list.html) (13,14). Probands fromthe 60 TKR-only families were also analyzed in the associationstudy, to further test the positive findings observed in kneecases.

RESULTS

When comparing the allelic scores for the 8variants, we noticed that the IL1B promoter SNPs �31and �511 were in complete linkage disequilibrium (LD),with each subject having identical genotypes for the 2SNPs. For example, individuals who had the genotype(��) at SNP �31 also had the genotype (��) at SNP�511, whereas those who had the genotype (��) at�31 also had the genotype (��) at �511. Therefore,these 2 SNPs provided the same information. We alsonoticed that the IL1RN SNP 9589 was in complete LDwith the IL1RN VNTR. We observed 4 alleles of theVNTR; allele I (4 repeat units) had a frequency of 0.690,allele II (2 repeat units) had a frequency of 0.286, allele

III (5 repeat units) had a frequency of 0.023, and alleleIV (3 repeat units) had a frequency of 0.002. Alleles I,III, and IV of the VNTR were in complete LD with the(�) allele of SNP 9589, while allele II was in completeLD with the (�) allele of 9589. Because of the very lowfrequency of alleles III and IV of the VNTR (combinedfrequency only 0.025), this polymorphism is effectively abiallelic marker. In the subsequent association analysis,we did not include data from the IL1B �31 SNP or theIL1RN VNTR.

Association analysis: cases versus controls. Ta-ble 4 lists the frequency and number of (�) and (�)alleles for the 6 SNPs in cases and controls. Cases werealso stratified according to sex, joint replaced (hip orknee), and sex combined with joint replaced. All SNPswere in Hardy-Weinberg equilibrium.

Table 5 lists the P values derived when thenumbers of (�) and (�) alleles in cases and controlswere compared. There were no significant differences(P � 0.05) in allele frequencies between the unstratified

Table 4. Frequency (no.) of alleles for each single-nucleotide polymorphism (SNP)*

Gene/SNP

IL1A/�899 IL1B/�3954 IL1B/5810 IL1B/�511 IL1RN/9589 IL1RN/11100

Cases, n � 557All 0.68/0.32 (748/346) 0.76/0.24 (842/260) 0.63/0.37 (691/399) 0.68/0.32 (745/349) 0.26/0.74 (278/80) 0.28/0.72 (314/788)Female 0.68/0.32 (460/212) 0.78/0.22 (524/150) 0.63/0.37 (424/246) 0.67/0.33 (446/222) 0.27/0.73 (182/482) 0.27/0.73 (182/492)Male 0.68/0.32 (288/134) 0.74/0.26 (318/110) 0.64/0.36 (267/153) 0.70/0.30 (299/127) 0.23/0.77 (96/326) 0.31/0.69 (132/296)Hip-only 0.70/0.30 (534/232) 0.78/0.22 (602/170) 0.61/0.39 (472/300) 0.68/0.32 (521/243) 0.26/0.74 (198/560) 0.28/0.72 (216/554)Knee-only 0.63/0.37 (167/99) 0.72/0.28 (193/75) 0.70/0.30 (182/78) 0.67/0.33 (182/90) 0.24/0.76 (65/203) 0.28/0.72 (76/194)Female hip-

only0.69/0.31 (332/148) 0.79/0.21 (379/103) 0.61/0.39 (299/189) 0.67/0.33 (318/158) 0.27/0.73 (127/345) 0.27/0.73 (128/350)

Male hip-only 0.71/0.29 (202/84) 0.77/0.23 (223/67) 0.61/0.39 (173/111) 0.70/0.30 (203/85) 0.25/0.75 (71/215) 0.30/0.70 (88/204)Female knee-

only0.65/0.35 (97/53) 0.76/0.24 (115/37) 0.69/0.31 (101/45) 0.65/0.35 (102/54) 0.28/0.72 (43/111) 0.28/0.72 (43/113)

Male knee-only 0.60/0.40 (70/46) 0.67/0.33 (78/38) 0.71/0.29 (81/33) 0.69/0.31 (80/36) 0.19/0.81 (22/92) 0.29/0.71 (33/81)Controls, n � 557

All 0.70/0.30 (763/327) 0.76/0.24 (841/267) 0.64/0.36 (708/398) 0.68/0.32 (747/349) 0.28/0.72 (315/795) 0.28/0.72 (307/799)Female 0.71/0.29 (304/126) 0.76/0.24 (325/101) 0.62/0.38 (264/160) 0.69/0.31 (284/130) 0.29/0.71 (126/302) 0.27/0.73 (113/309)Male 0.70/0.30 (459/201) 0.76/0.24 (516/166) 0.65/0.35 (444/238) 0.68/0.32 (463/219) 0.28/0.72 (189/493) 0.28/0.72 (194/490)

* Values are the frequencies (numbers) of alleles, reported as (�)/(�).

Table 5. P values derived when comparing the number of (�) and (�) alleles in cases and controls

Gene

Single-nucleotide

polymorphism

All casesvs. all

controls

Femalecases vs.female

controls

Malecases vs.

malecontrols

Hip cases vs.all controls

Knee casesvs. all

controls

Female hipcases vs.female

controls

Male hipcases vs.

malecontrols

Female kneecases vs.female

controls

Male kneecases vs.

malecontrols

IL1A �889 0.44 0.47 0.70 0.94 0.03 0.67 0.80 0.20 0.07IL1B �3954 0.83 0.63 0.66 0.32 0.22 0.45 0.75 0.96 0.08IL1B 5810 0.80 0.79 0.66 0.23 0.08 0.81 0.25 0.17 0.26IL1B �511 0.99 0.58 0.47 0.97 0.75 0.62 0.47 0.53 0.90IL1RN 9589 0.17 0.51 0.08 0.31 0.20 0.45 0.09 0.80 0.08IL1RN 11100 0.74 0.99 0.42 0.93 0.96 0.94 0.63 0.94 0.99

1522 LOUGHLIN ET AL

cases and the controls for any of the 6 SNPs. Stratifica-tion according to sex and joint replaced resulted in asignificant difference in the frequency of alleles at theIL1A �889 SNP between knee cases and controls (P �0.03); this difference involved an increased frequency ofthe (�) allele in knee cases (0.37 versus 0.30 in controls).Six additional differences in allele frequency betweenstratified cases and controls approached significance: 1)the frequency of the IL1A �889 (�) allele was 0.40 inmale knee cases versus 0.30 in male controls (P � 0.07);2) the frequency of the IL1B �3954 (�) allele was 0.33in male knee cases versus 0.24 in male controls (P �0.08); 3) the frequency of the IL1B 5810 (�) allele was0.70 in knee cases versus 0.64 in controls (P � 0.08); 4)the frequency of the IL1RN 9589 (�) allele was 0.81 inmale knee cases versus 0.72 in male controls (P � 0.08);5) the frequency of the IL1RN 9589 (�) allele was 0.75in male hip cases versus 0.72 in male controls (P � 0.09);and 6) the frequency of the IL1RN 9589 (�) allele was0.77 in male cases versus 0.72 in male controls (P �0.08).

These data may be interpreted in 2 ways. First,because we performed 54 tests (6 SNPs genotyped and 9strata tested), we would expect 2–3 positive results bychance alone (assuming each test is independent); there-fore, our results are not significant. Second, we observedallele sharing across a number of SNPs, which mayrepresent genuine, albeit marginal, differences betweencases and controls.

In an attempt to distinguish between these 2

possibilities, we genotyped the positive SNPs in addi-tional OA-affected individuals. We focused our effortson additional knee cases because, of the 6 comparisonsthat yielded a P value �0.1 in the joint-specific strata, 5were from a knee stratum and only 1 was from a hipstratum (Table 5). The additional affected individualswere probands from the 60 knee-only affected siblingpair families who are currently being studied in ourlinkage analysis project. These additional “patients” arehenceforth termed knee probands.

Association analysis: knee probands versus con-trols. We genotyped the �889, �3954, 5810, and 9589SNPs in the 60 knee probands (37 women and 23 men).Although there were no significant differences betweenprobands and controls in the frequency of the alleles atany of the 4 SNPs (Table 6), some of the differencesmerit comment.

IL1A �889. The (�) allele that was significantlyelevated in knee cases was also elevated in knee pro-bands, with a frequency of 0.35 compared with 0.30 incontrols; in male knee probands, the frequency was 0.39compared with 0.30 in male controls. These data supportthe original observation.

IL1B �3954. The (�) allele that was elevated inmale knee cases was also elevated in male knee pro-bands, with a frequency of 0.29 compared with 0.24 inmale controls. Again, this supports the original observa-tion.

IL1B 5810. The (�) allele that was elevated inknee cases had a reduced frequency in knee probands

Table 6. Number and frequency of (�) and (�) alleles in probands versus controls*

Gene/SNP, comparison

Number (frequency)

P

Probands Controls

(�) (�) (�) (�)

IL1A/�889All knee probands vs. controls 77 (0.65) 41 (0.35) 763 (0.70) 327 (0.30) 0.34Male knee probands vs. male controls 28 (0.61) 18 (0.39) 459 (0.70) 201 (0.30) 0.29Female knee probands vs. female controls 49 (0.68) 23 (0.32) 304 (0.71) 126 (0.29) 0.76

IL1B/�3954All knee probands vs. controls 86 (0.75) 28 (0.25) 841 (0.76) 267 (0.24) 1.0Male knee probands vs. male controls 30 (0.71) 12 (0.29) 516 (0.76) 166 (0.24) 0.67Female knee probands vs. female controls 56 (0.78) 16 (0.22) 325 (0.76) 101 (0.24) 0.90

IL1B/5810All knee probands vs. controls 67 (0.58) 49 (0.42) 708 (0.64) 398 (0.36) 0.22Male knee probands vs. male controls 26 (0.57) 20 (0.43) 444 (0.65) 238 (0.35) 0.31Female knee probands vs. female controls 41 (0.59) 29 (0.41) 264 (0.62) 160 (0.38) 0.65

IL1RN/9589All knee probands vs. controls 22 (0.20) 86 (0.80) 315 (0.28) 795 (0.72) 0.10Male knee probands vs. male controls 9 (0.20) 35 (0.80) 189 (0.28) 493 (0.72) 0.39Female knee probands vs. female controls 13 (0.20) 51 (0.80) 126 (0.29) 302 (0.71) 0.18

* Probands (n � 60) are from knee-only affected sibling pair families. SNP � single-nucleotide polymorphism.

IL-1 GENE CLUSTER AND SUSCEPTIBILITY TO KNEE OA 1523

(0.58 versus 0.64 in controls), which does not support theoriginal observation.

IL1RN 9589. The (�) allele that was elevated inknee cases was also elevated in knee probands, with afrequency of 0.80 in probands versus 0.72 in controls.This difference approached significance (P � 0.1) andsupports the original observation.

Association analysis: knee cases and knee pro-bands combined versus controls. We subsequently com-bined the data from knee cases and knee probands todetermine whether adding more affected individuals tothe cases had increased evidence for association. Table 7summarizes the allele frequencies for the 4 SNPs fromknee cases alone, knee probands alone, and knee casescombined with knee probands. This table also lists the Pvalues obtained when knee cases alone were comparedwith controls, when knee probands alone were com-pared with controls, and when knee cases combined withknee probands were compared with controls.

IL1A �889. The P value of 0.03 that was obtainedwhen all knee cases were compared with controls de-creased to 0.02 when knee cases were combined withknee probands. Furthermore, the P value that wasobtained when male knee cases were compared withmale controls (P � 0.07) achieved significance whenmale knee cases were combined with male knee pro-bands (P � 0.04). The OR for the �889 (�) allele was1.3 (95% CI 1.0–1.7) for knee cases and knee probandscombined and 1.5 (95% CI 1.0–2.1) for male knee casesand male knee probands combined.

IL1B �3954. The P value of 0.08, which wasobtained when male knee cases were compared with

male controls, remained constant when male cases andmale probands were combined. The OR for the �3954(�) allele was 0.67 (95% CI 0.5–1.0) for male knee casesand male knee probands combined, implying that thelow P value is not suggestive of a genuine association.

IL1B 5810. The P value of 0.08 observed for allknee cases increased to 0.48 when knee cases werecombined with knee probands, indicating that the orig-inal result, which was tending toward significance, wasprobably a false positive.

IL1RN 9589. The P value of 0.20 obtained whenall knee cases were compared with controls decreased to0.06 when knee cases were combined with knee pro-bands. In addition, the P value of 0.08 obtained whenmale knee cases were compared with male controlsreached significance (P � 0.05) when male knee caseswere combined with male knee probands. The OR forthe IL1RN 9589 (�) allele was 1.3 (95% CI 1.0–1.7) forknee cases and knee probands combined and 1.6 (95%CI 1.0–2.4) for male knee cases and male knee probandscombined.

Overall, the marginally positive and suggestiveassociation data observed for SNPs �889 and 9589,respectively, in knee cases were enhanced when kneecases and knee probands were combined. However, allof the CIs of the ORs for the associated alleles at these2 SNPs encompassed 1.0, indicating that these associa-tions could be false positives. The marginally positiveassociation data observed for SNPs �3954 and 5810 inknee cases were not supported when knee cases andknee probands were combined, suggesting that the orig-inal associations were false positive.

Table 7. Summary of allele frequencies for controls, knee cases, knee probands, and knee cases combined with knee probands*

Gene/SNP, subject group Controls CasesP, cases

vs. controls ProbandsP, probandsvs. controls Cases � probands

P, cases �probands vs.

controls

IL1A/�889All 0.70/0.30 0.63/0.37 0.03 0.65/0.35 0.34 0.64/0.36 0.02Males 0.70/0.30 0.60/0.40 0.07 0.61/0.39 0.29 0.60/0.40 0.04Females 0.71/0.29 0.65/0.35 0.20 0.68/0.32 0.76 0.66/0.34 0.24

IL1B/�3954All 0.76/0.24 0.72/0.28 0.22 0.75/0.25 1.0 0.73/0.27 0.30Males 0.76/0.24 0.67/0.33 0.08 0.71/0.29 0.67 0.68/0.32 0.08Females 0.76/0.24 0.76/0.24 1.0 0.78/0.22 0.90 0.76/0.24 0.94

IL1B/5810All 0.64/0.36 0.70/0.30 0.08 0.58/0.42 0.22 0.66/0.34 0.48Males 0.65/0.35 0.71/0.29 0.26 0.57/0.43 0.31 0.67/0.33 0.75Females 0.62/0.38 0.69/0.31 0.17 0.59/0.41 0.65 0.66/0.34 0.44

IL1RN/9589All 0.28/0.72 0.24/0.76 0.20 0.20/0.80 0.10 0.23/0.77 0.06Males 0.28/0.72 0.19/0.81 0.08 0.20/0.80 0.39 0.20/0.80 0.05Females 0.29/0.71 0.28/0.72 0.80 0.20/0.80 0.18 0.26/0.74 0.37

* Values are the frequencies of alleles, reported as (�)/(�).

1524 LOUGHLIN ET AL

Haplotype analysis. We next attempted to deter-mine whether the �889 and 9589 SNPs were highlight-ing the same association at the IL-1 gene cluster; i.e.,whether the associated (�) alleles at these 2 SNPsconstitute a single associated haplotype. We estimatedthe frequency of the 4 possible haplotypes for the 2SNPs using the EH� program. This estimation wasperformed on knee cases and knee probands combined,on male knee cases and male knee probands combined,on all controls, and on male controls (Table 8). Therewas a significant difference in the frequency of estimatedhaplotypes when knee cases and knee probands com-bined were compared with controls (P � 0.05). Therewas also a significant difference in the frequency ofestimated haplotypes when male knee cases and maleknee probands combined were compared with controls(P � 0.03). Both of these differences were attributed toan increase in the frequency of the �889 (�)/9589 (�)haplotype in cases and probands; the frequency of thishaplotype was 0.31 in knee cases and knee probandscombined and 0.36 in the male knee cases and male kneeprobands combined, compared with 0.24 in all controlsand 0.25 in male controls.

When we specifically examined the frequency ofthe (�)/(�) haplotype, it was significantly increased inknee cases and knee probands combined compared withcontrols (P � 0.01) and in male knee cases and maleknee probands combined compared with male controls(P � 0.009). The OR for this haplotype was 1.4 (95% CI1.1–1.8) in knee cases and knee probands combined and1.7 (95% CI 1.2–2.4) in male knee cases and male kneeprobands combined.

Overall, these data suggest that the (�) allele ofSNP �889 and the (�) allele of SNP 9589 are markersfor the same susceptibility haplotype. Neither SNP is, byitself, the causal variant that encodes for susceptibilitybut instead is in LD with the causal variant. Furtheranalysis of variants within the IL-1 gene cluster will be

required to fully characterize the susceptibility at theIL-1 ligand gene cluster.

Evidence from linkage analysis for a knee locusat 2q12–q13. Our group is currently in the process ofperforming a refined linkage analysis of our 2q linkage,using a cohort of 576 affected sibling pair OA families.Our original linkage encompassed �50 cM of 2q, from2q12 to 2q31 (5). The major linkage peak was not at2q12–q13, where the IL-1 gene clusters reside, butinstead at 2q24.3–q31. However, because our linkageencompassed the IL-1 genes, our refined analysis in-cludes the typing of 6 microsatellite markers in thisregion of chromosome 2, from 2p12 to 2q14.3 (Table 3).Multipoint linkage analysis of these 6 markers resultedin a maximum nonparametric linkage (NPL) score of0.23 (P � 0.39) for all 576 families (Figure 2). An NPLscore of 0.17 (P � 0.42) was obtained for the 382 familiesthat contain sibling pairs concordant for hip OA, and anNPL score of 1.1 (P � 0.14) was obtained for the 60families that contain sibling pairs concordant for kneeOA. Clearly, none of these NPL scores are significant.However, the absence of linkage in hip OA families andthe moderately suggestive linkage in knee OA familiesmay be providing corroborative (albeit weak) supportfor our association data, implying a role for the IL-1gene clusters in susceptibility to knee OA but not to hipOA. Furthermore, the peak of linkage for the 60 kneeOA families was at marker D2S160 (Figure 2), which islocated �2 cM from the IL-1 ligand cluster.

DISCUSSION

Identifying susceptibility genes for complex traitsis an arduous task that is confounded by clinical andgenetic heterogeneity, differences between populations

Figure 2. Multipoint linkage analysis of 2p12–q14.3. IL-1 �interleukin-1.

Table 8. Estimated number and frequency of haplotypes for theIL1A �899 and IL1RN 9589 single-nucleotide polymorphisms*

Haplotype,�889/9589

Kneecases and

kneeprobands

Male kneecases andmale kneeprobands

Allcontrols

Malecontrols

(�)/(�) 67 (0.18) 23 (0.15) 239 (0.22) 145 (0.22)(�)/(�) 168 (0.45) 69 (0.44) 521 (0.48) 309 (0.47)(�)/(�) 22 (0.06) 8 (0.05) 65 (0.06) 39 (0.06)(�)/(�) 115 (0.31) 56 (0.36) 261 (0.24) 165 (0.25)

* Values are the number (frequency).

IL-1 GENE CLUSTER AND SUSCEPTIBILITY TO KNEE OA 1525

in the frequency of disease genes, and the fact that anysingle causal mutation will account for only a fraction ofthe variance of the trait. Despite these limitations,encouraging breakthroughs are occurring, as exempli-fied by 2 recent reports of the identification of asusceptibility gene for Crohn’s disease on chromosome16 (15,16).

OA is a complex genetic disease that is amongthe most difficult to analyze, due to its high frequency inthe general population and its extensive clinical hetero-geneity. The degree to which clinical heterogeneitytranslates into genetic heterogeneity is unknown. How-ever, a number of epidemiologic studies have high-lighted potential differences in the degree of OA heri-tability between different joint groups and between the 2sexes (1,2,17). Results of these studies support theapproach of stratifying for site of disease and sex whenattempting to identify OA susceptibility genes. In thestratification analysis of our genome-wide linkage scan,we identified loci that were more relevant to disease at aparticular joint site or were restricted to a single sex(5,9,18). Although stratification of data is associatedwith a risk of generating a large number of false-positiveresults (because of the greater number of tests per-formed), it avoids the possibility of false-negative re-sults, which occur when data are not stratified andgenuine differences between strata exist.

From a functional perspective, the IL genes aregood candidates for harboring OA susceptibility (6,7).From a genetics perspective, there are also reasonablegrounds for examining these genes. In the Finnish study,the major linkage was to 2q12–q13, where the IL-1 recep-tor and ligand gene clusters reside (4); our original linkageto chromosome 2q also encompassed these genes (5).

These functional and linkage data prompted ouranalysis of the IL-1 ligand cluster. We genotyped 7 SNPsand 1 VNTR from within the genes encoding IL-1�,IL-1�, and IL-1Ra in a cohort of 557 cases and 557controls. No association was detected with any of thevariants. When we stratified our cases by site of disease(hip or knee), we detected a marginal association forSNP �889 in knee cases. We also detected differencesbetween knee cases and controls in allele frequencies forSNPs �3954, 5810, and 9589; these differences ap-proached significance. However, none of these resultswould merit much comment if corrections were madefor the large number of tests performed.

In an attempt to determine whether any resultswere genuine positives, we genotyped these 4 SNPs inthe probands from our knee-only affected sibling pairfamilies. None of the SNPs were associated in these

probands, although the association for SNP 9589 ap-proached significance. When cases and probands werecombined, however, the P values for SNPs �889 and9589 decreased with respect to the original P valuesobtained for knee cases alone. Increasing the number ofknee cases increased the evidence for association atthese 2 SNPs, suggesting that these associations may begenuine. Furthermore, when haplotypes between SNPs�889 and 9589 were constructed, the haplotype com-posed of the associated (�) alleles of each SNP wassignificantly elevated in knee cases and knee probands.These data suggest that the (�) alleles of these 2 SNPsdo not themselves encode for susceptibility but areinstead in LD with the causal mutation. Finally, ourrefined linkage analysis of chromosome 2q providedadditional, but very moderate, indications for a locus at2q13 that is more relevant to knee OA than to hip OA.

A German group recently reported an associa-tion between the IL-1 ligand gene cluster and OA (19).This group genotyped 1 variant from each of the genesencoding TNF�, IL-1�, IL-1Ra, and IL-6 in 61 patients(with end-stage knee or hip OA) and 254 controls. Theydetected a positive association to the IL1B SNP �3954and to the IL1RN VNTR. The �3954 allele that waselevated in their study was the (�) allele, which was thesame allele that was elevated (although not significantly)in our male knee cases.

As noted earlier, the IL1RN VNTR is in com-plete LD with the IL1RN 9589 SNP, and this SNPapproached significant association in our knee cases andwas significantly associated in our male knee cases. TheVNTR allele that was elevated in the German study wasallele II, which is in complete LD with the (�) allele of9589. In our study, we observed an elevation in thefrequency of the (�) allele of 9589, which is in completeLD with alleles I, III, and IV of the VNTR. These 2studies, therefore, demonstrated an increased frequencyof opposite alleles of the IL1RN VNTR; our studysupports the locus association of the German study butnot the allelic association. Whether the association ofopposite alleles is a reflection of differences between the2 populations examined (German versus UK) or is theresult of differences in the ascertainment of OA betweenthe 2 studies is open to speculation. Both groups re-cruited patients based on joint replacement surgery.However, the German group subdivided their patientsinto those who had tumor necrosis factor �–positivechondrocytes and those who had IL-1�–positive chon-drocytes. We performed no such subdivision.

Overall, these analyses have provided evidencefor an association of a haplotype within the IL-1 ligand

1526 LOUGHLIN ET AL

gene cluster to knee OA. There was no evidence for anassociation of this gene cluster to hip OA, which mayreflect heterogeneity of susceptibility between thesedifferent joint sites. The association highlights a diseasepathway that merits comprehensive examination and willhelp guide our efforts when targeting candidate genes instronger regions of linkage.

ACKNOWLEDGMENTS

We thank Professor Andrew Carr, Dr. Jai Chitnavis,and Ms Kim Clipsham, who organized the collection of patientand family genetic samples used in this study. That collectionwas funded in part by The Wishbone Trust.

REFERENCES

1. MacGregor AJ, Spector TD. Twins and the genetic architecture ofosteoarthritis. Rheumatology (Oxford) 1999;38:583–90.

2. Lanyon P, Muir K, Doherty S, Doherty M. Assessment of a geneticcontribution to osteoarthritis of the hip: sibling study. BMJ2000;321:1179–83.

3. Loughlin J. Genetic epidemiology of primary osteoarthritis. CurrOpin Rheumatol 2001;13:111–6.

4. Leppavuori J, Kujala U, Kinnunen J, Kaprio J, Nissila M, Helio-vaara M, et al. Genome scan for predisposing loci for distalinterphalangeal joint osteoarthritis: evidence for a locus on 2q.Am J Hum Genet 1999;65:1060–7.

5. Loughlin J, Mustafa Z, Smith A, Irven C, Carr AJ, Clipsham K, etal. Linkage analysis of chromosome 2q in osteoarthritis. Rheuma-tology (Oxford) 2000;39:377–81.

6. Goldring MB. The role of the chondrocyte in osteoarthritis.Arthritis Rheum 2000;43:1916–26.

7. Pelletier J-P, Martel-Pelletier J, Abramson SB. Osteoarthritis, aninflammatory disease: potential implication for the selection ofnew therapeutic targets. Arthritis Rheum 2001;44:1237–47.

8. Nothwang HG, Strahm B, Denich D, Kubler M, Schwabe J,Gingrich JC, et al. Molecular cloning of the interleukin-1 genecluster: construction of an integrated YAC/PAC contig and apartial transcriptional map in the region of chromosome 2q13.Genomics 1997;41:370–8.

9. Chapman K, Mustafa Z, Irven CM, Carr AJ, Clipsham K, Smith A,et al. Osteoarthritis-susceptibility locus on chromosome 11q, de-tected by linkage. Am J Hum Genet 1999;65:167–74.

10. Furutani Y, Notake M, Fukui T, Ohue M, Nomura H, Yamada M,

et al. Complete nucleotide sequence of the gene for humaninterleukin 1 alpha. Nucleic Acids Res 1986;14:3167–79.

11. Bensi G, Raugei G, Palla E, Carinci V, Buonamassa DT, Melli M.Human interleukin-1� gene. Gene 1987;52:95–101.

12. Zhao JH, Curtis D, Sham PC. Model-free analysis and permuta-tion tests for allelic associations. Hum Hered 2000;50:133–9.

13. Kruglyak L, Daly MJ, Reeve-Daly MP, Lander ES. Parametric andnonparametric linkage analysis: a unified multipoint approach.Am J Hum Genet 1996;58:1347–63.

14. Kong A, Cox NJ. Allele-sharing models: LOD scores and accuratelinkage tests. Am J Hum Genet 1997;61:1179–88.

15. Hugot J-P, Chamaillard M, Zouali H, Lesage S, Cezard J-P,Belaiche J, et al. Association of NOD2 leucine-rich repeat variantswith susceptibility to Crohn’s disease. Nature 2001;411:599–603.

16. Ogura Y, Bonen DK, Inohara N, Nicolae DL, Chen FF, Ramos R,et al. A frameshift mutation in NOD2 associated with susceptibil-ity to Crohn’s disease. Nature 2001;411:603–6.

17. Chitnavis J, Sinsheimer JS, Clipsham K, Loughlin J, Sykes B,Burge PD, et al. Genetic influences in end-stage osteoarthritis:sibling risks of hip and knee replacement for idiopathic osteoar-thritis. J Bone Joint Surg Br 1997;79:660–4.

18. Loughlin J, Mustafa Z, Irven C, Smith A, Carr AJ, Sykes B, et al.Stratification analysis of an osteoarthritis genome screen: sugges-tive linkage to chromosomes 4, 6 and 16. Am J Hum Genet1999;65:1795–8.

19. Moos V, Rudwaleit M, Herzog V, Hohlig K, Sieper J, Muller B.Association of genotypes affecting the expression of interleu-kin-1� or interleukin-1 receptor antagonist with osteoarthritis.Arthritis Rheum 2000;43:2417–22.

20. Rebeck GW. Confirmation of the genetic association of interleu-kin-1A with early onset sporadic Alzheimer’s disease. NeurosciLett 2000;293:75–7.

21. Pociot F, Mølvig J, Wogensen L, Worsaae H, Nerup J. A TaqIpolymorphism in the human interleukin-1� (IL-1�) gene correlateswith IL-1� secretion in vitro. Eur J Clin Invest 1992;22:396–402.

22. Guasch JF, Bertina RM, Reitsma PH. Five novel intragenicdimorphisms in the human interleukin-1 genes combine to highinformativity. Cytokine 1996;8:598–602.

23. El-Omar EM, Carrington M, Chow W-H, McColl KEL, BreamJH, Young HA, et al. Interleukin-1 polymorphisms associated withincreased risk of gastric cancer. Nature 2000;404:398–402.

24. Cantagrel A, Navaux F, Loubet-Lescoulie P, Nourhashemi F,Enault G, Abbal M, et al. Interleukin-1�, interleukin-1 receptorantagonist, interleukin-4, and interleukin-10 gene polymorphisms:relationship to occurrence and severity of rheumatoid arthritis.Arthritis Rheum 1999;42:1093–100.

25. Tarlow JK, Blakemore AIF, Lennard A, Solari R, Hughes HN,Steinkasserer A, et al. Polymorphism in human IL-1 receptorantagonist gene intron 2 is caused by variable numbers of an 86-bptandem repeat. Hum Genet 1993;91:403–404.

IL-1 GENE CLUSTER AND SUSCEPTIBILITY TO KNEE OA 1527