Association of the Interleukin-1 Gene Cluster With Radiographic Signs of Osteoarthritis of the Hip

Transcription

1 ARTHRITIS & RHEUMATISM Vol. 50, No. 4, April 2004, pp DOI /art , American College of Rheumatology Association of the Interleukin-1 Gene Cluster With Radiographic Signs of Osteoarthritis of the Hip Ingrid Meulenbelt, 1 Albert B. Seymour, 2 Marja Nieuwland, 1 Tom W. J. Huizinga, 1 Cornelia M. van Duijn, 3 and P. Eline Slagboom 1 Objective. To study the role of the interleukin-1 gene (IL1B) and the IL-1 receptor antagonist gene (IL1RN) in relation to the occurrence of radiographic osteoarthritis (ROA) in the hip, knee, and hand and disc degeneration of the spine. Methods. The study population consisted of a random sample of 886 subjects (ages years) from a population-based cohort (the Rotterdam study). Two polymorphisms within IL1B (3953C>T and 511C>T) and one within IL1RN (the variable-number tandem repeat [VNTR]) were analyzed and used in an association study of the occurrence of ROA. Haplotyping and simultaneous logistic regression analysis were performed to investigate whether the associations observed were independent. Results. Associations with a predisposition for hip ROA were observed for heterozygous and homozygous carriers of the rare IL1B allele 511T (crude odds ratio [OR] 1.8, 95% confidence interval [95% CI] and OR 2.9, 95% CI , respectively) and of the IL1RN VNTR allele 2 (crude OR 2.0, 95% CI and OR 3.3, 95% CI , respectively). An additive effect was observed for carriers of risk alleles of both polymorphisms, with a significant linear-by-linear association (P ). Funded by the Dutch Arthritis Association, the Netherlands Organisation for Scientific Research, and Pfizer Global Research & Development. 1 Ingrid Meulenbelt, PhD, Marja Nieuwland, Tom W. J. Huizinga, MD, PhD, P. Eline Slagboom, PhD: Leiden University Medical Center, Leiden, The Netherlands; 2 Albert B. Seymour, PhD: Discovery Pharmacogenomics, Pfizer Global Research & Development, Groton, Connecticut; 3 Cornelia M. van Duijn, PhD: Erasmus University Medical School, Rotterdam, The Netherlands. Address correspondence and reprint requests to Ingrid Meulenbelt, PhD, Molecular Epidemiology, Leiden University Medical Center, PO Box 9503, 2300 RA Leiden, The Netherlands. i.meulenbelt@lumc.nl. Submitted for publication March 3, 2003; accepted in revised form December 11, Conclusion. Our findings suggest that the IL-1 gene cluster polymorphisms may play a significant role in the pathogenesis of OA of the hip. Osteoarthritis (OA) is a joint disease characterized by degeneration of articular cartilage and remodeling of the subchondral bone with sclerosis. It has been demonstrated that genetic factors also play an important role in late-onset OA. Familial aggregation of OA (especially hand, knee, and hip OA) was observed in population-based studies (1 4). Candidate genes that may exert this genetic influence have been investigated by means of population-based association studies and genetic linkage studies in OA families. A large number of association studies using various definitions of OA have been performed, mainly investigating genes encoding structural proteins of the extracellular matrix of cartilage (e.g., COL2A1, CRTM, and AGN) or genes playing a role in the regulation of bone density and mass (e.g., VDR, IGF1, and ER), and have shown both positive and negative results for each of the genes (for review, see ref. 5). These differences in study results may indicate either false-positive or false-negative results, could be the result of differences in the OA phenotype investigated, or could be attributable to different linkage disequilibrium (LD) between the associated polymorphism and the causal mutation in various populations. Moreover, it is becoming increasingly clear that the genetic risk factors for OA at different joint sites may be distinct (6). Several groups of investigators have performed genome-wide screens of sibling pairs or nuclear families with OA and reported multiple chromosome areas (e.g., 2q13 14, 2q24 32, 2p24, 3p12, 4q32, 11q12, 9q33 34, 4q27, Xp11.3, and 7p22) that show positive linkage to specific OA phenotypes, such as radiographic OA (ROA) in the distal interphalangeal (DIP) joints, severe end-stage disease (based on hip replacement surgery), 1179

2 1180 MEULENBELT ET AL and nodal OA (7 13). When investigating end-stage disease (joint replacement surgery) (8), nodal OA (7), or DIP joint OA (11), the most consistent linkage was reported for a broad region on human chromosome 2q This particular region on chromosome 2 includes the interleukin-1 (IL-1) gene cluster. Other investigators could not confirm linkage in this region when investigating families with nodal OA of the hand and knee (which included many of the families mentioned above) (14) or specifically hand OA (15). In the latter study, however, some evidence of linkage was observed for the IL-1 receptor antagonist (IL-1Ra) gene (IL1RN). The IL-1 cluster contains 3 related genes within a 430-kb region: IL1A, IL1B, and IL1RN. These genes encode the proinflammatory cytokines IL-1 and IL-1 and their endogenous receptor antagonist IL-1Ra, respectively. Interindividual differences in cytokine production may have a genetic basis, and a proinflammatory cytokine profile has been associated with the outcomes of various infectious diseases (16,17). Several DNA variants within IL1B and IL1RN have been reported, including single-nucleotide polymorphisms (SNPs) in the promoter region ( 511C T) and in exon 5 (3953C T) of IL1B and a variable-number tandem repeat (VNTR) of an 86-bp repeat in intron 2 of IL1RN. As reported by Hall et al (18), the rare 511T allele (allele 2 in our study) in the promoter region of IL1B altered an activator protein 2 promoter element. Moreover, the effect of the 3953C T and 511C T polymorphisms on the production of IL-1 protein was extensively studied, showing a haplotype containing the C at 3953 and the T at 511 that was associated with increased secretion of IL-1 after lipopolysaccharide (LPS) stimulation alone, in 2 independent study populations. Although reports on the relationship between IL1RN VNTR genotypes and protein production have been contradictory (19,20), the second most common IL1RN allele, allele 2 (2 repeat units), has been associated consistently with various inflammatory diseases (e.g., inflammatory bowel disease, ulcerative colitis, and Helicobacter pylori associated gastric cancer) (17). IL-1 and IL-1Ra are of physiologic relevance in OA. Chondrocytes are known to respond to IL-1 by decreasing the synthesis of matrix components and increasing the synthesis of matrix metalloproteinases (MMPs). MMPs degrade extracellular matrix components in articular cartilage. IL-1Ra is the natural competitive inhibitor, acting by occupying the IL-1 cell surface receptor without triggering signal transduction. The presence of an innate proinflammatory cytokine profile may shift the metabolic balance during normal cartilage repair in favor of matrix loss, which may ultimately result in OA. Furthermore, a proinflammatory cytokine profile may influence clinical symptoms and disease progression once the OA process has been initiated (21). In this study, the effect of polymorphisms in the IL-1 gene (3953 and 511) and the IL-1Ra gene (VNTR in intron 2) was studied in relation to the occurrence of ROA in the hip, knee, and hand and disc degeneration of the spine in a random sample of 886 subjects ages years from a population-based cohort (the Rotterdam study) (22). PATIENTS AND METHODS Study population. Subjects were derived from a prospective, population-based cohort study of determinants and prognosis of chronic diseases in the elderly, the Rotterdam study (22), which comprises 7,983 Caucasian participants (response rate 78%). The current study was approved by the medical ethics committee of the Erasmus University Medical School, and written informed consent was obtained from each participant. In a random sample of persons between the ages of 55 and 65 years (n 963), radiographs were scored for the presence of OA of the knees and hips (23) and the hands and thoracolumbar spine (4). Definite OA at a particular joint site was defined as a Kellgren score (24) of 2. For hand OA, we considered subjects to be affected if they had ROA in at least 6 of 16 joint sites. This specific subgroup represents the 5% of cases of hand ROA in our study population that were most severe. Characteristics of the population (age, body mass index [BMI], and bone mineral density [BMD]) (25) were assessed at baseline. Of the 963 subjects whose radiographs were scored, 77 were excluded because of the absence of scoring of radiographs at some of the sites considered for ROA assessment. In the association analysis, controls were those subjects without ROA in the specific joint sites considered, because this control group is more robust than is the rare group that is completely negative for OA in all joint groups investigated (i.e., only 17% of persons ages years). Genotype measurements. Genomic DNA was isolated from all blood samples. In total, 886 subjects (347 women, 520 men, and 19 subjects whose sex was not recorded) were genotyped for 2 SNPs located on the IL-1 gene (3953C T and 511C T) and 1 VNTR located on the IL-1Ra gene (intron 2) (Figure 1). Analysis of the polymorphisms was based on polymerase chain reaction (PCR), which was performed using primers for 3953C T and 511C T as described by Loughlin et al (26) and for the VNTR as described by Moos et al (27). PCR was performed in a 15- l reaction volume containing 15 ng of genomic DNA. PCRs were performed in 96-well microtiter plates using recombinant Taq polymerase with the buffer supplied by the manufacturer (Amersham Pharmacia Biotech, Little Chalfont, UK). After PCR, IL1B polymorphisms were genotyped by restriction fragment length analysis. For the 3953C T polymorphism, restriction enzyme digestion of the 249-bp PCR fragment with Taq I either resulted in 2 fragments of 135 bp

3 IL-1 GENE CLUSTER POLYMORPHISMS AND PATHOGENESIS OF HIP OA 1181 Figure 1. Interleukin-1 gene cluster. Pairwise measures of linkage disequilibrium for the 3 polymorphisms are shown. VNTR variable-number tandem repeat. and 114 bp (allele C) or left the 249-bp fragment intact (allele T). For the 304-bp PCR fragment generated for the 511 polymorphism, restriction enzyme Ava I was used, which either resulted in 2 fragments of 113 bp and 191 bp (allele C) or left the 304-bp fragment intact (allele T). Restriction enzymes were purchased from Invitrogen Life Technologies (Carlsbad, CA). Restriction fragments were separated by electrophoresis through 2.5% and 3% agarose gels for 3953C T and 511C T, respectively. The IL1RN VNTR polymorphism was amplified by PCR, and the alleles were separated by electrophoresis through 1% agarose gels. The common alleles of the SNPs on IL1B, 3953C and 511C, were designated as allele 1, and the rare alleles, 3953T and 511T, were designated as allele 2. Alleles for the IL1RN VNTR were coded as follows: for allele 1, 4 repeat units of the 86-bp repeat; for allele 2, 2 repeat units; for allele 3, 5 repeat units; and for allele 4, 3 repeat units. Statistical and haplotype analysis. Characteristics of patients and controls (sex, age, and BMI) were compared using t-tests for independent samples. Allele frequencies were assessed by counting alleles and calculating sample proportions. Tests for goodness-of-fit into Hardy-Weinberg equilibrium were calculated using the HWE program of LINKUTIL ( To assess the strength of association in cases of ROA, a logistic regression model was used to estimate the odds ratios (ORs) (which are presented with 95% confidence intervals [95% CIs]). When possible, adjusted ORs were calculated using logistic regression models that adjusted for sex, age (in years), and BMI (kg/m 2 ). Age and BMI were used as continuous variables after determining that age and BMI as categorized variables did not significantly change the estimators, indicating that subjects were equally distributed over the age and BMI categories. All statistical analyses were performed with SPSS software, version 10, 2000 (SPSS, Chicago, IL). We first investigated allelic association of each locus separately. Because the marker loci are in close proximity to each other, we performed haplotype analysis to investigate whether underlying LD contributed to the nonindependence of these associations. Multiple-locus haplotype frequencies and the measures of pairwise LD were determined using the HAPLO program (28). This program implements a fairly standard method for estimating haplotype frequencies using data from unrelated individuals. It uses an expectation-maximization algorithm to calculate maximum-likelihood estimates of haplotype frequencies, given genotype measurements. The estimator of linkage disequilibrium D (where D h pq pq) indicates the difference in the observed (h pq ) and expected (pq) frequencies of haplotypes. Its maximum value depends on the allele frequencies and whether the rare alleles are associated together on a haplotype (positive value of D) or whether the common allele is associated with the rare allele (negative value of D). D (Lewontins ) is the fraction of D of its maximum (D max p pq) or minimum (D min pq) possible value. The measure r 2 is calculated as D 2 /(p [1 p] q [1 q]) (28). D and r 2 measures can take on values from 0 (no LD) to 1 (complete LD). However, D 1 when 2 alleles are in complete LD with divergent allele frequencies, whereas r 2 1 when 2 alleles are in complete LD with equal allele frequencies. The measure r 2 can be effectively used to identify redundant markers (when r 2 approaches 1) in an association study (29). Because the haplotype and genotype analyses that followed the initial allelic associations are not entirely independent tests, a strict Bonferroni correction may be too conservative. P values were, therefore, not corrected for multiple testing. RESULTS Characteristics of study population. The demographic characteristics of the study population are shown in Table 1. In total, genotyping for all polymorphisms was successful in 85% of DNA samples. Four alleles for the VNTR polymorphism were scored. Allele 1 had a frequency of 0.72, allele 2 had a frequency of 0.25, allele 3 had a frequency of 0.02, and allele 4 had a frequency of Because of the low frequency of alleles 3 and 4, we considered the VNTR polymorphism as biallelic and recoded alleles 3 and 4 to the most common allele 1. Allele 2 has been consistently associated with various

4 1182 MEULENBELT ET AL Table 1. Characteristics of the study subjects* Radiographic signs of osteoarthritis Characteristic Total None Hip Knee Hand Spine Number (%) 886 (100) 148 (17) 70 (8) 139 (16) 38 (4) 553 (62) Age, mean SD years BMI, mean SD BMD, mean SD * The total number (%) of genotypes available was as follows: for C3953T, 779 (87.9); for 511T, 723 (81.6); for ILRN, 771 (87.0). None no radiographic osteoarthritis in hip, knee, hand, or disc degeneration of the spine; BMI body mass index; BMD bone mineral density. Disc degeneration. inflammatory diseases and was considered the a priori risk allele. Allele frequencies of the IL1B and IL1RN polymorphisms. Allele distributions of the polymorphisms on IL1B (3953C T and 511C T) and IL1RN (VNTR; intron 2) were determined in the study population. Table 2 shows an allelic effect on hip ROA for the 3953 and 511 polymorphisms in IL1B and for the VNTR in IL1RN. For both the 511 and VNTR polymorphisms, the presence of allele 2 as compared with the presence of allele 1 was associated with an increased risk for hip ROA (P and P 0.001, respectively). The presence of allele 2 of 3953 was associated with a decreased risk for hip ROA (P 0.003). Haplotype analysis. To investigate whether the effect of these alleles is independent or is a result of LD between the IL1B and IL1RN alleles, haplotype frequencies of the 3 SNPs were estimated using the HAPLO program. We referred to haplotypes by naming the alleles that are present in order of the physical position within IL1B and IL1RN: the first nucleotide refers to the allele at position 3953, then 511, and finally the VNTR. Pairwise negative LD was observed for 3953/ 511 (D 0.06, D 0.72, and r ) and for 3953/VNTR (D 0.05, D 0.73, and r ). Pairwise positive LD was observed for 511/ VNTR (D 0.07, D 0.44, and r ) (Figure 1). The relatively low values of r 2 compared with the D values indicated that none of the markers can be considered redundant in an association study. Table 3 shows 2 haplotypes that are strongly associated with hip ROA: haplotype (P ) and haplotype (P ). These haplotypes carry the individual risk alleles of the VNTR and 511, respectively. A protective haplotype (P 0.012) carrying the protective allele of 3953 was also observed. The presence of 2 specific haplotypes (each carrying a different risk allele) that are strongly associated with hip ROA may indicate an independent predisposing effect of the VNTR and 511 alleles. Genotype relative risk in carriers of the interleukin gene cluster. In order to identify the risk of ROA in carriers of the interleukin alleles, genotype relative risks Table 2. Frequency of alleles for each single-nucleotide polymorphism* Phenotype IL1B 3953 IL1B 511 IL1RN VNTR Total 0.75/0.25 (1,168/390) 0.65/0.35 (941/505) 0.75/0.25 (1,153/389) No ROA 0.73/0.27 (192/70) 0.63/0.37 (156/92) 0.73/0.27 (190/70) No hip ROA 0.74/0.26 (1,055/371) 0.66/0.34 (875/447) 0.76/0.24 (1071/341) Hip ROA 0.86/0.14 (113/19) 0.53/0.47 (66/58) 0.63/0.37 (82/48) No knee ROA 0.76/0.24 (971/315) 0.65/0.35 (777/423) 0.75/0.25 (956/322) Knee ROA 0.72/0.28 (197/75) 0.67/0.33 (164/82) 0.75/0.25 (197/67) No hand ROA 0.75/0.25 (1,118/374) 0.65/0.35 (900/484) 0.75/0.25 (1,112/366) Hand ROA 0.76/0.24 (50/16) 0.66/0.34 (41/21) 0.64/0.36 (41/23) No disc degeneration 0.75/0.25 (450/148) 0.64/0.36 (354/196) 0.75/0.25 (448/146) Disc degeneration 0.75/0.25 (718/242) 0.66/0.34 (587/309) 0.74/0.26 (705/243) * Values are the frequency (number) of alleles, reported as allele 1/allele 2. VNTR variable-number tandem repeat; ROA radiographic osteoarthritis. P versus no hip ROA. P versus no hip ROA. P versus no hip ROA. P versus no hand ROA.

5 IL-1 GENE CLUSTER POLYMORPHISMS AND PATHOGENESIS OF HIP OA 1183 Table 3. Haplotype Estimated haplotype frequencies for IL1B 3953, IL1B 511, and IL1RN VNTR* Total (n 1,428) No hip ROA (n 1,306) Hip ROA (n 122) P (0.35) 479 (0.37) 28 (0.23) (0.08) 92 (0.07) 19 (0.16) (0.16) 196 (0.15) 34 (0.28) (0.16) 208 (0.16) 24 (0.20) (0.20) 275 (0.21) 14 (0.11) (0.02) 22 (0.02) 3 (0.02) (0.02) 33 (0.03) (0.00) 1 (0.00) * The common alleles of the single-nucleotide polymorphisms in the IL1B gene, 3953C and 511C, are designated as 1; the rare alleles, 3953T and 511T, are designated as 2. For the IL1RN variable-number tandem repeat (VNTR) polymorphism, allele 1 consists of the alleles with 4 repeat units, 5 repeat units, and 3 repeat units. VNTR allele 2 has 2 repeat units. ROA radiographic osteoarthritis. The total number of individuals for whom valid IL1B 511, IL1B 3953, and IL1RN VNTR genotypes were available is 714 (n 1,428 haplotypes). Hip ROA versus no hip ROA. were determined. Genotypes did not show deviations from Hardy-Weinberg equilibrium. Due to the absence of homozygotes for allele 2 of polymorphism 3953 among patients with ROA of the hip, homozygous and heterozygous carriers of allele 2 of polymorphism 3953 among the controls were pooled. A dose-response effect for hip ROA was observed among carriers of allele 2 of both the 511 and the VNTR polymorphisms. A protective effect for hip ROA was observed in carriers of allele 2 of the 3953 polymorphism (Table 4). Adjustment for sex, age, BMI, and BMD in all logistic regression analyses did not substantially alter the effects (data not shown). To test the independent effect of the 3 loci, genotypes of the polymorphisms were added simultaneously as covariates in the logistic regression analysis. Although the associations did not remain significant, the observed risk effects (ORs) for polymorphisms 3953, 511, and the VNTR did not change substantially, indicating an independent effect of the 3 polymorphisms (Table 4). When combining the risk alleles for both the 511 and VNTR polymorphisms, an additive effect was observed for subjects carrying 1 risk allele (OR 2.4, 95% CI , P ), 2 risk alleles (OR 4.1, 95% CI , P ), 3 risk alleles (OR 5.1, 95% CI , P ), or 4 risk alleles (OR 2.1, 95% CI , P ), with a significant linear-by-linear association (P ). DISCUSSION In this study, 2 polymorphisms of the IL-1 gene (3953 and 511) and 1 polymorphism of the IL-1Ra gene (VNTR in intron 2) were assessed by association Table 4. Risk effects for occurrence of hip ROA among carriers of IL1B and IL1RN alleles* Genotype (frequency) Crude OR (95% CI) Adjusted OR (95% CI) No. 1/1 1/2 2/2 Heterozygous P Homozygous P Heterozygous P Homozygous P IL1B 3953 No ROA (0.55) 48 (0.37) 11 (0.08) No hip ROA (0.55) 271 (0.38) 50 (0.07) Hip ROA (0.71) 19 (0.29) 0.5 ( ) ( ) IL1B 511 No ROA (0.40) 56 (0.45) 18 (0.15) No hip ROA (0.44) 297 (0.45) 75 (0.11) Hip ROA (0.27) 32 (0.52) 13 (0.21) 1.8 ( ) ( ) ( ) ( ) IL1RN VNTR No ROA (0.55) 48 (0.37) 11 (0.08) No hip ROA (0.57) 263 (0.37) 39 (0.06) Hip ROA (0.38) 32 (0.49) 8 (0.12) 2.0 ( ) ( ) ( ) ( ) * Adjusted odds ratios (ORs) accounted for the presence of the association of the other polymorphisms. See Patients and Methods for explanation of genotype frequency. ROA radiographic osteoarthritis; 95% CI 95% confidence interval; VNTR variable-number tandem repeat.

6 1184 MEULENBELT ET AL analysis for the occurrence of ROA of the hips, knees, and hands, and disc degeneration of the spine in a random sample of 886 subjects (ages years) from a population-based cohort (the Rotterdam study). A protective effect for hip ROA was observed in carriers of allele 2 of the 3953 polymorphism. A predisposing effect for hip ROA was observed in carriers of allele 2 of the 511 and VNTR polymorphisms (Table 2). Subsequent haplotype analysis showed that 2 separate haplotypes (1-1-2 and 1-2-1) were strongly associated with hip ROA (Table 3), indicating that these 2 haplotypes may act independently. Logistic regression analysis of genotypic data for the 3 separate loci showed a dose-response effect. This effect for each locus did not collapse when adjusted for the risk of the other loci, again indicating an independent effect of the loci (Table 4). Moreover, when combining the risk alleles, an additive effect was observed in subjects carrying increasing numbers of risk alleles of either the 511 or the VNTR polymorphism, with a linear-by-linear association (P ). Together, these data suggest an independent predisposition for hip ROA associated with allele 2 of the IL1B 511 and the IL1RN VNTR polymorphisms. Although less clear, a protective effect of allele 2 was observed for the 3953 polymorphism, which may also be independent. Haplotype (Table 3), carrying 2 individual risk alleles, was not associated with ROA of the hip. Absence of association for haplotype may occur when the predisposing functional variants are not present on this particular haplotype, which would indicate that the 511 and/or VNTR polymorphisms are not themselves functional. In spite of the significant linearbylinear association, the additive effect observed in subjects carrying increasing numbers of risk alleles at the 2 loci appears to break down in individuals carrying 4 risk alleles for ROA. The low number of individuals in this particular group, however, resulted in very large confidence interval limits (lower and upper), indicating that this risk is not accurately assessed. In contrast, the fact that such individuals are homozygous for the haplotype could indicate, once again, that this is not a risk haplotype. A relatively strong negative LD for polymorphism 3953 was observed with both 511 and the VNTR, whereas 511 showed a weaker positive LD with the VNTR. The measure D of LD, however, depends on the symbol (positive or negative). The difference observed in the measures of D of IL-1 markers may thus be explained by the fact that 3953 is in negative LD with 511 and the VNTR, whereas the 511 polymorphism is in positive LD with the VNTR. Although technically LD may differ among populations, the D measures and symbols (positive and/or negative) as observed in our Dutch population are similar to those observed in other studies (30 32). As reported by Hall et al (18), a haplotype carrying allele 2 of the 511 polymorphism in the promoter region of IL1B, together with allele 1 of 3953, was associated with increased secretion of IL-1 after stimulation of whole blood with LPS alone, in 2 independent study populations. Increased secretion of IL-1 may lead to a shift of the metabolic balance in cartilage in favor of matrix loss, which may ultimately result in OA. Furthermore, IL1RN VNTR allele 2 has previously been associated with a wide variety of inflammatory diseases (17). The association of hip ROA observed in carriers of allele 2 of the IL1B 511 and IL1RN VNTR polymorphisms appears to represent an innate proinflammatory and cartilage degradation prone profile. The observation that polymorphisms at IL1B 511 and IL1RN VNTR are associated with ROA in the hip joint only may be explained either by a relatively high response of chondrocytes in the hip cartilage to IL-1 stimulation (decreasing the synthesis of matrix components and increasing the synthesis of MMPs) or by the relatively large mechanical stresses at this joint leading to increased IL-1 production by the chondrocytes. As recently reported by Cole and Kuettner (33), chondrocytes within cartilage of the knee joint are more sensitive to IL-1 stimulation than are chondrocytes within cartilage of the ankle joint. The results of our study raise the question of whether cartilage of the hip joints may be even more sensitive to IL-1 stimulation. It has also been reported that excessive and continuous mechanical stress induces production of IL-1 by chondrocytes (34). Hip joints are subjected to especially high (multiplied) load stresses during normal walking, due to the femoral neck/lever arm effect counterbalanced by the force exerted by the abductor muscles (35). Both processes could contribute to the fact that the effects of polymorphisms within IL1B and IL1RN are most likely to be observed in patients with ROA of the hip. On the basis of some of these arguments, we would have expected to find an association with these loci in the knee joint also. The reason for this lack of association could be that the genetic effect of these loci on knee ROA is relatively small compared with the effects of other environmental risk factors. The results of several recent studies have indicated that development of OA (especially knee OA) may be influenced by secondary factors such as joint injury or BMI (36,37), which may have significantly decreased the power to

7 IL-1 GENE CLUSTER POLYMORPHISMS AND PATHOGENESIS OF HIP OA 1185 detect association between OA of the knee joint and the IL-1 gene cluster in our study. In our study population, a significant influence of BMI was observed for knee ROA (P ) but not for hip ROA (P 0.23), which supports the notion that the genetic risk factors for OA at different joint sites may very well be distinct (6). We did observe an association of IL1RN VNTR allele 2 with the presence of ROA in 6 or more joints of the hand (P 0.044). This specific group of subjects represents the 5% of cases of hand ROA in our study population that are most severe, and further investigation is needed. Previously, Leppavuori et al (11) reported linkage with chromosome 2q13 14, harboring the IL-1 gene cluster, for OA in the DIP joints. Other groups of researchers have investigated associations between the IL-1 gene cluster and the occurrence of end-stage symptomatic (joint replacement) knee or hip OA (26,27). In the study by Moos et al (27), subjects with joint replacement were subdivided into TNF low and TNF high phenotypes depending on the leading cytokine mediating the cartilage degradation process. For the IL1B 3953 polymorphism, those investigators reported a predisposing association for allele 2 in subjects with joint replacement and a TNF low phenotype; in contrast, we found a protective effect for this allele. Loughlin et al (26) did not observe an association with the IL1B 3953 polymorphism. The results reported by Moos et al (27) for the IL1B 3953 polymorphism were thus opposite to results obtained in our study. For the IL1B 511 polymorphism, which was not measured by Moos et al (27), we observed a significant association between the occurrence of hip ROA and the rare allele that is associated with increased IL-1 secretion in ex vivo experiments (18). In the study by Loughlin et al (26), however, association with the 511 polymorphism was absent. Finally, Moos et al (27) observed an association of IL1RN VNTR allele 2 and a TNF high phenotype that parallels the association with ROA in our study. Loughlin et al (26), however, demonstrated an association of the common allele of a SNP in the IL-1Ra gene (intron 3) with knee OA in men. The latter IL1RN allele was in complete LD with alleles 1, 3, and 4 of the IL1RN VNTR polymorphism (allele 1 in Tables 2 and 4). The association observed by Loughlin et al (26) thus differs from the results of our study and those of the study by Moos et al (27) with respect to the allele and the joint site that is associated. Together, the results of these studies may indicate a role for the IL-1 gene cluster in OA, although the specific allele/haplotype is unclear or has not yet been fully determined. Several reasons could account for the observed differences among studies, including falsepositive or false-negative results. The association of opposite alleles for the 3953 SNP and the VNTR observed in different studies may be a reflection of a difference in LD between the populations examined or may be explained by allelic heterogeneity. In the study by Moos and colleagues (27), subjects were subdivided into patients with TNF high or TNF low phenotypes. In our study and that by Loughlin et al (26), such a subdivision was not performed. Furthermore, it is generally known that OA may be considered a heterogeneous disease in which the correlation between radiographic and clinical symptoms is low, and different genes may play a role in the onset of various OA phenotypes. In the studies by Loughlin et al (26) and Moos et al (27), subjects had symptomatic end-stage (requiring joint replacement) knee or hip OA, whereas subjects in our study had radiographic signs of hip OA (Kellgren score of 2). Together, these differences may reflect the struggle in the study of OA to find a uniformly accepted phenotype. To further address the role of IL1B and IL1RN in OA, larger cohorts with well-characterized cases will have to be collected, which are suitable for more powerful stratification analysis according to disease site. Furthermore, several additional markers within genes should be examined to construct and compare haplotypes to delineate the optimal risk allele. Our results indicate an independent predisposition for the occurrence of especially hip ROA in carriers of the IL1B 511T allele and the IL1RN VNTR allele 2. Carriers of these alleles may represent a profile that includes proneness to cartilage degradation and/or an inflammatory response. ACKNOWLEDGMENTS We thank Leonie Willebrands for help with the genotyping and C. Bijkerk and H. Miedema for assessing Kellgren scores in the Rotterdam study. REFERENCES 1. Hirsch R, Lethbridge-Cejku M, Hanson R, Scott WW Jr, Reichle R, Plato CC, et al. Familial aggregation of osteoarthritis: data from the Baltimore Longitudinal Study on Aging. Arthritis Rheum 1998;41: Chitnavis J, Sinsheimer JS, Suchard MA, Clipsham K, Carr AJ. End-stage coxarthrosis and gonarthrosis: aetiology, clinical patterns and radiological features of idiopathic osteoarthritis. Rheumatology (Oxford) 2000;39: Wright GD, Regan M, Deighton CM, Wallis G, Doherty M. Evidence for genetic anticipation in nodal osteoarthritis. Ann Rheum Dis 1998;57:524 6.

8 1186 MEULENBELT ET AL 4. Bijkerk C, Houwing-Duistermaat JJ, Valkenburg HA, Meulenbelt I, Hofman A, Breedveld FC, et al. Heritabilities of radiologic osteoarthritis in peripheral joints and of disc degeneration of the spine. Arthritis Rheum 1999;42: Loughlin J. Genetic epidemiology of primary osteoarthritis. Curr Opin Rheumatol 2001;13: MacGregor AJ, Spector TD. Twins and the genetic architecture of osteoarthritis. Rheumatology (Oxford) 1999;38: Wright GD, Hughes AE, Regan M, Doherty M. Association of two loci on chromosome 2q with nodal osteoarthritis. Ann Rheum Dis 1996;55: Loughlin J, Mustafa Z, Smith A, Irven C, Carr AJ, Clipsham K, et al. Linkage analysis of chromosome 2q in osteoarthritis. Rheumatology (Oxford) 2000;39: Loughlin J, Mustafa Z, Irven C, Smith A, Carr AJ, Sykes B, et al. Stratification analysis of an osteoarthritis genome screen: suggestive linkage to chromosomes 4, 6, and 16 [letter]. Am J Hum Genet 1999;65: Chapman K, Mustafa Z, Irven C, Carr AJ, Clipsham K, Smith A, et al. Osteoarthritis-susceptibility locus on chromosome 11q, detected by linkage. Am J Hum Genet 1999;65: Leppavuori J, Kujala U, Kinnunen J, Kaprio J, Nissila M, Heliovaara M, et al. Genome scan for predisposing loci for distal interphalangeal joint osteoarthritis: evidence for a locus on 2q. Am J Hum Genet 1999;65: Stefansson SE, Jonsson H, Ingvarsson T, Manolescu I, Jonsson HH, Olafsdottir G, et al. Genomewide scan for hand osteoarthritis: a novel mutation in matrilin-3. Am J Hum Genet 2003;72: Demissie S, Cupples LA, Myers R, Aliabadi P, Levy D, Felson DT. Genome scan for quantity of hand osteoarthritis: the Framingham Study. Arthritis Rheum 2002;46: Gillaspy E, Spreckley K, Wallis G, Doherty M, Spector TD. Investigation of linkage on chromosome 2q and hand and knee osteoarthritis. Arthritis Rheum 2002;46: Stankovich J, Sale MM, Cooley HM, Bahlo M, Reilly A, Dickinson JL, et al. Investigation of chromosome 2q in osteoarthritis of the hand: no significant linkage in a Tasmanian population. Ann Rheum Dis 2002;61: De Jong BA, Huizinga TW, Bollen EL, Uitdehaag BM, Bosma GP, van Buchem MA, et al. Production of IL-1 and IL-1Ra as risk factors for susceptibility and progression of relapse-onset multiple sclerosis. J Neuroimmunol 2002;126: Witkin SS, Gerber S, Ledger WJ. Influence of interleukin-1 receptor antagonist gene polymorphism on disease. Clin Infect Dis 2002;34: Hall SK, Perregaux DG, Gabel CA, Woodworth T, Durham LK, Huizinga TW, et al. Polymorphic variation in the promoter region of the interleukin-1 gene correlates with interleukin-1 protein secretion levels. Submitted for publication. 19. Tountas NA, Casini-Raggi V, Yang H, di Giovine FS, Vecchi M, Kam L, et al. Functional and ethnic association of allele 2 of the interleukin-1 receptor antagonist gene in ulcerative colitis. Gastroenterology 1999;117: Danis VA, Millington M, Hyland VJ, Grennan D. Cytokine production by normal human monocytes: inter-subject variation and relationship to an IL-1 receptor antagonist (IL-1Ra) gene polymorphism. Clin Exp Immunol 1995;99: Smith MD, Triantafillou S, Parker A, Youssef PP, Coleman M. Synovial membrane inflammation and cytokine production in patients with early osteoarthritis. J Rheumatol 1997;24: Hofman A, Grobbee DE, de Jong PT, van den Ouweland FA. Determinants of disease and disability in the elderly: the Rotterdam Elderly Study. Eur J Epidemiol 1991;7: Odding E. Locomotor disability in the elderly [dissertation]. Rotterdam, The Netherlands: Erasmus University, Kellgren JH. Atlas of standard radiographs of arthritis: report of the Subcommittee on Standardization of X-ray Methods. In: Kellgren JE, editor. The epidemiology of chronic rheumatism. Vol. II. Oxford: Blackwell Scientific; Burger H, van Daele PL, Algra D, van den Ouweland FA, Grobbee DE, Hofman A, et al. The association between age and bone mineral density in men and women aged 55 years and over: the Rotterdam Study. Bone Miner 1994;25: Loughlin J, Dowling B, Mustafa Z, Chapman K. Association of the interleukin-1 gene cluster on chromosome 2q13 with knee osteoarthritis. Arthritis Rheum 2002;46: Moos V, Rudwaleit M, Herzog V, Hohlig K, Sieper J, Muller B. Association of genotypes affecting the expression of interleukin-1 or interleukin-1 receptor antagonist with osteoarthritis [published erratum appears in Arthritis Rheum 2001;44:1715]. Arthritis Rheum 2000;43: Long JC, Williams RC, Urbanek M. An E-M algorithm and testing strategy for multiple-locus haplotypes. Am J Hum Genet 1995;56: Johnson GC, Esposito L, Barratt BJ, Smith AN, Heward J, Di Genova G, et al. Haplotype tagging for the identification of common disease genes. Nat Genet 2001;29: Cox A, Camp NJ, Nicklin MJ, di Giovine FS, Duff GW. An analysis of linkage disequilibrium in the interleukin-1 gene cluster, using a novel grouping method for multiallelic markers. Am J Hum Genet 1998;62: Joos L, McIntyre L, Ruan J, Connett JE, Anthonisen NR, Weir TD, et al. Association of IL-1 and IL-1 receptor antagonist haplotypes with rate of decline in lung function in smokers. Thorax 2001;56: El Omar EM, Carrington M, Chow WH, McColl KE, Bream JH, Young HA, et al. Interleukin-1 polymorphisms associated with increased risk of gastric cancer. Nature 2000;404: Cole AA, Kuettner KE. Molecular basis for differences between human joints. Cell Mol Life Sci 2002;59: Fujisawa T, Hattori T, Takahashi K, Kuboki T, Yamashita A, Takigawa M. Cyclic mechanical stress induces extracellular matrix degradation in cultured chondrocytes via gene expression of matrix metalloproteinases and interleukin-1. J Biochem (Tokyo) 1999;125: Maquet P. Biomechanics of hip dysplasia. Acta Orthop Belg 1999;65: Manek NJ, Hart D, Spector TD, MacGregor AJ. The association of body mass index and osteoarthritis of the knee joint: an examination of genetic and environmental influences. Arthritis Rheum 2003;48: Cooper C, Snow S, McAlindon TE, Kellingray S, Stuart B, Coggon D, et al. Risk factors for the incidence and progression of radiographic knee osteoarthritis. Arthritis Rheum 2000;43: