SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Objective

Osteoarthritis (OA), characterized by late-onset degeneration of articular cartilage, is recognized to have a genetic component. We examined the role of 26 single-nucleotide polymorphisms (SNPs) from 24 candidate genes in OA susceptibility and progression.

Methods

We compared human complementary DNA libraries from OA-affected and normal cartilage and synovium and selected 22 genes in addition to the estrogen receptor α and vitamin D receptor genes. Based on the availability of polymorphisms, we proceeded to test whether genetic variation at those genes affected susceptibility to or progression of radiographic knee OA over a 10-year period in 749 women (mean age 64 years) from the longitudinal Chingford Study.

Results

After adjusting for age and body mass index, we observed significant associations at ADAM12, BMP2, CD36, COX2, and NCOR2 with 3 OA susceptibility traits (presence/absence of joint space narrowing [JSN], presence/absence of osteophytes, and Kellgren/Lawrence [K/L] score). For the OA progression traits (change over 10 years in the K/L score, osteophyte grade, and JSN grade), we found significant associations with ADAM12, CILP, OPG, and TNA. Overall, we observed 15 associations with nominal significance (P < 0.05) and, by permutation analysis, found that such a number would be observed by chance only 3.8% of the time. Although these tests require replication, the stronger genetic associations observed are unlikely to be attributable simply to multiple comparisons.

Conclusion

Our results suggest that OA severity and progression have a multigenic and feature-specific nature. These findings should encourage the development of genetic diagnostics for OA progression based on multiple SNPs and help unravel some of the complex disease mechanisms in OA.

Osteoarthritis (OA) is the most common cause of musculoskeletal disability related to aging and is characterized by late-onset degeneration of articular cartilage (1). An imbalance of joint functioning initiates the disease process, which is then worsened through biochemical changes in the collagen in the joint (2). At the end stage of the disease, a failure of the cellular response occurs, with a full-thickness loss of articular cartilage, thickening of subchondral bone, and the appearance of large osteophytes (3, 4). Little is known about the initiating events in OA or the relative importance of bone remodeling compared with that of cartilage degradation. Nevertheless, the degradation of cartilage matrix components is likely to be attributable to increased synthesis and activation of extracellular proteinases, with insufficient synthesis of new matrix macromolecules also being involved. Although OA is often described as a noninflammatory arthropathy, proinflammatory cytokines are now implicated as important mediators in the disease (2).

The incidence of OA is directly correlated with age (5). We know most about the role of recognized risk factors for prevalent knee OA (obesity, knee injury, and physical activity) and have some data on incident disease. However, relatively little is known about risk factors for progression (6), despite this information being of far greater interest clinically.

A genetic contribution to OA has been suggested in several epidemiologic studies (7–10). Genetic factors comprise an essential component of the etiology of OA, as the results of several studies have indicated. Heritability for OA appears to be stronger for quantity of disease than for disease prevalence, which may reflect problems of disease definition. Some twin studies have shown that 39–65% of the variation in knee and hip OA in women in the general population can be attributed to genetic factors (7, 8, 11). High heritability (61–76%) of knee cartilage volume in women, based on magnetic resonance imaging (12), has also been recently reported. Other population family studies, however, have failed to find significant genetic influences on the knee compared with other joint sites (13, 14), for reasons that are unclear.

Various chromosomal regions have shown genetic linkage with OA. Chromosomes 2, 4, and 16 were identified in multiple genome scans and are therefore likely to encode susceptibility genes (15). Association analyses of candidate genes have been also carried out for vitamin D receptor (VDR) (16–19), type II collagen (COL2A1) (16, 20, 21), type I collagen (COL1A1) (18, 21), estrogen receptor α (ESR1) (18, 22, 23), transforming growth factor β (TGFB1) (24), insulin-like growth factor 1 (IGF1) (25), and aggrecan (26). Some of these studies demonstrated positive associations between genetic variation at these loci and some OA-related phenotypes, while others failed to replicate the initial reports (although many of the initial and replication studies had inadequate sample sizes) (21). Moreover, for the most part, the chromosomal linkage regions do not encompass the major cartilage structural genes that many investigators consider to be the most likely susceptibility loci (27). Identification of the genetic pathways involved in the pathogenesis of OA still represents a substantial challenge.

Characterization of differentially expressed genes provides a powerful tool for identifying molecules that may be involved in the pathogenesis of a disease. Indeed, gene expression patterns are commonly used to discover novel genes involved in the development of cancer (for review, see refs. 28 and29) by comparing the gene expression profiles of normal and tumor cells. In this study, we compared the abundance of complementary DNA (cDNA) transcripts (copies of messenger RNA) in 4 Incyte LifeSeq (Incyte, Palo Alto, CA) libraries (normal and OA-affected synovium, and normal and OA-affected cartilage). Having selected 22 genes whose transcripts appeared to be differentially expressed in OA-affected cartilage relative to normal tissue, we screened them for polymorphisms or searched for single-nucleotide polymorphisms (SNPs) in the literature or in public domain databases. In addition to the above-mentioned genes, we also studied SNPs at the vitamin D receptor and the estrogen receptor, which have also been implicated in genetic susceptibility to osteoporosis. The genetic variation at these 24 genes was then correlated with susceptibility to and progression of radiographic knee OA, using DNA samples obtained from members of the Chingford Study cohort.

We hypothesize that genes differentially expressed in healthy and OA cartilage and synovium could contribute to OA pathogenesis, and that genetic variation at some of these genes will affect susceptibility to and/or progression of radiographic OA of the knee.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Subjects.

The subjects were 749 women ages 43–67 years at baseline (1988–1989) who were participating in the Chingford Study, a population-based study of OA and osteoporosis. The Chingford Study cohort comprises 1,003 women derived from the age/sex register of a large general practice in North London, who are similar to the UK population for most demographic variables (30). For all 749 of the subjects studied here, anteroposterior extended-view weight-bearing radiographs of the knees were obtained at baseline and ∼10 years later (range 9–11 years). Views were standardized with the backs of the knees in contact with the cassette, the patella centralized over the lower end of the femur, and the beam centered 2.5 cm below the apex of the patella, with a tube-to-film distance of 100 cm. Radiographs were read by examiners who were blinded to the clinical information, using an atlas of radiographic features to obtain a global Kellgren/Lawrence (K/L) score (0–4 scale) (7) and to evaluate individual features of osteophytes and joint space narrowing (JSN) (0–3 scale for both) (31). Progression was evaluated on an individual basis, because the genes of an individual are the subject of interest. The knee with the score reflecting the greatest severity was selected for all analyses.

The change in scores for radiographic JSN and osteophytes and the K/L grade was determined to be the difference between the score for the most severely affected knee at year 10 and the score for the same knee at baseline. That same knee was used for the cross-sectional part of the study. Progression was defined as an increase of ≥1 K/L grade in the medial or lateral compartment of the knee. Paired films were read by an investigator who was blinded to clinical details. Reproducibility of this method showed that intraobserver agreement was high for identification of osteophytes (κ = 0.88) and JSN (κ = 0.69), and interobserver agreement was good for identification of osteophytes (κ = 0.69) and JSN (κ = 0.54). Reproducibility for reading change in longitudinal radiographs yielded intraobserver agreements of κ = 0.79 for osteophytes and κ = 0.70 for JSN. Little difference in reproducibility was seen between blinded or side-by-side readings, and the main radiographic analysis was therefore performed with baseline and longitudinal films side by side (32). Height, weight, and details of concomitant diseases, operations, or medications were also recorded. Blood was stored in EDTA, and DNA was extracted by standard phenol and salt methods. At both baseline and 10 years later, all subjects completed a standardized questionnaire on medical history. The Guys & St Thomas' Trust and the Waltham Forest Trust ethics committees approved the study protocol. After study procedures were explained to participants, they gave written consent.

cDNA libraries.

The expression levels in 4 cDNA libraries from the LifeSeq database (March 2001; Incyte Pharmaceuticals, Palo Alto, CA) were compared. The information for each library provided by the manufacturer was as follows. 1) The library of normal cartilage (library name CARGNOT01; clone count 2,011) was constructed using 1.0 μg of poly(A) RNA isolated from pooled microscopically normal cartilage. The tissue was obtained from 4 donors: a 57-year-old white man who died of a heart attack, a 34-year-old white man who died of cardiac failure, a 32-year-old white man who died of a gunshot wound, and a 17-year-old woman who died of an aortic aneurysm. 2) The library of OA cartilage (library name CARGDIT01; clone count 7,229) was constructed using 1 μg of poly(A) RNA isolated from diseased cartilage tissue obtained from a 71-year-old woman whose clinical history included OA. 3) The library of normal synovium (library name SYNONOT01; clone count 4,046) was constructed using 2 μg of poly(A) RNA isolated from synovial tissue removed from a 75-year-old white man. 4) The library of OA synovium (library name SYNOOAT01; clone count 5,674) was constructed using 1 μg of poly(A) RNA isolated from knee synovial membrane tissue obtained from an 82-year-old woman with OA. Although age could confound this part of the study, if the differences of expression abundance were attributable solely to age, we would then be studying genes involved in chondrocyte senescence and thus potentially involved in OA pathogenesis.

The expression pattern of 54 genes was considered to be substantially different between the cDNA cartilage libraries of OA and normal tissue. The list of such genes is as follows: AACT, ACLP, ADAM9, ADAM12, ADLICAN, AGC1, BMP2, CAPN4, CD36, CDO1, CHI3L2, CHM-I, CILP, COL10A1, COL11A2, COMP, CRTL1, CTSL, DAF, FGF1, FMOD, FST, G0S2, GADD34, IBSP, IER3, IGFBP6, IHH, INHBA, JUN, LIF, LUM, METTL1, MIA, NCOR2, OGN, OMD, OPG, PDCD6IP, PPP1R5, PRELP, PRSS11, RIN1, SCRG1, SCYA20, SDC2, SDC4, SOD2, SOD3, STATI2, TEM1, TNFAIP6, TNRC3, and TNA. A higher priority for SNP screening was assigned to those genes with sequence homology to molecule classes commonly accepted to be potentially good therapeutic intervention targets (e.g., receptors, enzymes, secreted molecules). Based on these 2 criteria and budget constraints, only 12 genes were actually screened. The SNPs available in the public domain were selected for another 10 genes of interest.

SNP discovery.

SNPs in AACT, ACLP, ADLICAN, BGN, BMPR1A, COMP, COX2, CTSL, DAF, MMP3, OGN, and SOD3 were identified using the Incyte Pharmaceutical proprietary fluorescence-based single-strand conformation polymorphism (SSCP) method. Fluorescence-labeled primers were synthesized, and polymerase chain reaction (PCR) was performed on 47 DNAs from a Coriel derived Human Diversity Panel. The PCR products were electrophoresed on an ABI Prism 377 sequencer (Applied Biosystems, Foster City, CA), and 8% nondenaturing, 12-cm SSCP gels were used. The resulting traces were aligned in ABI Prism Genotyper software (Applied Biosystems), and, where variant traces (indicating an underlying polymorphism) were found, an example of each variant type was sequenced.

Genotyping.

Genomic DNA was set on 384-well plates. Genotyping was carried out using Assays-on-Demand SNP Genotyping products from Applied Biosystems, in which the primers are labeled with a reporter dye at the 5′ end of each probe: the VIC dye was linked to the 5′ end of one of the allele's probe, and the FAM dye was linked to the 5′ end of the other allele's probe. The PCR amplification was carried out in a total reaction volume of 5 μl using reagents supplied by the vendor containing both probes and primers (Applied Biosystems). Amplification was carried out in a KBiosystems Super Duncan thermal cycler (Essex, UK), using the specific temperature cycling profile shown below. Prereaction incubations were as follows: 95°C for 10 minutes (AmpliTaq Gold [Applied Biosystems] activation, template denaturation) for 40 cycles, denaturation at 95°C for 15 seconds, anneal/extend at 60°C for 60 seconds. Cleavage by the AmpliTaq Gold DNA polymerase separates the reporter dye from the quencher dye, which results in increased fluorescence by the reporter. The increase in fluorescence signal occurs only if the amplified target sequence is complementary to the probe. Thus, the fluorescence signal generated by PCR amplification indicates which alleles are present in the sample. A substantial increase in VIC dye fluorescence indicates homozygosity for allele 1. A substantial increase in FAM dye fluorescence indicates homozygosity for allele 2. A substantial increase in both fluorescence signals indicates allele 1–allele 2 heterozygosity. Each 384-well plate was read using SDS instrumentation, and alleles were called using the SDS software package (Applied Biosystems).

In addition to the 22 genes shown in Table 1, we also included SNPs from 2 well-characterized osteoporosis candidate genes, namely, ESR1 (ERα; map position 6q25.1; accession number NM_000125) and VDR (map position 12q12–q14; accession number NM_000376). The ESR1Pvu II polymorphism falls within intron 1, whereas the VDRTaq I polymorphism encodes a synonymous change in exon 9. Both of these polymorphisms have been analyzed in previous studies for their relationship with OA susceptibility (11, 17, 18).

Table 1. Abundance of expression of 22 genes in 4 cDNA libraries from normal and OA-affected tissue*
Locus symbolGenBank accession no.Gene nameMap positionOA synoviumOA cartilageNormal synoviumNormal cartilage
  • *

    Values in synovium and cartilage are the expression abundance, or frequency. cDNA = copies of messenger RNA; OA = osteoarthritis.

AACTNM_001085Serine proteinase inhibitor, clade A (α-1 antiproteinase, antitrypsin), member 314q32.10.550.40
ACLPNM_001129AE binding protein 1; aortic carboxypeptidase-like protein7p130.120.06
ADAM12NM_003474A disintegrin and metalloproteinase domain 12 (meltrin α)10q26.30.05
ADLICANNM_015419Adhesion protein with leucine-rich repeats and immunoglobulin domains related to perlecanXp22.330.06
BGNNM_001711BiglycanXq220.140.240.120.20
BMP2NM_001200Bone morphogenetic protein 220p120.110.03
BMPR1ANM_004329Bone morphogenetic protein receptor; type IA10q22.30.03
CD36NM_000072CD36 antigen (type I collagen receptor; thrombospondin receptor)7q11.20.140.10
CILPNP_003604Cartilage intermediate-layer protein, nucleotide pyrophosphohydrolase15q220.050.190.25
COMPNM_000095Cartilage oligomeric matrix protein19p13.10.121.190.94
COX2NM_000963Prostaglandin-endoperoxide synthase 2 (prostaglandin G/H synthase and cyclooxygenase)1q250.04
CTSLNM_001912Cathepsin L9q21–q220.120.100.01
DAFNM_000574,  AY055758Decay-accelerating factor for complement (CD55, Cromer blood group system)1q320.620.03
IBSPNM_004967Integrin-binding sialoprotein (bone sialoprotein, bone sialoprotein-II)4q21–q250.04
MMP3NM_002422Matrix metalloproteinase 3 (stromelysin 1, progelatinase)11q22.35.172.130.50
NCOR2NM_006312Nuclear receptor corepressor 212q240.03
OGNNM_033014Osteoglycin (osteoinductive factor, mimecan)9q220.08
OPGNM_002546Tumor necrosis factor receptor superfamily, member 11b (osteoprotegerin)8q240.590.10
SOD3NM_003102Superoxide dismutase 3, extracellular4p160.120.40
TIMP1NM_003254Tissue inhibitor of metalloproteinase 1 (collagenase inhibitor)Xp11.3–p11.230.330.180.570.05
TNANM_003278Tetranectin (plasminogen-binding protein)3p22–p21.30.090.010.07
TNFAIP6NM_007115Tumor necrosis factor α–induced protein (hyaluronate-binding protein, TSG6)2q23.30.040.10

Statistical analysis.

Genetic associations.

The association between SNP genotypes and the presence/absence of osteophytes, radiographic OA (defined as a K/L score of ≥2), and JSN was tested using a logistic regression model that included age and body mass index (BMI) as covariates. Genetic association with the presence of OA was also tested, although 94 samples with a K/L score of 1 were excluded because evidence suggests that 50% of patients with K/L grade 1 will go on to develop full OA (32).

Analyses of variance (ANOVAs) were used to compare the mean progression of radiographic grade (change over 10 years) between SNP genotypes. Age and BMI were used as covariates in the ANOVA. After adjustment for age and BMI, smoking status and knee injury did not show a significant or nearly significant correlation with any of the traits studied. Less than 2% of participants reported knee injury, which might explain the lack of significance of this correlation, and such a small number of injury cases is unlikely to confound any genetic associations.

The outcome variables used were the change over 10 years in K/L score, osteophyte grade, and JSN grade. For the binary traits, the odds ratios (ORs) derived by logistic regression, with 95% confidence intervals (95% CIs), are presented. For the quantitative traits (change over 10 years in radiographic grade), the adjusted means of the radiographic grade of each trait (with standard error) are shown.

Multiple comparisons and permutation test.

To correct for the effects of multiple testing, we used a resampling permutation method. Permutation methods are well established as a robust approach for obtaining overall significance levels while minimizing Type II errors (see refs. 33–35). Such methods can be extended to multiple testing scenarios (36), and examples of their application to human genetics are not uncommon (see ref. 37).

In this study, the permutation tests rely on the assumption that the OA-related phenotypes of an individual are fixed, and the null hypothesis (i.e., no genetic association) is that the genotypes at the loci studied have no effect on the clinical traits. Under the null hypothesis of no genetic association, the genotypes of each individual should be exchangeable. We randomly shuffled the observed clinical values (keeping the 6 phenotypic traits together to preserve the correlation between them) over the 26 SNP genotypes of the 24 genes studied (keeping the 26 genotypes of each individual together) and computed the test statistics in these new samples. The test statistics used were a chi-square for binary traits comparing genotype frequencies and an F value from the ANOVA for the quantitative traits (change in K/L score, osteophyte grade, and JSN grade). This procedure was repeated 500 times, generating an empirical distribution of the test under the hypothesis of no association between clinical traits and SNP genotypes. The P value corresponding to each of these test statistics given the number of degrees of freedom was then computed. Two separate questions were addressed: first, what is the probability of observing by chance the total number of P values less than 0.05 that are observed in the true set, and second, what is the probability of observing by chance an SNP associated at a certain significance level with a certain number of phenotypes. All statistical analyses were performed using the S-Plus software package (release 3, 2000; Insightful, Seattle, WA).

RESULTS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Candidate genes and SNPs.

The abundance of 22 genes in cDNA libraries derived from OA and normal cartilage and OA and normal synovium is shown in Table 1. The genetic polymorphisms that were found in those 22 genes plus SNPs at ESR1 and VDR are shown in Table 2. Genotyping was carried out at these SNPs in 749 samples obtained from subjects in the Chingford cohort. No statistically significant deviations from Hardy-Weinberg equilibrium were detected.

Table 2. SNPs genotyped and analyzed for association with knee OA*
GeneSNP descriptionSNP referenceSNP aliasMinor alleleFrequency, %HWEFlanking sequence
  • *

    The single-nucleotide polymorphism (SNP) alias comprises the gene symbol followed by an underscore and the position of the polymorphism in the gene. OA = osteoarthritis; HWE = Hardy-Weinberg equilibrium; Ins = GTG; UTR = untranslated region.

AACTAla (G)9Thr(A)rs4934AACT_9A49.90.12AATGGAGAGAATGTTACCTCTCCTG[A/G]CTCTGGGGCTCTTGGCGGCTGGGTT
AACTC > G intron 1Present studyAACT_intG44.70.75GGGACTCTGGGCACTTCCACTGCTG[C/G]GGAAGCAGGGTCGAGCAGGGCGACC
ACLPC > T intron 18Present studyACLP_intT25.10.94TACAGACCCAAAGTCAGCCCCTCTC[C/T]GGACCAGGCCCCGCCCACAGCCCCT
ADAM12Gly(G)48Arg(C)rs3740199ADAM12_48G43.70.45TGATGAAGTTGTCAGTGCCTCTGTT[G/C]GGAGTGGGGACCTCTGGATCCCAGT
ADLICANGly(C)2663Asp(T)Present studyADLICAN_2663C33.10.45TTCCCCAGAAGTGGTAACAGGTATG[C/T]CTGTCTGTAGCCTCCCAGGCTGGGCC
BGNG > T intron 7Present studyBGN_intT46.30.14CCTCTGGTGTGGCCCTTGAAGTAGG[G/T]TGCAGAGGCAACAGCAAAATGCCTCC
BMP2Ser(T)87Ser(C)rs1049007BMP2_87C38.50.31CTGGGGCGGGTGAGCCCGGCTGACC[C/T]GAGTGCCTGCGATACAGGTCTAGCAT
BMP2Ser(T)190Arg(A)rs235768BMP2_190A38.80.83GATTCACCAACCTGGTGTCCAAAAG[A/T]CTGGTCACGGGGAATTTCGAGTTGGC
BMPR1ADel > Ins intron 10Present studyBMPR_intIns12.30.92GAAAAACATACATCTACTATTAAGA(Ins/Del)AATCATAGTGTCTATATTTTTTCTTGT
CD365′-UTR A > C (−120)rs1049654CD36_5pA43.30.99GATTCTTTCTGTGACTCATCAGTTC[A/C]TTTCCTGTAAAATTCATGTCTTGCTG
CILPThr(C)395Ile(T)Nakamura et al (49)CILP_395T37.70.86TCCAAGGTTGCCCAGCTGATTGTCA[C/T]AGGTAAGCCTGTCTGGGTCCCTGGG
COMPAsn(A)386Asp(G)Present studyCOMP_386G4.90.92CGACCGGATCCGCAACCAGGCCGAC[A/G]ACTGCCCTAGGGTACCCAACTCAGA
COX2Val(G)102Val(C)Present studyCOX2_102C17.00.93TTCGAAATGCAATTATGAGTTATGT[C/G]TTGACATGTAAGTACAAGTGTCTTT
CTSLC > T intron 10Present studyCTSL_intT46.00.94TGAAACTTCCCCAGAAAGAATAGTC[C/T]TGGCTGTTGAGAAGTTTTAGTCAGAG
DAFA > G intron 7Present studyDAF_intG32.90.32TTTACGCAGAGTCCTTCAGCAGCAC[A/G]TAAGTCCACTAATGTACATTCCCCA
ESR1T > C intron 1Yaich et al (50)ESR1_intC47.40.56GAAAACGGCAACGGCAGCAGCGGAG[A/G]AGACAATGGAGAAGAAGGGGAAGAA
IBSPGly(A)195Glu(G)rs1054627IBSP_195A28.30.97TCATCTGAGTTCCAAATGTCCCAGC[T/C]GTTTTATGCTTTGTCTCTGTTTCCC
MMP3C > T intron 4 (5766)Present studyMMP3_int4C47.80.12ATAAATTGGTCCCTATTTAAGAAAT[C/T]GAGAACAATAGTTACTTATTTTTTA
NCOR2Thr(A)1699Ala(G)rs2229840NCOR2_1699G16.40.11CACCTCGCAGCAGATGCACCACAAC[A/G]CGGCCACCGCCATGGCCCAGCGAGCT
OGN3′-UTR A > GPresent studyOGN_3pG4.70.43TGTACTTTCATTTATATGTTGTACC[A/G]ATAGAGGTTAAAAGTATGACCCTATC
OPG5′-UTR C > Trs1564858OPG_5pT11.10.57GGATTTGGAGTGGTGCAAGCTGGTA[C/T]GTGTCAATGTGCAGCAAAATTAATTA
SOD33′-UTR C > TPresent studySOD3_3pT33.40.74CCACTCTGAGGTCTCACCTTCGCCT[C/T]TGCTGAAGTCTCCCCGCAGCCCTCTCC
TIMP1Leu(C)124Leu (T)rs1043945TIMP1_124T47.70.07AGCTCAGGCTGTTCCAGGGAGCCAC[A/G]AAACTGCAGGTAGTGATGTGCAAGAGT
TNASer(A)106Gly(G)rs13963TNA_106G38.90.13CTGCATCTCGCGCGGGGGCACCCTG[A/G]GCACCCCTCAGACTGGCTCGGAGAAC
TNFAIP6Arg(G)144Gln(A)rs1046668TSG_144G13.50.83GGTGGCGTCTTTACAGATCCAAAGC[A/G]AATTTTTAAATCTCCAGGCTTCCCA
VDRIle(T)365Ile(C)rs 731236VDR_365C41.00.99CTGGGGTGCAGGACGCCGCGCTGAT[T/C]GAGGCCATCCAGGACCGCCTGTCCAACA

OA rates in the cohort.

Table 3 shows characteristics of the 749 female subjects (mean age 63.9 years; mean BMI at 10 years 26.6 kg/m2) and also for the subsets of patients with and without radiographic OA. The prevalence of radiographic OA was 21.2% at baseline and 38% by the tenth year of followup. This corresponds to an incidence of ∼2% per year, which is comparable with published data from other studies (5, 6). The phenotypes that we set out to study (prevalence of OA, presence of osteophytes, presence of JSN, change in K/L grade, change in JSN grade, change in osteophyte grade) are significantly correlated (see Table 4). However, although these clinical characteristics are not independent, each of these traits provides some nonredundant information that can be affected by an individual's genotype.

Table 3. Characteristics of the 749 Chingford Study participants whose DNA was used in this study*
CharacteristicTotal (n = 749)Normal (n = 469)OA (n = 280)
  • *

    Except where indicated otherwise, values are the mean or the mean ± SD. Individuals were classified as normal if they did not have disease in either knee at both examinations, and as having osteoarthritis (OA) if they had a Kellgren/Lawrence (K/L) score of ≥2. Changes in the K/L score, osteophyte grade, and joint space narrowing grade were calculated as the difference between the grade at the most severely affected knee at year 10 minus the grade at baseline.

Age in 2000, years63.87 ± 6.1663.6766.90
Body mass index at year 1, kg/m225.27 ± 4.0024.6026.40
Body mass index at year 10, kg/m226.65 ± 4.6525.9027.90
Age at menopause, years48.12 ± 50.348.0048.30
% with joint space narrowing23.115.735.3
Change in K/L score0.56 ± 0.880.181.18
Change in osteophyte grade0.38 ± 0.390.000.52
Change in joint space narrowing grade0.11 ± 0.270.030.26
Table 4. Correlation between phenotypic traits in this study*
 AgeBMI, year 1OA, yes/noΔ K/L scoreΔ osteophyte gradeΔ JSN gradeOsteophytes, yes/noJSN, yes/no
  • *

    Upper diagonal shows the squared correlation coefficient (R2); lower diagonal shows the P values of correlation. BMI = body mass index; OA = osteoarthritis; Δ = change in; K/L = Kellgren/Lawrence scale; JSN = joint space narrowing; NS = not significant.

Age 0.0100.0650.0040.0200.0000.0630.002
BMI at year 10.007 0.0540.0280.0580.0190.0590.009
OA<1 × 10−6<1 × 10−6 0.3010.4130.0140.9690.048
Δ K/L score0.08<1 × 10−6<1 × 10−6 0.4090.0220.1960.030
Δ osteophyte grade0.002<1 × 10−6<1 × 10−6<1 × 10−6 0.0310.4260.037
Δ JSN gradeNS0.00030.0010.00013 × 10−6 0.0120.322
Osteophytes, yes/no<1 × 10−6<1 × 10−6<1 × 10−6<1 × 10−6<1 × 10−60.004 0.046
JSN, yes/noNS0.013<1 × 10−65 × 10−6<1 × 10−6<1 × 10−6<1 × 10−6 

Genetic association results.

For the 3 OA susceptibility traits (presence/absence of JSN, osteophytes, radiographic OA), we found significant associations at ADAM12, BMP2, CD36, COX2, and NCOR2. For the 3 OA progression traits (change over 10 years in K/L score, osteophyte grade, and JSN grade), we found significant associations with SNPs in ADAM12, CILP, OPG, and TNA.

The extent and direction of the genetic associations (adjusted for age and BMI) for the genes for which a nominally significant (P < 0.05) or nearly significant (P < 0.07) association with knee OA susceptibility and progression was observed are shown in Tables 5 and 6. Three genes were associated with susceptibility to radiographic knee OA measured as either a K/L score of ≥2.0 or the presence of osteophytes. These genes are ADAM12, NCOR2, and CD36. The candidacy of ADAM12 is particularly strong given the low P values for a range of phenotypes. The genetic association with the presence of radiographic OA was tested by comparing individuals with a K/L score of <2 and those with a K/L score of ≥2, as shown in Table 7. Excluding samples with a K/L score of 1 from the analysis did not affect our results substantially, except for CILP_395, which appeared to be significantly associated with OA (P < 0.022) if the samples with a K/L score of 1 were excluded but was not significantly associated when the samples with a K/L score of 1 were included (P < 0.10). In terms of JSN, in addition to NCOR2 we observed a strong significant association with 2 BMP2 polymorphisms and a more modest association with COX2.

Table 5. Magnitude and direction of genetic associations with OA progression*
Trait, SNP aliasGenotypeMean change in gradeSEF(df)P
  • *

    OA = osteoarthritis; SNP = single-nucleotide polymorphism; df = degrees of freedom; K/L = Kellgren/Lawrence; JSN = joint space narrowing.

  • Values are for the comparison between genotypes.

Change in K/L grade     
 ADAM12_48GG0.4480.0703.29(1,737)<0.070
 GC + CC0.5900.035  
 CILP_395CC0.4740.0545.30(1,689)<0.021
 CT + TT0.6440.044  
 TNA_106AA0.6980.0803.58(1,720)<0.058
 AG + GG0.5390.036  
      
Change in osteophyte grade     
 ADAM12_48GG0.2290.0608.31(1,737)<0.004
 GC + CC0.4240.030  
 ESR_intTT0.4710.0533.53(1,732)<0.060
 TC + CC0.3530.032  
 OPG_5pGG0.3550.0313.95(1,711)<0.047
 GT + TT0.4920.067  
 TNA_106AA0.5120.0706.13(1,720)<0.014
 AG + GG0.3570.030  
      
Change in JSN grade     
 AACT_9GG0.2130.0404.13(2,695)<0.016
 GA0.0810.031  
 AA0.0720.040  
 TNA_106AA0.1020.0243.54(1,720)<0.06
 AG + GG0.0550.011  
 TSG_144AA0.1320.0253.31(1,697)<0.07
 AG + GG0.0420.042  
Table 6. Magnitude and direction of genetic associations with OA susceptibility*
Trait, SNP aliasGenotype (no.)% with traitOR (95% CI)Wald χ2P
  • *

    OA = osteoarthritis; SNP = single-nucleotide polymorphism; OR = odds ratio; 95% CI = 95% confidence interval; JSN = joint space narrowing.

  • Values are for the comparison between genotypes.

  • Radiographic OA was defined as a Kellgren/Lawrence grade ≥2.

Presence of osteophytes     
 ADAM12_48GG (0)26.71.92 (1.26–2.91)9.35<0.002
 GC + CC (1)40.3   
 CD36_5pAA (0)42.10.76 (0.60–0.95)5.96<0.015
 AC (1)39.2   
 CC (2)32.1   
 NCOR2_1699GG + GA (0)28.81.60 (1.11–2.31)6.47<0.011
 AA (1)40.5   
      
Radiographic OA     
 ADAM12_48GG (0)27.31.84 (1.22–2.79)8.39<0.004
 GC + CC (1)40.3   
 CD36_5pAA (0)41.40.77 (0.61–0.96)5.43<0.020
 AC (1)39.2   
 CC (2)32.1   
 NCOR2_1699GG + GA (0)29.31.57 (1.09–2.25)5.88<0.015
 AA (1)40.5   
      
Presence of JSN     
 BMP2_87TT (0)17.91.68 (1.15–2.46)7.18<0.007
 CT + CC (1)27.2   
 BMP2_109TT (0)17.71.72 (1.18–2.52)7.91<0.005
 AT + AA (1)27.5   
 COX2_102CC + CG (0)28.90.66 (0.46–0.95)4.88<0.027
 GG (1)21.3   
 NCOR2_1699GG + GA (0)17.31.54 (1.02–2.32)4.23<0.040
 AA (1)24.9   
Table 7. Genetic association between polymorphisms at OA candidate genes and 6 OA progression and susceptibility traits*
GeneSNPPresence of JSNPresence of osteophytesOA (K/L score ≥2)Change in JSN gradeChange in osteophyte gradeChange in K/L scoreSNP from permutation test
  • *

    Associations are expressed as P values. Only those single-nucleotide polymorphisms (SNPs) that showed a nominally significant association (P < 0.05) are shown. SNPs that had P values less than 0.05 for association with radiographic osteoarthritis (OA), observed excluding from analysis the 94 samples with a Kellgren/Lawrence (K/L) score of 1, were as follows: ADAM_48 (P < 0.015), CD36_5p (P < 0.049), CILP_395 (P < 0.022), and NCOR_1699 (P < 0.046).

  • JSN = joint space narrowing; NS = not significant.

  • Probability of a SNP showing as many significant P values at a given level in 500 permutations.

AACTAACT_90.013NS
ADAM12ADAM12_480.0020.0040.0040.014
BMP2BMP2_870.007NS
BMP2BMP2_1900.005NS
CD36CD36_5p0.0150.020NS
CILPCILP_3950.021NS
COX2COX2_1020.027NS
NCOR2NCOR2_16990.0400.0110.0150.142
OPGOPG_5p0.047NS
TNATNA_1060.014NS

We found several genes that appeared to influence OA progression, measured as the change in K/L grade over 10 years. The polymorphisms at ADAM12, CILP, and TNA appeared to correlate with change over time in the K/L score, although only the Thr>Ile change in CILP reached statistical significance. The SNPs at ADAM12, ESR1, OPG, and TNA all affected change in the osteophyte grade. The strongest association was with the ADAM12 genotype (P < 0.004): carriers of 1 or 2 arginines at this site (CC or GC) had almost twice the change in osteophyte grade as did the glycine homozygotes. Finally, only the Ala>Thr change at codon 9 of AACT was significantly associated with change in the JSN grade (P < 0.02). Polymorphisms at TNA and TNFAIP6 also showed some effect on this trait, although not a statistically significant one.

Overall, we observed 15 associations with nominal significance (P < 0.05). By permutation analysis we found that the expected number of P values less than 0.05 that would be observed by chance if the genotypes and phenotypes were unassociated would be 7.09. The probability of observing 15 or more P values less than 0.05 was 0.038, thus rejecting the hypothesis that all of these many significant associations are spurious. In terms of individual SNPs, the P values derived from permutation analysis for a SNP showing as many associations at the observed significance levels are presented in Table 7. With the exception of the SNP in ADAM12 (P < 0.014), we cannot exclude the possibility of the other genetic associations being attributable purely to chance.

DISCUSSION

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

In this study, we have examined the role in OA susceptibility and progression of genetic variation at many candidate genes not previously explored. Having selected most of those genes based on their expression pattern would suggest that, in principle, these genes should have a role in the pathogenesis of OA. Although this list of genes likely to be involved in OA is by no means exhaustive, we consider it an important first step toward evaluating the role of multiple genes in the pathogenesis and development of knee OA.

We have identified 9 genes whose genotype correlates with knee OA susceptibility and/or progression and could have potential diagnostic or therapeutic value by pointing to possible new disease mechanisms. The study has illustrated the diversity and complexity of the genetics of OA and highlights the difficulties of picking candidates a priori on the basis of likely physiologic function.

Several potential limitations to the study must be mentioned. Along with the possibility of observing false-positive results, we must also consider the possibility of observing false-negative results. In particular, for most genes only 1 polymorphism has been studied, and thus it is quite feasible that genetic variation at these genes could still be involved in susceptibility and/or progression to OA; by not studying every variant at each gene and the haplotype combinations, we might have missed such effects. Therefore, our inability to detect a genetic association at any given gene does not preclude a potentially important role for that gene in the susceptibility to or progression of OA. To more fully test the potential of the more interesting genes highlighted by this research, a followup study that explores the role of haplotypes on the genetic susceptibility to OA is clearly desirable.

Radiologic classification of progression using traditional extended views of the knee in the standing position is now considered suboptimal for clinical trials, but the uniquely long time course of this study lessens the minor problems seen with shorter 2–3-year studies (38). We did not have true quantitative data, because scoring systems are semicontinuous, and these usually fit standard models less well than do fully continuous data. Furthermore, any errors of classification would tend to be random, because the scoring was blinded to genotyping and therefore favors a null result.

Of the 9 genes associated with OA traits, 6 were associated with only 1 trait. It is likely that some of these associations were observed because of the multiple tests carried out, as the corrected P values in Table 7 indicate. In contrast, with only 1 exception (see Table 4), the proportion of the variance in an OA-related trait that can be explained by any of the other OA-related phenotypes in this study, even if highly statistically significant, was <50%. This suggests that to some extent, different risk factors, and therefore potentially different genetic risk factors, could be at play in the various phenotypes. The strongest test of the value of the current genetic associations (as for any similar study) lies not in the P values or biologic plausibility but in the ability to reproduce some of these in independent cohorts. Lohmueller and coworkers (39) have suggested guidelines to interpret the value of genetic association results. In their opinion, a single nominally significant result should be viewed as tentative until it has been independently replicated.

In this study, we observed strong genetic associations at genes not previously studied for their role in OA. The most striking was the association between an amino acid change at ADAM12 (meltrin α) and several progression and susceptibility features. This gene encodes a metalloprotease and has been shown to be important in mediating cellular interactions and responses (40). It appears to regulate the formation of macrophage-derived giant cells, possibly by mediating the effects of 1,25-hydroxyvitamin D3 on cell–cell fusion (which itself has been implicated in OA). Moreover, the addition of antisense ADAM12 messenger RNA to mononuclear osteoclast precursor cells results in a 50% decrease in giant cell formation (41). Its role in osteoclast formation might explain the association of this gene with osteophyte presence and progression. Another interesting gene to emerge from our study is NCOR2, which is a silencing mediator for retinoid and thyroid hormone receptors and also a modulator of both basal and ligand-activated transcription of genes controlled by RAR/RXR heterodimers in a dose-dependent manner (42, 43).

In addition, the genotype at BMP2 appeared to affect the presence and extent of JSN. This gene has recently been implicated as being responsible for a strong linkage signal to bone mineral density (BMD) in a number of large Icelandic families (44). The inverse relationship between BMD and OA and the likely genetic correlation (45, 46) suggest that genes such as BMP2 may have a dual role in both diseases. Thus, the data presented here provide an excellent framework for other researchers who are interested in examining the genetic influences of OA susceptibility and progression.

Polymorphisms considered as haplotype combinations at ESR1 have been recently associated with radiographic OA of the knee in the elderly (22), and in particular with the presence of osteophytes. The association with the presence of osteophytes in this study supports the results found in the Rotterdam cohort and adds weight and validity to our other findings.

We were not able to detect any significant association between the VDR SNP studied here and any of the traits. Polymorphisms at this gene have been previously implicated in the risk of OA (47) and with the presence of osteophytes (16) in particular. In an earlier analysis (in a sample that partially overlaps with this one in this study), we observed a positive association for the Taq restriction fragment length polymorphism and knee osteophytes (17). Differences between the genetic markers, the lack of full coverage, and the lower power of SNP analyses may have produced the discrepancy. However, other authors have been unable to reproduce such VDR associations (18, 19, 48), and the current position is unclear.

None of the genes studied here showed a very large effect on any of the traits. In fact, the largest OR observed in this study was 1.92, for ADAM12 and the presence of osteophytes. Such results indicate that multiple genes with differing effects on cartilage, bone, and inflammation and small effects individually are involved in OA susceptibility and progression and are likely to be feature specific. Nonetheless, it is possible that combinations of genes might prove more useful than any individual polymorphism for detecting large effects on the risk of developing OA. This study raises the hope of the emergence of a genetic diagnostic technique to predict OA susceptibility and progression based on multiple SNPs. It also illustrates the complexity and diversity of disease mechanisms involved in OA.

Acknowledgements

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Justin Brooking and Andrew Dearlove are acknowledged for their contribution in the genotyping. We thank Paula Smith and Tom Weaver for their assistance and support during earlier stages of this project. Katrin Scurrah and Toby Andrew provided comments on the manuscript. We are grateful to the staff and patients of Chingford Hospital for their support—especially Maxine Daniels and Dr. Alan Hakim. We thank 2 anonymous reviewers for their valuable comments and suggestions on an earlier version of the manuscript.

REFERENCES

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES