Introduction

Chondrocalcinosis (CC) most commonly results from calcium pyrophosphate (CPP) crystal deposition (CPPD) [1]. CPPD may present as acute CCP crystal arthritis, CPPD with osteoarthritis (OA), chronic CPP crystal inflammatory arthritis, or asymptomatic CC [1]. Age, OA, diuretic use, and joint injury are recognized risk factors for CC [1, 2]. Additionally, metabolic diseases that elevate extracellular pyrophosphate (ePPi) levels (hyperparathyroidism, hypomagnesemia, and hypophosphatasia), hemochromatosis, and familial predisposition are uncommon risk factors [1, 2]. Though rare, familial predisposition is reported from several countries and different ethnic groups [3–9]. The pattern of inheritance is usually autosomal dominant. The main clinical phenotype is characterized by early onset (in the 20s or 30s) of acute CPP crystal arthritis with florid polyarticular CC and variable severity of accompanying structural arthritis/OA. However, a second phenotype with later onset in the sixth to seventh decades and oligo-articular CC that more closely resembles sporadic CPPD has also been reported [10]. This latter familial form may be more common than is recognized, the late onset of disease expression and geographic dispersal of families tending to mask such predisposition. An association with benign childhood fits appears unique to one UK family with early-onset polyarticular CC, and the responsible gene—CC gene 2 (CCAL2) on chromosome 5p15—was first identified in this family [11]. Other kindreds with CC due to mutations at this locus [3] have been reported, and the responsible gene was subsequently identified as the ankylosis human (ANKH) gene [12]. The other reported locus in an American family with premature OA and associated CPPD (CCAL1) is on chromosome 8q [13], and a specific gene predisposing to CPPD at this site has not been identified.

Although several kindreds with familial CPPD are reported, the evidence for genetic contribution to sporadic CC is conflicting [3]. For example, the -4bpG > A transition in the 5′ untranslated region (5′ UTR) of ANKH (which encodes the trans-membrane PPi transport protein ANKH) associated with CC in a study of 128 CC cases and 475 healthy controls [14], whereas hereditary contribution to CC was not detected in a larger sibling study (n = 1,841) [15]. Therefore, the objectives of this study were to validate the reported association between -4bpG > A transition in the 5′ UTR of ANKH and CC, to examine whether this is independent of other risk factors of CC such as age and OA, and to investigate whether other candidate single-nucleotide polymorphisms (SNPs) in genes involved in PPi metabolism—ANKH, tissue non-specific alkaline phosphatase (TNAP), and ectonucleotide pyrophosphatase 1 (ENPP1) ( Figure 1)—or iron overload (for example, high ferritin and transferrin) associate with CC.

Figure 1
figure 1

Extracellular pyrophosphate metabolism. ANKH, (progressive) ankylosis human; Mg++, magnesium; NTP, nucleotide tri-phosphate; PC1, plasma cell glycoprotein 1; Pi, phosphate; PPi, pyrophosphate; TNAP, tissue non-specific alkaline phosphatase.

Full size image

Methods

Study design and participants

The case-control study used data from the Genetics of Osteoarthritis and Lifestyle (GOAL) and Nottingham Osteoarthritis Case-Control (NOAC) studies [16, 17] (Table 1). The GOAL study was approved by the Nottinghamshire Research Ethics Committee (UK). The NOAC study was approved by the North Nottinghamshire Research Ethics Committee and the Leicestershire, Northamptonshire, and Rutland Research Ethics Committee. All study participants gave written informed consent. These studies, originally set up to examine the genetics of large-joint OA, recruited Caucasians with (a) hip or knee OA severe enough to require joint replacement surgery or (b) no radiographic or clinical features of hip or knee OA [16, 17]. All participants were residents of Nottinghamshire, UK. Information about disease and demographic variables were collected at the study visit. Genomic DNA was extracted from the peripheral blood leukocytes by using standard protocols. GOAL study participants had radiographs of hands, knees, and pelvis; and the NOAC study participants had radiographs of the index knee or hip joint awaiting replacement. These radiographs were scored for CC at the knees, hips, wrists, and symphysis pubis and for metacarpophalangeal joint (MCPJ) calcification by a single trained observer (Sally Doherty).

Table 1 Demographics of study participants
Full size table

Cases and controls

Cases were participants with CC at any joint, whereas controls did not have CC at any joint x-rayed.

Single-nucleotide polymorphism selection

Seventeen SNPs were selected on the basis of published associations with various phenotypes (Table 2).

Table 2 Association between all single-nucleotide polymorphisms and chondrocalcinosis in the Genetics of Osteoarthritis and Lifestyle study
Full size table

Genotyping

Genotyping was carried out at AstraZeneca laboratories in Macclesfield, UK, by using the Taqman method and at Kbioscience Ltd (Hertfordshire, UK) by using the Kompetitive Allele Specific PCR (KASPar) chemistry. All 17 SNPs were genotyped in the GOAL study. All selected SNPs in any gene which contained at least one SNP that associated with CC with an uncorrected P ≤0.10 in the GOAL study were genotyped in the NOAC study participants.

Covariates

Data about age (years), height (centimeters), and weight (kilograms) were collected at the study visit. Height and weight were used to calculate body mass index (BMI) (kg/m2). OA was defined as knee or hip OA clinically severe enough to warrant consideration of joint replacement surgery.

Statistical analysis

Mean and standard deviation (SD) and number (percentage) were used for descriptive purposes. Chi-square test and student t test were used to compare categorical and continuous variables. Cases with CC were compared with controls without CC. All SNPs were checked for Hardy-Weinberg equilibrium (HWE). Data from the GOAL and NOAC studies were pooled together for analyzing genetic risk. Genotype odds ratio (ORGENOTYPE)—the OR for association between increasing number of minor alleles of an SNP and CC—was calculated. Binary logistic regression was used to calculate aORGENOTYPE (95% confidence interval, or CI) adjusting for age (tertiles), gender, BMI (tertiles), and OA. Additionally, ORGENOTYPE was meta-analyzed with published studies by using fixed effects analysis given the lack of heterogeneity between studies. Meta-analyses were performed by using R V.2.13.1 [26]. Other analyses were carried out by using SPSSv14. Statistical significance for genetic association was set at P ≤0.003 after application of Bonferroni correction for multiple tests. Linkage disequilibrium for the four ANKH SNPs studied was estimated from unphased genotype data by using the Haploview 4.2 version [27].

Results and discussion

The descriptive characteristics of study participants are presented in Table 1. Three thousand one hundred forty-one GOAL and 1,800 NOAC study participants were included. The mean age (SD), number of females (percentage), and mean BMI (SD) of the GOAL study and NOAC study participants were 66.6 (7.9) and 70.5 (9.2) years, 1,520 (48.4%) and 1,045 (58.1%) women, and 29.3 (5.3) and 29.5 (5.6) kg/m2, respectively. GOAL study participants were significantly younger (P <0.001), were less likely to be female (P <0.001), and had similar BMI (P = 0.22) compared with the NOAC study participants.

Six hundred fifty-eight (13.2%) participants had CC at any site. All genotyped SNPs were in HWE. As TNAP571 [28] was a rare mutation in the GOAL study population (minor allele frequency = 0.01), it was excluded from further analysis. Three ANKH and one HFE SNPs associated with CC with an uncorrected P ≤0.10 in the GOAL study population (Table 2). Thus, all ANKH (-4bpG > A 5′ UTR, rs3045, rs39968, and rs875525) and HFE (rs1800562 and rs1799945) SNPs genotyped in the GOAL study were further genotyped in the NOAC study participants. The four ANKH SNPs were not in linkage disequilibrium (Table 3). Among the six SNPs genotyped in both the GOAL and NOAC studies, the -4bpG > A 5′ UTR of ANKH associated with CC after adjustment for age, gender, BMI, and OA; aORGENOTYPE was 1.39 (1.14-1.69) (P = 0.001) (Table 4). There was a trend toward an association between rs3045 and CC; aORGENOTYPE was 1.31 (1.09-1.58) (P = 0.005); however, it was not statistically significant after application of Bonferroni correction (Table 4). The frequency of rare and common allelic variants in the ANKH and HFE SNPs genotyped in both the GOAL and NOAC studies is shown in an additional file (Additional file 1: Table S1).

Table 3 Linkage disequilibrium between the genotyped single-nucleotide polymorphisms in ankylosis human ( ANKH ) gene
Full size table
Table 4 Association between chondrocalcinosis and single-nucleotide polymorphisms in the combined dataset
Full size table

Comparison with published studies

Two studies have reported on the association between CC and HFE SNPs [20, 21]. Of these, one study reported genotype data about rs1800562, and rs1799945 without reporting full data on compound heterozygote numbers [20], whereas the other did not report detailed genotype data [21]. Therefore, the latter study was not included in the meta-analysis. In a meta-analysis of the data of the present study and of the study by Alizadeh et al. [20], the values for pooled ORGENOTYPE (95% CI) for association between rs1799945 and rs1800562 and CC were 1.20 (1.04-1.39) (P = 0.015) and 1.08 (0.88-1.33) (P = 0.445), respectively. Similarly, the results of the present study and the previous study that reported an association between CC and -4bpG > A transition in the 5′ UTR of ANKH were meta-analyzed [14]. The pooled ORGENOTYPE (95% CI) was 1.36 (1.13-1.61) (P = 0.001).

The present study confirms the association between CC and -4bpG > A transition in the 5′ UTR of ANKH. It is the first study to show that this association is independent of age and OA, which are the two major established risk factors of CC. This study also raises the possibility that other SNPs in ANKH (for example, rs3045) may also associate with CC. However, in keeping with previous reports, there was no association between SNPs in TNAP or ENPP1 and CC [29].

ANKH encodes a multipass transmembrane protein (ANKH) in joints and other tissues and participates in the export of intracellular PPi [30, 31]. PPi cannot diffuse across cell membranes passively, and ANKH is the principal way in which intracellular PPi reaches the extracellular environment. ANKH-mediated control of PPi levels regulates tissue calcification and susceptibility to arthritis [30, 31]. The autosomal dominant mutations in ANKH are thought to confer a gain in PPi transport function leading to increased extracellular PPi levels [14, 32]. Functional assays show that the -4bpG > A transition in the 5′ UTR of ANKH reduces intracellular PPi (a surrogate for increased transcellular PPi export) and increases ANKH expression in vitro[14, 18]. The minor alleles of rs3045 also result in lower intracellular PPi levels in vitro, providing external validity to our finding of a possible association between this polymorphism and CC [18]. These two SNPs are not in linkage disequilibrium. The minor allele frequency for the -4bpG > A transition in 5′ UTR of ANKH and rs3045 is higher in this study than that in the multi-ethnic 1000 Genomes Project. (See Additional file 2: Table S2 for genotype frequencies of the selected SNPs in the 1000 Genomes Project). This may explain, in part, why CPPD is more common in Caucasians than in other ethnicities.

Though not statistically significant, results from the data suggest that rs1799945 (HFE SNP associated with smaller iron overload), but not rs1800562 (HFE SNP associated with greater iron overload), may associate with CC. Similar findings have been reported previously [20]. The lack of association between homozygosity for minor allele at rs1800562 and CC may be due to a channeling bias (that is, patients with hemochromatosis are excluded from these studies).

This is the largest reported study of genetic risk factors for CC. This study has several strengths. First, the analysis of genetic risk was adjusted for factors that associate with CC, specifically age and OA, and also for gender, which associates with iron overload. Moreover, correction for multiple testing was applied to reduce the chances of a type I error. However, this study has several caveats. First, this is a hospital-based study carried out by reconstituting cases and controls within cohorts assembled primarily to examine risk factors for knee or hip OA within the East Midlands region of the UK. Cases with mild to moderate large-joint OA were not included. The study sample therefore does not resemble a community-based population and is restricted to one area of the country. Moreover, as more than 78% of participants had severe large-joint OA, the results may be confounded by their OA status. However, to minimize any confounding and to improve the generalizability of these findings, we have adjusted for OA at the knee and hip. Second, participants of the NOAC study did not have the same extensive radiographic phenotyping of CC as participants of the GOAL study. As a result, some NOAC study participants who may have CC at distant joints without CC at the joint to be replaced will be misclassified as ‘CC negative’ controls. This misclassification is likely to minimize the genetic association and does not invalidate the genetic associations observed in this study.

Conclusions

This study validates an established genetic association with CC and shows that this is independent of age and OA. This study also raises the possibility that other SNPs in ANKH associate with CC. Larger studies with greater power are required to confirm these findings. Finally, the findings of this study derived from a hospital-based cohort warrant confirmation in a community-based population including cases with mild to moderate disease and in other countries.

Authors’ information

AMV and MD are joint senior authors.