Research Article

Role of the Mouse ank Gene in Control of Tissue Calcification and Arthritis

See allHide authors and affiliations

Science  14 Jul 2000:
Vol. 289, Issue 5477, pp. 265-270
DOI: 10.1126/science.289.5477.265


Mutation at the mouse progressive ankylosis(ank) locus causes a generalized, progressive form of arthritis accompanied by mineral deposition, formation of bony outgrowths, and joint destruction. Here, we show that theank locus encodes a multipass transmembrane protein (ANK) that is expressed in joints and other tissues and controls pyrophosphate levels in cultured cells. A highly conserved gene is present in humans and other vertebrates. These results identify ANK-mediated control of pyrophosphate levels as a possible mechanism regulating tissue calcification and susceptibility to arthritis in higher animals.

Arthritis is one of the most common human health afflictions, affecting 50% of people over 65 and accounting for $100 billion in medical costs and lost productivity each year (1). As the population ages, the health and economic tolls of arthritis are predicted to increase dramatically in the future (2). Despite the tremendous individual and social impact of arthritis, its underlying causes are not well understood. Twin studies suggest that genetic factors account for half to two-thirds of human arthritis cases, including the most common types such as osteoarthritis, rheumatoid arthritis, and ankylosing spondylitis (3–5). Several genetic susceptibility factors have been identified, including specific major histocompatibility complex (MHC) alleles and rare mutations in cartilage matrix genes (6).

Mice carrying the ank mutation have been studied as a model of arthritis (7–12). The autosomal recessive ank mutation causes an abnormal, flat-footed gait in young mice due to decreased mobility of ankle and toe joints. Loss of joint mobility becomes more severe with age and spreads to most joints throughout the limbs and vertebral column, leading to complete rigidity and death around 6 months of age (7).

Hydroxyapatite crystals develop in articular surfaces and synovial fluid of ank mice, accompanied by joint space narrowing, cartilage erosion, and formation of bony outgrowths or osteophytes that cause fusion (ankylosis) and joint immobility (7,8, 10, 11). Although the distribution and severity of phenotypes do not precisely mimic any single form of human arthritis, many of the pathological features in ankmice are reminiscent of cardinal features in several important arthritic diseases, including ectopic calcification seen in mineral deposition disease, cartilage erosion and osteophyte formation seen in osteoarthritis, and vertebral fusion seen in ankylosing spondylitis (7, 8, 10, 11).

Despite extensive studies over the last 20 years, the genetic defect in ank mice has remained elusive. Here, we identify the ank gene product, study its functional properties, and propose a model that may explain its pathogenic role in arthritis and mineral formation.

Genetic and physical mapping. Previous studies mapped theank locus to proximal mouse chromosome 15 (7). To narrow the location of the gene, we generated 923 F2ank/ank mice from a cross withMus musculus castaneus (13). Linkage analysis identified two microsatellite markers that failed to recombine with theank trait in the high-resolution genetic cross, placing the markers at less than 0.06 centimorgan (cM) from the ankmutation (Fig. 1).

Figure 1

Genetic and physical map of the ank locus. High-resolution genetic mapping localizes the ank mutation to a region between D15Mit251 and D15Mit55. The nonrecombinant marker D15Mit20 was used to initiate a chromosome walk in the ank region. Black bars indicate BssH II sites in overlapping BAC clones. End probes (black circles) from BACs A and H map to opposite sides of ank (number of recombinants shown), and define an interval of ∼400 kb that must contain the mutation. The rescuing BAC C is highlighted in red and further narrows the critical region to ∼190 kb.

The D15Mit20 marker was used as a molecular entry point to build a physical map of the ank region (14). Multiple rounds of chromosome walking led to the isolation of a ∼400-kb interval in bacterial artificial chromosome (BAC) clones (Fig. 1). End probes from BACs A and H recognized three and two recombination events, respectively, that mapped on opposite sides of ank in the high-resolution mapping cross (Fig. 1), showing that the ank mutation must lie in the cloned region.

In vivo complementation and BAC transgenic mice. To narrow the candidate interval, we tested whether individual BAC clones could rescue ank mutant phenotypes in transgenic mice (15). Control ank/ank mutants were unable to grasp a wire cage top at weaning age due to stiffness in the digits. Histological sections of the joints in these mice showed obvious arthritic features like joint space narrowing, cartilage erosion, accumulation of debris in the joint space, and ectopic calcification in and around the affected joints (Fig. 2, D to F). In contrast, ank/ank mice carrying BAC C showed normal digit mobility and normal joint morphology (Fig. 2, G to I). Functional and histological rescue were seen in three independent BAC C transgenic lines evaluated at both 5 weeks and 3 months of age. These data suggest that the ank mutant phenotypes arise from decreased function of a gene contained in BAC C.

Figure 2

BAC C rescues the ank phenotype. (A to C) Hind limb interphalangeal joints from 3-month-old wild-type mice show normal empty joint space flanked by smooth, unmineralized articular cartilage. Joints were stained with alizarin red (A), or sectioned and stained with hematoxylin and eosin (B) or with von Kossa's calcium staining method (C). Scale bars: 250 μm in (A) and 50 μm in (B). (D to F) Age-matched joints of ank/ank mutant mice show joint space narrowing, articular cartilage erosion [asterisks in (E) and (F)], calcified debris on articular cartilage and in joint space [arrowheads in (E) and (F)], and excessive calcification (F). (Gto I) Joints of ank/ank mutant mice carrying BAC C are indistinguishable from wild-type mice.

Identification of the ank mutation. To identify candidate genes in the critical region, we used shotgun sequencing to generate ∼450 kb of random sequence from BAC C (16). BLAST searches identified portions of 11 candidate groups in the nonredundant and expressed sequence tag (EST) databases of GenBank. None of the candidate genes showed any obvious relation to arthritis or displayed any differences in wild-type and mutant samples by Northern or Southern blot analysis. However, further sequencing studies (16) identified a single nucleotide G to T substitution in one of the candidate genes inank mutant mice (Fig. 3B). This base change creates a nonsense mutation in the corresponding open reading frame (ORF) and is predicted to destroy a Hinf I restriction site in genomic DNA. Amplification of the corresponding region from wild-type and mutant mice followed by Hinf I digestion confirmed the sequence change in ank mutant mice (Fig. 3C) (16). This sequence change was not found in 17 wild-type inbred strains surveyed (16), indicating that the G to T substitution is not a common strain polymorphism. Most importantly, the nucleotide change was not present in the JGBF strain in which the ank mutation arose (7), strongly suggesting that the single base substitution is theank mutation.

Figure 3

ank mutation disrupts a highly conserved hydrophobic protein. (A) Mouse and human ANK proteins are nearly identical. Dashed boxes or solid boxes denote regions predicted by the Kyte-Doolittle algorithm (18) to be either likely or certain transmembrane domains, respectively. Black bars, exon boundaries in the gene; red arrowhead, position ofank mutation. Mouse and human sequences have been deposited in GenBank (accession numbers AF274752 and AF274753, respectively). (B) Sequence traces showing a G to T substitution inank mutant mice (small arrowheads in chromatogram), changing the codon for Glu440 to a stop codon. (C) Confirmation of sequence change in genomic DNA. A 58-bp region amplified around the site of the ank mutation is cleaved by Hinf I in wild-type mice, partially cleaved in ank/+heterozygous mice, and not cleaved in ank/ank mice.

Features of the predicted ANK protein. The nucleotide sequence of this gene matches two split GenBank entries for an unpublished gene previously identified in mitogen-stimulated murine prostate carcinoma cells (17). We refer to the gene as ank, and its corresponding protein as ANK, in accordance with the name of the original mouse locus. The ank gene is predicted to encode a 492 amino acid protein (Fig. 3A), with an expected molecular mass of 54.3 kD and an isoelectric point of 8.01. The predicted protein contains three potential N-linked glycosylation sites and multiple putative phosphorylation sites. A hydropathy analysis with the Kyte-Doolittle algorithm (18) revealed 9 to 12 hydrophobic stretches (Fig. 4A). Most are ∼20 residues long, as would be expected for membrane-spanning regions in an integral multipass transmembrane protein. Further analysis with sequence homology and motif scanning algorithms did not identify any significant functional domains in ANK, suggesting that ANK is a previously unknown protein.

Figure 4

Expression and immunolocalization of ANK protein. (A) Kyte-Doolittle hydropathy plot of ANK sequence with red X denoting the site of ank mutation. Blue bars denote position of peptide epitopes used to generate polyclonal antibodies to ANK protein. (B) ANK protein was expressed in COS-7 cells by transfection with pCMV-ANK, and its subcellular localization was determined by confocal microscopy using Ab1 or Ab3 under permeabilizing conditions. Green signal shows strong immunoreactivity to ANK on cell surface; red signal indicates nuclear staining by the DNA-intercalating dye 7-AAD (Molecular Probes). Similar results were seen with Ab2 (not shown). (C) A possible model of ANK protein topology.

Protein expression and subcellular localization. To characterize the ANK protein, we generated an expression construct (pCMV-ANK) by cloning the ank ORF into a mammalian expression vector (pCMV) (19). This construct was transfected into COS-7 monkey kidney cells (20), and protein extracts were collected and analyzed by Western blotting with rabbit polyclonal antibodies raised against various peptide epitopes in ANK (Fig. 4A) (19). These antibodies detected a major protein band of ∼50 kD in pCMV-ANK–transfected but not pCMV-transfected cells (21). To determine the subcellular localization of ANK protein, we used the same antibodies for indirect immunofluorescence on transfected cells (19). Cells transfected with pCMV did not show any appreciable signal (21). In contrast, cells transfected with pCMV-ANK exhibited an intense ring of immunoreactivity on the cell surface, suggesting that ANK is a membrane protein. A subset of cells (30 to 50%) also showed weaker immunoreactivity in the cytoplasm in addition to the cell surface ring (21). Similar results were seen with three different antibodies [Ab1 and Ab3 (Fig. 4B) and Ab2 (21)]. The ANK signal was detectable only in permeabilized cells, suggesting that the epitopes recognized by Ab1, Ab2, and Ab3 are likely to be cytoplasmic. Figure 4C shows one possible transmembrane topology that is consistent with the results from sequence prediction and immunofluorescence experiments (22).

Expression pattern. Northern blot analysis revealed that the ank mRNA is expressed in many tissues in adult mice including heart, brain, liver, spleen, lung, muscle, and kidney (Fig. 5A) (23). Although soft tissue phenotypes have not previously been reported in ank mice, we have observed increased calcification in kidneys of adults, consistent with an important role for the gene in nonskeletal tissues (24). In situ hybridization analysis of developing mouse limbs showed strong ank expression in the developing articular cartilage of joints in the shoulder, elbow, wrist, and digits (Fig. 5, B, D, and F) (23), tissues that are severely affected in adult ank mutants.

Figure 5

The ank gene is expressed in many tissues, including joints. (A) ank cDNA probes detect ∼3.5- and ∼4.0-kb transcripts (arrowheads) in multiple mouse adult tissues with Northern analysis of ∼2-μg mRNA from heart (H), brain (B), liver (L), spleen (S), lung (Lu), skeletal muscle (M), and kidney (K). The two transcript sizes seen are consistent with different usage of polyadenylation sites in the 3′ untranslated region of ank EST clones. (B toG) In situ hybridization analysis of developing mouse forelimbs (embryonic day 16). Antisense probes show ankexpression in articular surfaces of limb joints (arrowheads) between the scapula and humerus (B), the humerus and ulna (D), and the carpal and phalangeal bones (F). No appreciable signal was detected with sense control probes in (C), (E), and (G). Scale bars: 250 μm in (B), 100 μm in (D), and 250 μm in (F).

Cloning of a human ortholog of the ank gene. Database searching with the full-length mouse ank sequence identified no clear homologs in the extensive genome sequences available for bacteria, yeast, worms, and flies. In contrast, the primary sequence of the ank gene is highly conserved in EST clones from many vertebrates, including zebrafish, rats, mice, cows, and humans. Using the exon/intron structure of the mouse gene and the partial sequence from human EST clones, we amplified and sequenced the complete coding region of human ANK (25). The human ANK protein is nearly identical to mouse ANK over its entire length, with only 9 amino acid substitutions out of 492 amino acids (Fig. 3A). Radiation hybrid mapping shows that the human gene maps to a homologous region on human chromosome 5p, closely linked to D5S1954(26). Interestingly, the genetic defects in several human pedigrees with arthritis and chondrocalcinosis map to the same chromosomal region (27–29). Like ankmutants, affected family members show precocious calcification, pain, or arthritis in synovial joints. The sequence identity and map location suggest that the function of the ank gene may be highly conserved in vertebrates.

ANK functions in pyrophosphate regulation. Increased levels of intracellular inorganic pyrophosphate (PPi) have previously been reported in cultured skin fibroblasts and lymphoblasts from one of the human families whose genetic defect maps to the chromosome 5p region containing ANK (30). To determine if ank mutants have a similar defect in PPi metabolism, we established primary skin fibroblast cultures from wild-type and mutant mice (31). Fibroblasts from ank mutants displayed about twofold increase in intracellular PPi levels (31) over wild-type cells (Fig. 6A) (P < 0.01), a change similar to that previously reported for cells from the human patients (30).

Figure 6

ANK activity controls pyrophosphate levels in cultured cells. (A) Primary skin fibroblasts isolated from adult wild-type (wt) or age- and sex-matched ank/ank mutant (mut) mice show striking differences in levels of intracellular and extracellular pyrophosphate (PPi). In this and all other panels, error bars denote standard deviations from quadruplicate pyrophosphate assays conducted on each sample. All results were confirmed in at least three independent experiments. (B) Expression of ANK protein reverses the alterations in PPilevels. Wild-type and ank mutant fibroblasts were transfected with a control vector (+ pCMV) or the same vector driving expression of ANK (+ pCMV-ANK), and PPi levels were measured ∼48 hours later. (C) Overexpression of ANK in COS cells causes dramatic drop in intracellular PPi and rise in extracellular PPi levels. COS-7 cells were transfected with a control vector (+ pCMV) or the ANK expression construct (+ pCMV-ANK). PPi levels were determined ∼48 hours later. (D) The effect of the ANK gene is blocked by the anion transport inhibitor probenecid. COS-7 cells were transfected with pCMV or pCMV-ANK. After ∼30 hours, the media were replaced with control medium or with medium containing 2.5 mM probenecid. Intracellular PPi levels were determined 12 hours later.

PPi and its derivatives are potent inhibitors of calcification both in vitro and in vivo (32). PPi is present in synovial fluid, plasma, and urine at levels sufficient to block calcification and has been postulated to be a natural inhibitor of hydroxyapatite formation in fluids outside cells (33, 32). To test whether the defect inank mice might also affect extracellular PPi, we collected conditioned media from wild-type and mutant fibroblasts and measured their PPi levels (31). Theank mutation caused a three- to fivefold decrease in extracellular PPi levels (Fig. 6A) (P < 0.005), in contrast to its stimulatory effect on intracellular PPi. Thus, the ank gene appears to regulate both intra- and extracellular levels of an important inhibitor of hydroxyapatite crystal formation.

To determine if the cloned ank gene could rescue the PPi abnormalities in the mutant cells, we transfected wild-type and ank mutant fibroblasts with pCMV-ANK and assayed PPi levels (34). Both the increase in intracellular PPi and the decrease in extracellular PPi observed in ank mutant cells were abolished when ANK expression was restored in these cells (Fig. 6B), indicating that ank is the gene responsible for the mutant phenotypes.

To explore how perturbations in ANK activity change cellular PPi levels, we also investigated the effect of high-level ANK expression in COS-7 cells (34). Whereas a loss of ANK function in mutant fibroblasts increased intracellular PPi levels and decreased extracellular PPilevels, a gain of ANK function in COS-7 cells caused a dramatic drop in intracellular PPi to undetectable levels (Fig. 6C) (P < 0.001) and a two- to fourfold rise in PPi levels in the conditioned medium (Fig. 6C) (P < 0.001). In contrast, introduction of theank mutation into the expression construct drastically reduced its effects on PPi levels (intracellular pyrophosphate levels of 100%, 0%, and 48% were seen after transfection with control, wild-type, and mutant constructs, respectively).

The opposite effects of ANK on PPi levels inside and outside cells could be explained if ANK normally functions in PPi transport. To test this model, we examined the effects of probenecid, a weak organic acid that inhibits anion transport (35). Previous studies suggest that articular chondrocytes have a PPi-transport activity that is blocked by 5-mM probenecid (36). When transfected COS-7 cells expressing ANK were treated with 2.5 mM probenecid, the ANK-mediated drop in intracellular PPi was completely blocked (Fig. 6D). This suggests that ANK is functioning through a probenecid-sensitive anion transport mechanism, or that probenecid has additional effects on cells that can counteract ANK function.

Discussion. We have identified a previously unknown, multiple-pass transmembrane protein as the product of theank locus. The original mouse mutation creates a nonsense codon that truncates the COOH-terminal region of the protein and greatly reduces its activity in vitro. The contrasting effects of the ANK gene product on intracellular and extracellular PPilevels are most simply explained if ANK acts as a transmembrane transporter that shuttles PPi between intracellular and extracellular compartments (Fig. 7). Because PPi is charged and polar, any movement of PPithrough the cell membrane is likely to involve a specialized channel or transporter. A loss of this transporter function would result in intracellular accumulation and extracellular drop in PPi, as observed in the ank mutant cells. In contrast, an increase in transporter activity would lead to an increase in extracellular levels and a drop in intracellular levels, as seen in the transfected COS cell assays. The multipass transmembrane structure of the ANK protein and its sensitivity to the anion transport inhibitor probenecid are also consistent with a direct role in PPitransport. It is possible that ANK affects transport or cotransport of other small molecules in addition to PPi, or controls intracellular and extracellular PPi through independent metabolic effects both inside and outside cells.

Figure 7

Working model for the role of ANK in tissue calcification and arthritis. Expression of wild-type ANK leads to elaboration of extracellular PPi, most likely by a transport mechanism. Local elaboration of PPi provides a natural inhibitor of hydroxyapatite deposition, blocking unwanted mineralization in articular cartilage and other tissues. With loss of ANK activity, extracellular PPi levels drop, intracellular PPi levels rise, and unregulated calcification begins in the joints, triggering a destructive cycle (49) of mineral deposition, inflammation, and osteophyte formation leading to arthritis.

How could a defect in PPi regulation lead to the dramatic joint calcification and arthritic destruction seen inank mice? Many previous studies have shown that PPi is a potent inhibitor of calcification, bone mineralization, and bone resorption in vitro and in vivo (32). PPi was purified from urine in a screen for natural agents that inhibit formation of calcium phosphate crystals (32, 33). Addition of PPi to organ culture blocks mineralization of growing bone, and injection in vivo blocks ectopic calcification induced by vitamin D (37,38). Human defects in alkaline phosphatase, an enzyme that degrades PPi, lead to an increase in PPi levels and a severe block in skeletal mineralization in vivo (39). Conversely, genetic defects in a cell surface ectoenzyme that normally generates extracellular PPi from nucleotide triphosphate cause ectopic mineralization of joints and ligaments in the mouse mutant tiptoe walking (40) and may be associated with spinal ligament ossification in humans (41).

The ability of PPi to inhibit mineral deposition has been widely applied in dentistry to control unwanted mineralization. The active ingredient in most tartar control toothpaste formulations is PPi, which inhibits formation of the calcified mineral deposits typically found in tartar (42). The ankgene may provide a natural form of tartar control for the synovial joints of vertebrates. Articular cartilage is one of the few locations that normally remains unmineralized in the vertebrate skeleton. Theank gene is expressed in developing articular surfaces and may help maintain the unmineralized state by providing a local source of PPi to inhibit hydroxyapatite formation. In the absence of normal ank activity, mineralization extends unhindered throughout articular cartilage, hydroxyapatite deposits form in synovial fluid, and new bone is deposited in and around joints, showing that the gene is essential for normal joint maintenance.

If PPi dysregulation is the underlying cause ofank phenotypes, the arthritic defects in ank mice should be rescued when normal PPi levels are restored. This is difficult to test with PPi itself, which is degraded rapidly when administered systemically (32). However, structural analogues of PPi have been developed that are more stable in vivo, including bisphosphonates (32) and phosphocitrate. Like PPi, phosphocitrate is a potent inhibitor of hydroxyapatite crystal deposition (43). Daily injections of phosphocitrate block mineral deposition, osteophyte formation, and joint immobility defects in ank mutants (44), providing additional evidence that ANK controls mineral deposition in vertebrate joints.

Deposition of calcium-containing crystals in joints is common in humans. Up to 5% of the human population shows radiographic evidence of chondrocalcinosis, with the incidence rising to over 40% in the elderly (45, 46). More than 60% of osteoarthritis patients show evidence of calcium pyrophosphate crystals, hydroxyapatite crystals, or both, in synovial fluid (47, 48). A longstanding debate in the rheumatology field has been whether such crystals are a primary cause of arthritis or a secondary consequence of joint damage (45,49–51). Calcium crystals can stimulate release of proteases and inflammatory cytokines from cultured synovial cells, and direct injection of crystals into joints causes arthritis in experimental animals. Joint damage in turn can also stimulate secondary crystal deposition (45, 49–51).Progressive ankylosis and tiptoe walking mice provide clear examples of generalized arthritis syndromes that can be traced to primary defects in the control of mineral formation. These mutants strongly suggest that genetic defects in the mechanisms that control PPi levels can be an important primary cause of osteophyte formation and joint destruction in arthritis (Fig. 7).

Several human pedigrees with joint abnormalities have been mapped to the same chromosomal region as human ANK(27–29). Affected family members display precocious calcification in and around affected joints and clinical symptoms variously described as chondrocalcinosis, pseudogout, pseudorheumatoid arthritis, or pseudoosteoarthritis. Fibroblasts from patients of one of these families show elevation of intracellular PPi (30), an alteration similar to that seen in fibroblasts from ank mice. It will be interesting to test whether familial forms of human arthritis linked to chromosome 5p, as well as more common forms of mineral deposition and joint disease seen in the general human population, can be explained by genetic variation in the human ANK gene. ANK activity may also be a useful target for the development of new therapies for arthritis and many other human diseases involving abnormal mineral deposition.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: kingsley{at}


Stay Connected to Science

Navigate This Article