The mineralisation disorder pseudoxanthoma elasticum (PXE) is associated with mutations in the transporter protein ABCC6. Patients with PXE suffer from calcified lesions in the skin, eyes and vasculature, and PXE is related to a more severe vascular calcification syndrome called generalised arterial calcification of infancy (GACI). Mutations in ABCC6 are linked to reduced levels of circulating vitamin K. Here, we describe a mutation in the zebrafish (Danio rerio) orthologue abcc6a, which results in extensive hypermineralisation of the axial skeleton. Administration of vitamin K to embryos was sufficient to restore normal levels of mineralisation. Vitamin K also reduced ectopic mineralisation in a zebrafish model of GACI, and warfarin exacerbated the mineralisation phenotype in both mutant lines. These data suggest that vitamin K could be a beneficial treatment for human patients with PXE or GACI. Additionally, we found that abcc6a is strongly expressed at the site of mineralisation rather than the liver, as it is in mammals, which has significant implications for our understanding of the function of ABCC6.
Pseudoxanthoma elasticum [PXE; Online Mendelian Inheritance in Man (OMIM) #264800] is a congenital disorder which causes ectopic mineralisation of the eyes, skin and arterial walls, and, in most cases, is associated with mutations in the ABCC6 gene (Bergen et al., 2000; Chassaing et al., 2005; Uitto et al., 2014). The closely related disease generalised arterial calcification of infancy (GACI; OMIM #208000) is characterised by severe vascular mineralisation that is usually lethal in the neonatal period (Cheng et al., 2005). Mutations in either ABCC6 or ENPP1 can cause GACI, and it has been suggested that PXE and GACI represent a spectrum of symptoms from the same underlying causes (Nitschke et al., 2012; Nitschke and Rutsch, 2012). For a recent review of these and similar mineralisation disorders, see Li et al. (2014).
Much of the research into PXE has focused on the role of the protein encoded by ABCC6, a putative efflux transporter of an unknown factor. In mice, Abcc6 is expressed in arterial endothelial cells, cornea, retina, neurons and kidney proximal tubules, but the highest expression is found in the liver (Beck et al., 2003; Matsuzaki et al., 2005; Scheffer et al., 2002). Homozygous Abcc6−/− mice feature spontaneous calcification of the eye, vasculature and skin (Gorgels et al., 2005; Klement et al., 2005). Several experiments using these models have given weight to the ‘metabolic hypothesis’ that PXE is a disorder caused by insufficient levels of a circulating agent, excreted from the liver by ABCC6. Transplanting skin from wild-type mice onto Abcc6−/− animals resulted in mineralisation of the grafted tissue, whereas the reverse operation did not (Jiang et al., 2008). Surgical joining (parabiosis) of Abcc6−/− and wild-type mice halted ectopic mineralisation in the knockout animal (Jiang et al., 2010b), suggesting that the blood of wild-type mice carries an as-yet-unknown anti-mineralisation factor. One candidate for this factor is fetuin-A (alpha-2-HS-glycoprotein/Ahsg), a small liver-derived protein which can inhibit hydroxyapatite precipitation (Schäfer et al., 2003). Serum levels of fetuin-A are reduced in both human PXE patients and Abcc6−/− mice, and recombinant overexpression of fetuin-A successfully restored normal mineralisation in knockout mice (Jiang et al., 2010a, 2007).
Another candidate for this unknown factor is vitamin K (Borst et al., 2008). Patients with PXE have reduced levels of serum vitamin K (Vanakker et al., 2010), a cofactor required by the enzyme gamma-glutamyl carboxylase (GGCX) to convert glutamic acid into gamma-carboxyglutamate (Gla) in certain proteins, conferring a high affinity for calcium. Three Gla proteins are directly implicated in bone or soft tissue mineralisation: osteocalcin (OC; bone gamma carboxyglutamate protein/Bglap – Mouse Genome Informatics), matrix Gla protein (MGP) and Gla-rich protein (GRP) (Theuwissen et al., 2012; Viegas et al., 2008). Functional loss of GGCX results in symptoms very similar to PXE (Vanakker et al., 2006), and in mice, the pathological mineralisation phenotype of the Abcc6−/− genotype is accelerated by the concomitant knockout of Ggcx (Li and Uitto, 2010). Arterial calcification has been reported in patients using vitamin K antagonists such as warfarin, probably due to under-carboxylation of OC, GRP and especially MGP (Chatrou et al., 2012; Theuwissen et al., 2012). Similar results were obtained in Abcc6−/− mice (Li et al., 2013). However, ectopic mineralisation in Abcc6−/− mice was not reduced by dietary administration of vitamin K1 or K2 (Brampton et al., 2011; Gorgels et al., 2011; Jiang et al., 2011) despite successfully raising the vitamin K concentration in tissues and serum; interestingly, this increase was significantly subdued in knockout mice and was accompanied by hepatic lesions, suggesting that Abcc6−/− mice have an impaired ability to absorb, metabolise or distribute vitamin K (Brampton et al., 2011).
Here, we describe a zebrafish mutant, gräte (grt; abcc6ahu4958), identified in a forward genetic screen, with a mutation in the abcc6a gene. gräte fish show signs of excessive mineralisation in the craniofacial and axial skeleton but appear otherwise normal. A transgenic reporter revealed unexpected abcc6a expression at craniofacial bone elements and in the notochord, but not in the liver. Significantly, administration of vitamin K counteracted the hypermineralisation phenotype of abcc6a−/− and enpp1−/− embryos, whereas administration of warfarin exacerbated the phenotype in both lines.
RESULTS AND DISCUSSION
Characterisation of the gräte phenotype as a model for PXE
Zebrafish homozygous for the gräte allele featured hypermineralisation of the axial skeleton, resulting in mineralised structures appearing in the intervertebral space (Fig. 1A). Adult mutants survived for at least one year, but were shorter than siblings (Fig. 1B). Craniofacial bone elements within embryos appeared to be more mineralised (Fig. 1C). Elsewhere, ectopic mineralisation is infrequently present in skin on the ventral side; interestingly, this is equally common in grt+/− and grt−/− embryos (Fig. 1D). Quantifying Alizarin Red staining (Fig. 1E) revealed that mineralisation was significantly more advanced in grt−/− embryos as early as 6 dpf, and that heterozygous embryos had an intermediate phenotype (Fig. 1F,G). Compared with siblings, juveniles at 6 weeks featured an undulating spine and vertebrae were shorter (mutant, 162±2.1 µm; sibling, 174±1.8 µm; s.e.m., P<0.001) and thicker (mutant, 106±6.0 µm; sibling, 85±3.8 µm; s.e.m., P=0.01) than those in siblings, with large mineralised nodules developing on the margins of the intervertebral space (Fig. 1H-K).
Using whole-genome sequencing (Mackay and Schulte-Merker, 2014), a T→G mutation was detected at chr6:10974761 in the abcc6a gene. The gräte phenotype co-segregated with markers flanking the abcc6a locus (Fig. 2A). This gene encodes a putative efflux transporter of the ATP-binding cassette (ABC) superfamily, with two transmembrane domains and two catalytic nucleotide-binding domains (NBD) (Fig. 2B). The identified mutation results in the substitution L1429R in a highly conserved region of NBD-2 containing the Walker B motif (Fig. 2C). This motif of four hydrophobic residues is essential for binding to ATP (Geourjon et al., 2001; Walker et al., 1982). Most of the known human PXE-causing mutations are in the second ABC or NBD domains (Le Saux et al., 2001) (Fig. 2B), including an I1424T substitution immediately preceding the human equivalent of zebrafish L1429 (L1425). The detectable phenotype in heterozygous embryos (not reported in human patients or the mouse model) suggests L1429R to be highly deleterious. Importantly, the phenotype was considerably variable from clutch to clutch, indicating that external factors can influence the extent of ectopic mineralisation. This is congruent with the variable human symptoms of PXE even among families with the same genetic lesion in ABCC6 (Uitto et al., 2014).
In contrast to the phenotype described above, morpholino knockdown of abcc6a has previously been reported to cause oedemas and high mortality in embryonic zebrafish (Li et al., 2010), even though expression was reduced by only 54-81%. The discrepancy between morpholino and grt−/− phenotypes might be attributed to off-target effects of the morpholinos, even though co-injection of morpholino with wild-type mouse Abcc6 mRNA caused a complete rescue of the phenotype (Li et al., 2010).
Superficially, the gräte phenotype does not match that of human patients with PXE, who experience mineralisation of the skin and angioid streaks in the retina (Finger et al., 2009), with no reported vertebral abnormalities. Knockout mice mirror the human symptoms (Klement et al., 2005). A possible cause of this discrepancy is the greater propensity for mineralisation to occur in the fibrillar collagen of the notochord sheath. Similarly, the enpp1−/− zebrafish (dragonfish) features extensive hypermineralisation of the axial skeleton, unlike human patients with GACI (Apschner et al., 2014).
abcc6a is expressed in osteoblasts and not the liver
In situ hybridisation (ISH) at 5 dpf showed abcc6a expression in regions of developing bone such as the lateral-ventral edge of the operculum (Fig. 3A), but not in the liver (unlike mammalian ABCC6). Two orthologues of ABCC6 exist in the zebrafish genome, but whole-mount ISH revealed abcc6b expression in the operculum and parasphenoid as well as the cartilage of the ear, with no hepatic expression (Fig. 3B). Fetuin-A (ahsg) was highly expressed in the liver (Fig. 3C). Expression of either abcc6a or ahsg was not altered by the gräte allele (data not shown) and expression patterns of the vitamin K-dependent genes ggcx and vkor were not associated with bone elements (supplementary material Fig. S1).
To facilitate further analysis of abcc6a expression, a reporter construct was prepared by introducing the GAL4 element into the start codon of the abcc6a gene via BAC recombineering (Bussmann and Schulte-Merker, 2011). The mosaic expression of Tg(abcc6a:gal4;uas:gfp) at 7 dpf revealed GFP in the notochord and in cells near the operculum and cleithrum (Fig. 3D; supplementary material Fig. S3). Crossing the stable GAL4 line with a line expressing UAS:RFP and the early osteoblast marker osterix:GFP (Spoorendonk et al., 2008) confirmed that abcc6a was co-expressed with osterix (osx; Sp7 transcription factor/sp7 – Zebrafish Model Organism Database) in some cells in the operculum and cleithrum (Fig. 3E,F); expression in these bone elements first appeared at 4 dpf, about one day after that of osx. By contrast, expression of abcc6a appeared very early (24 hpf) in the notochord and neural tube (Fig. 3E). In older fish (20 dpf), osx+ osteoblasts were present in the developing neural and hemal arches of the vertebrae, whereas abcc6a was strikingly expressed in the intervertebral disc regions (Fig. 3G,I), structures that are affected most by the gräte allele (Fig. 1K). The late osteoblast marker osteocalcin:GFP is co-expressed with abcc6a in some, but not all, cells around the operculum (Fig. 3H). Based on these observations, we believe that abcc6a labels a population of mature osteoblasts. ABCC6 expression in mammalian osteoblasts or in other cells at the site of mineralisation has not been reported in the literature.
It has been hypothesised that Abcc6 has an endocrine role, exporting a ligand from the liver into the circulation. Our results show that zebrafish Abcc6a functions locally at the site of mineralisation, suggesting that the transported ligand in fish is not liver derived. Jansen et al. recently reported that ABCC6 overexpression induces nucleotide release in vitro (Jansen et al., 2013). These nucleotides are rapidly converted by ENPP1 into pyrophosphate (PPi), a potent inhibitor of mineralisation (Jansen et al., 2013; Nitschke et al., 2012). In a follow-up study, Jansen et al. reported that PPi secretion from the livers of Abcc6−/− mice was dramatically lower than that of wild-type mice, and they suggested that ABCC6 is an ATP efflux transporter (Jansen et al., 2014). In line with this compelling hypothesis, we postulate that zebrafish Abcc6a secretes ATP from cells at the site of mineralisation, increasing PPi locally, in contrast with the hepatically derived PPi in mammals.
We have recently described a zebrafish allele, dragonfish (dgf), with a nonsense mutation in enpp1, resulting in ectopic mineralisation of the skin and axial skeleton in embryos, and the eyes and bulbus arteriosus of the heart in adults similar to GACI symptoms (Apschner et al., 2014). The axial phenotype of dgf mutants is considerably more severe than grt mutants, but dgf+/− embryos are phenotypically normal. We did not observe an additive effect from these two alleles: grt+/−; dgf+/− embryos were indistinguishable from grt+/−; dgf+/+, and the axial hypermineralisation phenotype of dgf mutants was not exacerbated by the grt genotype (data not shown). Furthermore, transgenic overexpression of ENPP1 reduced mineralisation in all embryos, regardless of grt genotype (not shown). Both of these results suggest that zebrafish Abcc6a is one of several sources of nucleotides for Enpp1.
Vitamin K reduces hypermineralisation
Patients with PXE are reported to have low serum concentrations of vitamin K (Vanakker et al., 2010), but vitamin K was not effective in treating PXE in mouse models. We tested the effect of vitamin K on gräte embryos by supplementing the media from 4-8 dpf with 80 µM phylloquinone (vitamin K1). Vitamin K1 reduced hypermineralisation in grt−/− and grt+/− embryos, resulting in significant rescue of the gräte phenotype (Fig. 4A,B).
Warfarin is a potent antagonist of vitamin K, reducing serum levels by inhibiting its recycling. As warfarin has been reported to accelerate the mineralisation phenotype of Abcc6−/− mice (Li et al., 2013), we raised grt−/− zebrafish embryos in the presence of sodium warfarin from 4-8 dpf. Mortality was observed at concentrations exceeding 120 µM. At 60 µM, warfarin stimulated an approximately twofold increase in mineralisation in embryos of all genotypes, resulting in dramatic hypermineralisation of grt−/− embryos (Fig. 4A,C). Warfarin also stimulated ectopic mineralisation in the ventral skin in ∼20% of treated embryos regardless of genotype (supplementary material Fig. S2).
To test whether the observed activity of vitamin K is specific to that of abcc6a, we administered vitamin K1 (80 µM) to dragonfish (enpp1−/−) embryos, which feature extensive axial hypermineralisation due to an inability to produce PPi (Apschner et al., 2014; Huitema et al., 2012). In these embryos, vitamin K1 administration from 4-8 dpf provided the same protective effect as seen in gräte, and warfarin similarly exacerbated the phenotype (Fig. 4D-F). Dragonfish embryos exhibit ectopic mineralisation in the skin much more frequently than gräte embryos, enabling the effect of vitamin K to be tested on this phenotype. Curiously, vitamin K did not reduce the incidence or extent of skin mineralisation, which affected about half of the dgf−/− embryos regardless of treatment (Fig 4G-I).
These results show for the first time that vitamin K supplementation reduces hypermineralisation caused by mutations in either of the genes implicated in PXE and GACI. We propose that the reduced serum level of vitamin K seen in PXE patients is a consequence of its utilisation by GGCX in an attempt to restrict mineralisation, and not a consequence of the loss of ABCC6; subsequent administration of vitamin K could increase MGP carboxylation by GGCX, thus preventing hypermineralisation. This could also explain why Abcc6−/− mice were more resistant to increases in serum levels caused by a vitamin K-rich diet (Brampton et al., 2011). One prediction of this model is that serum vitamin K would also be depleted in GACI patients (ENPP1−/−).
The potential of vitamin K as a treatment for PXE or GACI is of enormous clinical significance, but the positive results shown here contrast with negative results seen in mice from three separate research groups. We believe this contrast unlikely to be simply a matter of bioavailability. It is possibly related to the difference in affected tissues; vitamin K-dependent anti-mineralisation factors such as MGP might have a bigger impact in the axial skeleton than in the skin. In any case, the near-identical response of abcc6a and enpp1 mutants to vitamin K and warfarin treatment renders strong support for the notion that the human PXE and GACI syndromes are closely related clinical entities. Finally, the expression of abcc6a in non-hepatic organs, which we report here, sheds new light on the cellular source of the ABCC6 ligand.
MATERIALS AND METHODS
Zebrafish were maintained in standard husbandry conditions (Nüsslein-Volhard and Dahm, 2002) in accordance with Dutch regulations for ethical treatment of experimental animals. Embryos were raised in E3 media (5 mM NaCl, 17 µM KCl, 330 µM CaCl2, 330 µM MgSO4) at 28°C.
Identification of the gräte mutation
A forward-genetic screen was performed using Alizarin Red staining as described previously (Spoorendonk et al., 2010). The mutation in abcc6a was detected using whole-genome sequencing, as reported elsewhere (Mackay and Schulte-Merker, 2014). Subsequent genotyping was performed using the KASP SNP detection assay mix (LGC Genomics) with the following primers:
wild type: 5′-GAAGGTGACCAAGTTCATGCTAGACAAAAGTTCTGGTGCT-3′; mutant: 5′-GAAGGTCGGAGTCAACGGATTGACAAAAGTTCTGGTGCG-3′; common reverse: 5′-GTCCAGTGCAGCTGTTGCCTCAT-3′.
Whole-mount ISH and BAC transgenesis
To generate the abcc6a:gal4 transgene, the GAL4FF transcriptional activator was recombined in place of the ATG site of abcc6a in bacterial artificial chromosome (BAC) #DKEY-252I22, as described (Bussmann and Schulte-Merker, 2011). The following primers were used: Abcc6a_gal4_f: 5′-GAAGCAGGATACACAGCAGGGATAGAGACAGCCTCAGGACCAG-ACGAGTGACCATGAAGCTACTGTCTTCTATCGAAC-3′; Abcc6_neo_r: 5′-GAAATGTTTGCTCACCCATAGAGGGTCAAGTCCACTTAGACT-GCAAAAGGTCAGAAGAACTCGTCAAGAAGGCG-3′.
Vitamin K/warfarin treatment and mineralisation assay
Vitamin K1 or sodium warfarin (Sigma-Aldrich) were dissolved in 1:1 DMSO:ethanol or water to a working stock concentration of 40 or 60 mM, respectively. Embryos (4 dpf) were incubated with compounds in E3 media in the dark until 8 dpf. Embryos were fixed, bleached with H2O2, stained with Alizarin Red (0.005%) in 1% KOH and 0.2% Triton X-100 (Sigma-Aldrich), and imaged using an Olympus SZX16 stereomicroscope. To quantify Alizarin Red staining, images were processed using ImageJ (NIH). Mineralisation in each genotype-treatment group was compared by a two-tailed t-test.
S.S.-M. acknowledges support from the Smart Mix Programme of the Netherlands Ministry of Economic Affairs. L. Lleras provided valuable advice.
E.W.M. performed experiments and analysed data. A.A. constructed the abcc6a BAC transgene. E.W.M. and S.S.-M. conceived experiments and wrote the manuscript.
This work was supported by the Deutsche Forschungsgemeinschaft (DFG), Cells-in-Motion Cluster of Excellence [EXC 1003 – CiM]. A.A. received a DOC Fellowship (Austrian Academy of Sciences).
The authors declare no competing or financial interests.