New tools for studying osteoarthritis genetics in zebrafish

Summary Objective Increasing evidence points to a strong genetic component to osteoarthritis (OA) and that certain changes that occur in osteoarthritic cartilage recapitulate the developmental process of endochondral ossification. As zebrafish are a well validated model for genetic studies and developmental biology, our objective was to establish the spatiotemporal expression pattern of a number of OA susceptibility genes in the larval zebrafish providing a platform for functional studies into the role of these genes in OA. Design We identified the zebrafish homologues for Mcf2l, Gdf5, PthrP/Pthlh, Col9a2, and Col10a1 from the Ensembl genome browser. Labelled probes were generated for these genes and in situ hybridisations were performed on wild type zebrafish larvae. In addition, we generated transgenic reporter lines by modification of bacterial artificial chromosomes (BACs) containing full length promoters for col2a1 and col10a1. Results For the first time, we show the spatiotemporal expression pattern of Mcf2l. Furthermore, we show that all six putative OA genes are dynamically expressed during zebrafish larval development, and that all are expressed in the developing skeletal system. Furthermore, we demonstrate that the transgenic reporters we have generated for col2a1 and col10a1 can be used to visualise chondrocyte hypertrophy in vivo. Conclusion In this study we describe the expression pattern of six OA susceptibility genes in zebrafish larvae and the generation of two new transgenic lines marking chondrocytes at different stages of maturation. Moreover, the tools used demonstrate the utility of the zebrafish model for functional studies on genes identified as playing a role in OA.


Introduction
Osteoarthritis (OA) is an increasingly common degenerative joint condition, estimated to affect more than 100 million people worldwide and more than 40% of people over 70 years of age 20 . OA has a complex, multifactorial aetiology and treatment options remain limited; however there is increasing evidence of a genetic component to OA (reviewed by Cornelis et al 2 ). It has been estimated from a number of twin studies that the genetic contribution to OA is between 39% and 60% in hip and knee OA, respectively 3,4 . To date, however, while many genes have been shown to be differentially expressed between osteoarthritic and healthy chondrocytes by RT-PCR and microarray analyses 5,6,7 relatively few genes have been identified through association studies to have reached genome wide significance; those that have include Growth/Differentiation factor (GDF5), a cluster of six genes on 7q22 (comprising PRKAR2B, HPB1, COG5, GPR22 DUS4L and BCAP29) and recently MCF2L 8,9,10 .
There is increasing evidence showing that there are significant similarities between OA progression and the normal developmental process of endochondral ossification, whereby a cartilage template is progressively replaced by bone (reviewed by Pitsillides and Beier 11 ). Endochondral ossification occurs during embryonic development and continues postnatally in the cartilage growth plates of long bones, which give the potential for continued skeletal growth. In the cartilage growth plate, chondrocytes are organised into zones of progressive maturation beginning with the metabolically inactive resting chondrocytes, which become activated and proliferate; following proliferation they enter a pre-hypertrophic and finally a hypertrophic state (reviewed by Mackie et al 12 ). Hypertrophic chondrocytes, which are regarded as terminally differentiated, remodel their matrix through expression of proteases such as Matrix metalloprotein 13 (MMP13) and secretion of different collagens, in particular Type X Collagen 13 .
Ectopic chondrocyte hypertrophy, marked by expression of Type X Collagen, is seen as a hallmark of OA in humans and in diverse animal models from mice to sheep (reviewed by van der Kraan and van den Berg 14 ). Genetic changes that stimulate chondrocyte hypertrophy lead to increased incidence of OA 14 raising the prospect that inhibition of chondrocyte entry to hypertrophy might be a therapeutic target in OA. Moreover, these parallels between endochondral ossification and OA raise the prospect that developmental models could be used to study the processes that lead to chondrocyte hypertrophy. Therefore, tools that allow investigators to follow the differentiation state of chondrocytes and the onset of hypertrophy in real time in vivo will be particularly valuable.
Zebrafish, along with another small teleost species medaka, have long been used as model organisms for developmental biology. They owe their popularity in part to the rapid external development of their larvae, to their amenability to genetic manipulation and also, importantly, to the translucency of the larvae, which allows detailed observation of organogenesis in the living fish. The zebrafish craniofacial skeleton is of comparable complexity to that of terrestrial vertebrates, and contains bones of both dermal and chondral origins, which form from neural crestderived cells relatively early in development 15 . Importantly both the key regulators of skeletal development and the control of the major signalling pathways are highly conserved between mammals and teleosts 16 . As such, findings in fish are likely to be applicable to mammalian osteogenesis. Therefore, to increase the utility of the zebrafish model for functional in vivo studies into OA, we generated a new transgenic col10a1 reporter line to enable monitoring of chondrocyte hypertrophy in live fish in real time.
We describe, for the first time in any animal model, the spatiotemporal expression of the OA associated gene mcf2l during early development. Additionally we describe zebrafish expression of parathyroid hormoneerelated protein (pthrp) and the mRNA expression of three key collagens and gdf5 are shown. These tools provide a platform from which to probe the function of these genes during cartilage development in vivo.

Zebrafish husbandry
Zebrafish were maintained as described 17 . All experiments on zebrafish were approved by the local ethics committee and the Home Office (Project licence number 30/2863).

Skeletal staining
The protocol for bone and cartilage staining was as described previously 18 .

In situ hybridization
As previously described 19 In situ hybridisation probes previously used were gdf5 20 , col10a1 21 .

Generation of new probes
Total RNA was extracted from zebrafish at 2 days post fertilization (dpf), 5 dpf and adult fin using EZNA total RNA kit following the manufacturer's protocol. 1 mg RNA was taken and used as a template to synthesize cDNA using promega MMLV-RT following the manufacturer's protocol. The cDNAs were pooled and 1 mg of the pooled cDNA was used as the template for a 50 ml PCR reaction using Roche Fidelity Taq (with conditions as per manufacturer's protocol) and a simple 35 cycle extension programme. Primers sequences used for probe synthesis are shown in Table I. The PCR product was cleaned with a PCR purification kit (E.Z.N.A) and 8 ml was used for RNA Dig labelled probe transcription as previously described 22 . To generate mcf2l, col9a2, and col1a2 probes we amplified T3 polymerase tailed fragments from total cDNA and used the cleaned PCR product as a template for probe synthesis. The pthrp PCR product was cloned into pGEMT, digested with NdeI, and transcribed with primer T7 for antisense probe synthesis. Probes were synthesized as previous described 22 .

Transgenic lines
The Tg(Col2a1aBAC:mcherry) stable line was generated by injection of the modified BAC construct previously described 18,23 along with tol2 mRNA. Injected larvae were grown to sexual maturity and the F2 generation was screened for evidence of germline integration. The osterix reporter line Tg(OlSp7:mCherry) zf131 has been previously described 24 .
The primers for amplification of the tol2 sites and for confirmation of integration have been previously published 25 . The stable col10a1:citrine reporter line was generated by injection of the modified BAC construct containing citrine under the control of the col10a1 promoter. Larvae showing strong mosaic transgene expression were grown to maturity and their offspring screened to identify germline carriers.  200) overnight at 4 C and then incubated with fluorescent conjugated secondaries (Dylight 488, and 550 1/500) for 3 h at RT, as previously described 18 . Larvae were viewed on a Leica SP5 confocal.

Cryosectionning
For cryosectionning larvae were fixed, incubated in 20% sucrose overnight, snap frozen in OCT and 20 mM sections cut on a Leica cryostat at À20 C. Sections were rehydrated in PBS and viewed on a compound microscope.

Phylogenetic analyses
Phylogenetic analyses were performed using the 'Phylogenetic Tree' web resource available from the Computation Biochemistry Research group at the Swiss Federal Institute of Technology (http:// www.cbrg.ethz.ch/services/PhylogeneticTree) 26 . Sequences were input and analysis was run as in distance mode with the results displayed as an unrooted tree in the case of the mcf2l analysis and as a rooted tree for the pthlp analyses.

Mcf2l expression
The MCF2L locus has recently been identified as having a genome wide association with human hip/knee 8 . We identified two possible zebrafish homologues of Mcf2l (mcf2l) on LG1 and LG9 using BLAST searches. To establish which is the closest orthologue of mammalian Mcf2l we performed synteny analysis using Ensembl and generated a phylogenetic tree [ Fig. 1

Pthlh/Pthrp expression
Expression of PTHrP/PTHLH has been shown to be increased at the protein level in cartilage from patients with OA 28 , although the level of the Pthrp at the transcript level has been shown to be decreased in surgically induced OA in the rat 29 . We identified two possible zebrafish homologues of PTHLH; to establish which is the closest homologue we generated a phylogenetic tree [ Fig. 2(A)] and established that the homologue which we denote as pthlha is more closely related to other vertebrate Pthlh genes than the gene which we denote pthlhb. In zebrafish pthlha was expressed at 72 hpf in the pectoral fins and in the cartilages that form the ventral jaw [ Fig. 2(B)]. At 120 hpf, pthlha expression could be seen in the operculum [ Fig. 2(C, C 0 )], and by this stage jaw cartilage expression of pthlh was mainly restricted to the 5th branchial arch (a cartilaginous structure which is subsequently mineralised) at the position where the teeth will later form. Additionally, strong expression was seen in the mesonephros pronephros [ Fig. 2(C)].

Gdf5 expression
GDF5 variants have been demonstrated to show association with human OA 10,30 . In mouse, Gdf5 becomes restricted to joints and is required for correct joint specification 31 . In zebrafish, gdf5 expression has been previously described in pharyngeal arch cartilages up to 96 hpf 20 . Gdf5 is also expressed in the developing fin joints through larval and early adult life 32 . At 96 and 120 hpf we saw expression restricted to the jaw joints that form between the Meckel's cartilage and palatoquadrate and the tip of the ceratohyal [ Fig. 2(EeE 00 )], suggesting that it has a conserved function in the specification of joints in zebrafish.

Collagen expression
We here describe the expression of three collagen types involved in skeletal development (col2a1, col9a1, and col10a1), and we have generated and analysed two transgenic reporter lines (col10a1 and col2a1) in vivo. These specific collagens were chosen on the basis that all are expressed in chondrocytes at specific stages of maturation during endochondral ossification 12 .
Mutations in COL9A2 have been linked by numerous groups to autosomally dominant multiple epiphyseal dysplasia (MED), a disease which manifests from childhood with axial limb deformities, joint pain and gait abnormalities and predisposes sufferers to early onset OA 33 . In zebrafish, col9a2 was expressed strongly in the otic capsule of the ear at 72 hpf, and was more weakly detected in the jaw cartilage elements [ Fig. 2(GeG 0 )]. At 120 hpf, col9a2 transcripts were present throughout cartilagenous structures of the ventral jaw and remained strongly expressed in the otic capsule [ Fig. 2(FeF', HeH 0 )].
Col10a1 expression is a hallmark of chondrocyte hypertrophy in mammals 34 . In zebrafish, at 72 hpf, col10a1 was expressed in the cleithrum, operculum, and parasphenoid, bone elements that form through intramembranous ossification [ Fig. 2(IeI 0 )]. At 120 hpf, col10a1 was still strongly expressed in dermal bone elements, but could, at this stage of development, also be detected in chondral bone elements such as the centre of the ceratohyal [ Fig. 2 24 ], thus confirming as previously reported that col10a1 is expressed in osteoblasts as well as in chondrocytes during zebrafish 21 , and gar 35 development. At later stages of development, (shown here at 14 dpf), col10a1 was expressed throughout the vertebral column [ Fig. 3(B)], including the neural and hemal arches [ Fig. 3(B)]. Col2a1:mCherry was expressed in all chondrocytes during early skeletal development [ Fig. 3(CeE)], while at later stages col10a1 becomes expressed in a subset of chondrocytes as they maturate to hypertrophy [ Fig. 3(DeE) and data not shown]. As shown in Fig. 3(D 00 and E 00 ), in a double transgenic line Tg(col10a1BAC:mcitrine) hu7050 ; Tg(col2a1aBAC:mCherry) hu5900 it was easy to distinguish the less

Transgenic reporter fish faithfully recapitulate the endogenous expression of col2a1a and col10a1
We generated a stable transgenic line marking chondrocytes by injection of the col2a1a:BAC mCherry construct previously described 18 . In this line (also described in 23 mCherry labels all chondrocytes at early stages of development. To demonstrate that mCherry expression marks cells which express col2a1a we used immunohistochemistry to detect the mCherry protein and to detect the Collagen II protein. As expected the cells which express col2a1a:mCherry are surrounded by matrix containing the Collagen II protein [ Fig. 3(F, F 0 )]. To demonstrate that the col10a1 transgenic line recapitulates the expression of endogenous col10a1 mRNA we performed an in situ for col10a1, in 5 dpf larvae, which we detected with an anti-digoxigenin-rhodamine secondary antibody and immune labelled for GFP/citrine which we detected with 488conjugated secondary antibody [ Fig. 3(G)]. The overlay of the two antibodies demonstrates that all regions in which col10a1 mRNA is present show co-expression of citrine [ Fig. 3(GeG 00 )].

Discussion
Osteoarthritis associated genes are dynamically expressed during zebrafish development In this study, we describe the spatiotemporal expression of the OA associated gene mcf2l during early development. Mcf2l (previously identified as Ost and Dbs) is a guanine nucleotide exchange factor, which in purified form, catalyses nucleotide exchange on RhoA and Cdc42 36 . In 5-week-old rat brain sections, Ost/Mcf2l expression was seen in neurons and a-tanycytes 36 . However, to date, the developmental expression of Mcf2l is unknown. Functionally, Mcf2l has been shown to stimulate migration of breast carcinoma cells and of Schwann cells 37,38 . Furthermore, Mcf2l has been identified, through genome-wide association (GWA) studies, to have a significant association with OA. However, with only limited expression and functional data, it is currently unclear how the genetic variants contribute to OA, although the fact that it plays a role in Schwann cell migration has lead to predictions that it could be involved in the pain response to OA 38,39 .
We demonstrate that mcf2l is dynamically expressed in a range of cell types during development, including Kupffer's vesicle. We also observe diffuse expression of mcf2l in the brain throughout development, consistent with the strong expression seen in the brain in rat 36 . Importantly, we have also observed expression of mcf2l in the developing jaw cartilages, which suggests that mcf2l has a function in cartilage development, offering another potential explanation for how MCF2L could play a role in OA.
Additionally, we have characterised the expression pattern of pthrp in zebrafish. PTHrP has been shown to regulate the entry of chondrocytes into hypertrophy, and application of PTHrP can block the ability of RUNX2 to induce expression of hypertrophic markers such as COL10A1 in culture 40 . Here, we show that pthrp is dynamically expressed during zebrafish development. Pthrp is only transiently expressed in cartilage elements, becoming restricted to the 5th branchial arch, adjacent to where the first tooth will attach to the arch at 120 hpf 41 . PthrP is required for tooth eruption in mice 42 , suggesting a likely conservation of function between teleosts and tetrapods. We also observed strong expression of pthrp in the developing mesonephros. No significant overlap in expression is seen with the zebrafish parathyroid hormones pth1 and pth2, whose expression is limited to the lateral line and sense organs 43 . Gdf5, in zebrafish, becomes restricted to the site of joints in the developing cartilage as is the case in mice 31 . Additionally, a number of other genes implicated in human OA such as FRZB/sfrp3 44 and ASPN 45 show expression patterns that hint at conserved functions in zebrafish 46 (data available on zfin.org).

Collagens
Various collagens have been implicated in OA susceptibility and pathogenesis; mutations in col1a1 are associated with osteogenesis imperfecta both in the zebrafish and in humans 47,48,49 . Only weak association has been seen between COL1A1 and OA through GWAS 56 , but other studies report increased levels of COL1A1 in osteoblasts from OA 51 , suggesting a shift towards an osteophytic phenotype. Type II Collagen breakdown fragments are frequently used as a biomarker for OA, while synthesis of the pro collagen gene COL2A1 has been shown to be increased in many models of OA (see for example review by Garvican et al 52 ). COL9A2 has been linked through candidate gene association studies to OA of the hip 53 . Ectopic chondrocyte hypertrophy is seen as a hallmark of OA 14 , and Col10a1 expression is the best characterised marker of hypertrophic chondrocytes 54 . We show here that all four collagen genes show dynamic expression in the developing skeleton of the zebrafish and that in zebrafish, as in mammals, col10a1 can be used as a marker of hypertrophic chondrocytes in addition to marking zebrafish osteoblasts.

Zebrafish as a model for OA
There are many existing animal models established for the study of OA ranging from small rodents to large mammals such as sheep 1 . One might think, therefore, that there is little point in adding another model to the list. Zebrafish, although their skeleton is subject to different loading due to their aqueous environment, do have a number of key advantages for studies of OA genetics that may complement those in existing animal models. One example is that, due to their translucency during larval development, organogenesis, even deep tissues such as the skeleton can be viewed microscopically in vivo. In the case of transgenic reporter lines, such as the ones described in this manuscript, detailed observations on the location and behaviour of the cells expressing the genes in live fish can be made, something which is not possible in the existing animal models, for which the depth and mineralisation of the joints limits the options for imaging. Already a number of zebrafish cartilage skeletal mutants have been demonstrated to share pathology with human disease; these include the heparin sulphate proteoglycan mutants such as ext2/dak and papst/pic which model the human osteosarcoma condition Multiple Hereditary Exostoses (HME) 50,55,56,57 . The ongoing generation of an increasing number of transgenic lines marking bone via promoters such as osterix 24,58,59 and osteocalcin 60 or cartilage by col2a1 28 pathways (reviewed in Hammond and Moro 23 ) will further benefit fish skeletal research.
It has previously been shown that all five of the genes in the OA susceptibility locus on 7q22 are expressed in zebrafish, and two of these, cog5 and dus4l show expression in cartilage at 5 dpf 9 . Therefore, taken together with the published expression of frzb/sfrp3 61 , asporin (direct data submission to zfin.org), and the expression patterns detailed in this manuscript, candidate OA Fig. 3. Col10a1, osterix and col2a1 transgenic reporter expression expression of tg(col10a1BAC:citrine) A, A 0 B, C, E 0 , E 00 , F 0 and F 00 , tg(col2a1BAC:mcherry) C, D, D 00 , E 0 and E 00 and Tg(OlSp7:mCherry)zf131 A 0 (A) Col10a1BAC:citrine reporter expression in live zebrafish at 72 hpf, (A,A 0 ), 120 hpf (DeE 00 ) and 14 dpf (B). B is of the vertebral column at the level of the cloaca. (F and F 0 ) 96 hpf col2a1BAC:mCherry transgenic larvae fixed and labelled with anti-mCherry to detect transgene activity in green and with IIeII3B3 antibody (DSHB) which detects type II collagen protein in red, green labelled cells with transgene activity are surrounded by matrix positive for collagen 2 protein. G) Fluorescent in situ for col10a1 mRNA detected with anti-digoxigenin rhodamine. G 0 ) anti-GFP/citrine staining and G 00 shows the overlay of G and G 0 , showing that all regions that stain for col10a1 mRNA are also positive for the fluorescent transgenic protein. A, A 0 , DeD 00 F and GeG 00 are lateral views and EeE 00 and F 0 are ventral. All are projections of confocal stacks, all presented with anterior facing left. cl ¼ cleithrum, m ¼ maxilla, ch ¼ ceratohyal, op ¼ operculum, mc ¼ Meckel's cartilage, 5ba ¼ 5th branchial arch and teeth, hs ¼ hyosymplectic, br ¼ branchiosteal rays, na ¼ neural arches, ha ¼ haemal arches. genes identified through GWAS have homologues in zebrafish, and most have been shown to be expressed in the skeletal tissues at various stages of development. The expression of these 'OA' genes in zebrafish can be seen in the developing skeletal system and it is likely that they will have conserved functions during skeletogenesis with their mammalian homologues. An advantage of studying the function of OA genes in zebrafish is the relative ease of genetic manipulation in the zebrafish, such as morpholinos for transient knockdown of protein translation 67 and transgenic approaches for overexpression of genes of interest, either globally or under the control of a promoter of interest. There have also been recent advances in the ability to rapidly generate stable mutant lines 63,64,65 with zinc finger nucleases (ZFNs) or transcription activator-like effector nucleases (TALENs); together these techniques allow the functions of genes to be dissected in vivo. This is essential, as functional analyses demonstrating that zebrafish can develop pathologies relevant to OA will be required to establish zebrafish as a model for OA.
A major stumbling block in the management of OA is the paucity of pharmaceutical therapies. The development of zebrafish tools relevant to the study of OA raises the prospect of using zebrafish for compound screens; the protocols for such screens are well described in zebrafish 62,66 . These have the potential to identify novel modifiers of cartilage and bone phenotypes; streamlining the path into drug discovery programmes to test for compounds with therapeutic properties in OA and related diseases. As expression of type X Collagen is one of the hallmarks of chondrocyte hypertrophy, double transgenic lines of zebrafish carrying reporters for both type II Collagen tg(col2a1BAC:mCherry) hu5900 and Type X Collagen tg(col10a1BAC:citrine) hu7050 , as described here, could be used in drug screens to identify compounds that can block the onset of chondrocyte hypertrophy and thus potentially to prevent OA progression.
In summary, we have described the developmental expression of a number of genes with relevance to OA in the zebrafish and we describe a new transgenic reporter line for Type X collagen, which can be used to study chondrocyte hypertrophy in live fish. This can be used as a platform for further research into the functions of the OA genes described here in the developing skeleton and may have uses in large scale screening programmes.

Author contributions
CLH and SSM designed and conceived the study, CLH conducted experiments and wrote the manuscript. REM, CS, LHB, REHS and LFAH performed and analysed experiments.
The research was funded by an Arthritis Research UK career development award to CLH. SSM is supported by FP7 TREAT_OA, and also gratefully recognizes the support of the Smart Mix Programme of the Netherlands Ministry of Economic Affairs and the Netherlands Ministry of Education, Culture and Science.