ATX-LPA1 axis contributes to proliferation of chondrocytes by regulating fibronectin assembly leading to proper cartilage formation

The lipid mediator lysophosphatidic acid (LPA) signals via six distinct G protein-coupled receptors to mediate both unique and overlapping biological effects, including cell migration, proliferation and survival. LPA is produced extracellularly by autotaxin (ATX), a secreted lysophospholipase D, from lysophosphatidylcholine. ATX-LPA receptor signaling is essential for normal development and implicated in various (patho)physiological processes, but underlying mechanisms remain incompletely understood. Through gene targeting approaches in zebrafish and mice, we show here that loss of ATX-LPA1 signaling leads to disorganization of chondrocytes, causing severe defects in cartilage formation. Mechanistically, ATX-LPA1 signaling acts by promoting S-phase entry and cell proliferation of chondrocytes both in vitro and in vivo, at least in part through β1-integrin translocation leading to fibronectin assembly and further extracellular matrix deposition; this in turn promotes chondrocyte-matrix adhesion and cell proliferation. Thus, the ATX-LPA1 axis is a key regulator of cartilage formation.


Results
Loss of ATX-LPA 1 signaling results in dyschondroplasia in zebrafish. LPA-related genes are highly conserved in vertebrates. In zebrafish and mice, the amino acid sequences of ATX and LPA 1 are 67 and 85% identical, respectively. The genes for ATX and the six LPA receptors in vertebrates are completely conserved in zebrafish 13,14 . We employed TILLING 15 and identified a zebrafish lpa 1 mutant with a vu374 mutation at G309 that results in a premature stop codon (Fig. S1a) in the first extracellular loop of LPA 1 . Homozygous lpa 1 mutants are able to reach adulthood and are fertile but display a craniofacial malformation: a round-shaped cephalic region that is also phenocopied in lpa 1 mutant mice 6 (Fig. 1a).
A search for other mutant and morphant zebrafish with similar phenotypes identified several cases, e.g. the foxe1 mutant. In most of these cases, cartilage formation was impaired 16,17 . In addition, lpa 1 mRNA was highly expressed at 72 and 96 hour post fertilization (hpf) in the zebrafish embryos, and the expression pattern overlapped with that of sox9a mRNA (Fig. 1c), a marker of chondrocytes, which shows that chondrocytes express lpa 1 in the cartilage. Staining the cartilage of the wild type and lpa 1 mutant with alcian blue revealed that the mutant embryos had disorganized jaw cartilage at both 96 (Fig. 1a) and 120 hpf (data not shown). Both Meckel's and ceratohyal cartilages (Fig. S1b), the two main cartilages of the lower jaw, were deformed. The lengths of Meckel's and ceratohyal cartilages were shorter in the lpa 1 mutants (Fig. 1a and Fig. S1c). Similar abnormalities were observed when LPA 1 was down-regulated either by injection of a morpholino antisense oligonucleotide (MO) against LPA 1 13 or by treatment with an LPA 1 antagonist, Ki16425 ( Fig. 1b and Fig. S1d-f), which is known to be active against zebrafish LPA 1 14 . ATX is an LPA-producing enzyme that was found to be expressed in the cartilage (Fig. 1c). Knockdown of ATX with a high dose of ATX MO1 (2.5 ng) was previously shown to induce severe vascular defects 14 . When a low dose of ATX MO1 (0.3 ng) or ATX MO2 (3.2 ng) was injected, most of the embryos survived at 120 hpf but had rounded-shaped heads ( Fig. 1b and Fig. S1e), similar to the heads of LPA 1 mutant embryos. ATX morphant embryos also displayed impaired cartilage formation as illustrated by the loss of gill cartilage, i.e. deformation of Meckel's and ceratohyal cartilages (Fig. 1b). We conclude that the loss of LPA 1 signaling results in dyschondroplasia in zebrafish embryos and that ATX is the main LPA-producing enzyme in cartilage tissues.
To determine how loss of ATX-LPA 1 signaling affects the behavior of chondrocytes in cartilage tissues, we employed col2:EGFP transgenic zebrafish, which expressed EGFP protein specifically in chondrocytes under the control of the col2a1a promoter 18 (Fig. 1d). At 120 hpf, chondrocytes in both Meckel's and ceratohyal cartilages maintained their intercalated and stacked organization in control embryos (Fig. 1d). In contrast, uneven sized and irregularly aligned chondrocytes were observed in LPA 1 and ATX MO-injected embryos (Fig. 1e) as well as LPA 1 antagonist (Ki16425)-treated embryos (data not shown).
The cartilage elements of the jaw are largely derived from cranial neural crest cells (CNCCs) that arise from dorsal and lateral regions of the neural ectoderm at 12 hpf and migrate into the area of the pharyngeal arches at 24-48 hpf, where the cells differentiate into chondrocytes to form the jaw cartilage at 72 hpf 17,19 . The expression patterns of slug and sox10, which are markers of CNCCs, were normal in both LPA 1 and ATX morphant embryos (Fig. S1g), and the expression pattern of sox9a was also normal (data not shown). Thus, the migration of CNCCs and the maturation and differentiation of chondrocytes from CNCCs appeared to be unaffected by the loss of ATX-LPA 1 signaling. It should be noted that bone tissues are not formed in zebrafish embryos until 120 hpf 19 , suggesting again that ATX-LPA 1 signaling has a critical role in chondrogenesis.
Loss of ATX-LPA 1 signaling results in dyschondroplasia in mice. We next examined the role of ATX-LPA 1 signaling in cartilage formation in mice. LPA 1 KO mice showed reduced anteroposterior growth of skull bone, and shorter femur, tibia and humerus (Fig. 2a-e). We focused on cartilage tissues in the cranial base (Data are mean ± s.d., n = 8-10, * P < 0.05, ** P < 0.01, *** P < 0.001) (f) Loss of ATX-LPA 1 signaling leads mislocalization of chondrocytes in mice. Sections of intersphenoid synchondrosis in wt, LPA 1 KO and ATX flox/− mice at P0 were stained with H&E or immunostained with anti-Col II antibody. Scale bar: 20 μm and 5 μm in magnified view. (g) The numbers of chondrocyte in intersphenoid synchondrosis per unit area of wt and LPA 1 KO mice at P0, shown by relative ratio (Data are mean ± s.d., n = 4, ** P < 0.01).
Scientific RepoRts | 6:23433 | DOI: 10.1038/srep23433 (Fig. S2a) because the base is important for anteroposterior growth of skull bone. In fact, many mutant mice with defects in the base showed abnormal formation of skull bones like LPA 1 KO mice [20][21][22] . We found that intersphenoid synchondrosis (the cartilage that links bones at the cranial base) ossified earlier in LPA 1 KO mice at 3 weeks of age (Fig. S2b). At the cellular level, alignment of chondrocytes in the cartilage tissue was disturbed (Fig. 2f) and the number of cells was also significantly lower in LPA 1 KO mice (Fig. 2g). Similar mislocalization of the chondrocytes was observed in other cartilage tissues such in the costa and femur (Fig. S2c).
Since global ATX KO mice are embryonic lethal because of impaired vascular formation 9 , we set out to produce conditional ATX KO mice. We produced mice with various combinations of ATX wild type, ATX-flox and null alleles, i.e., ATX +/+ , ATX +/flox , ATX flox/flox , ATX +/− and ATX flox/− mice. We found that one flox allele insertion significantly decreased the serum ATX activity about 15% (Fig. S2d). Interestingly, mice with both ATX-flox and ATX-null alleles (ATX flox/− mice) showed phenotypes similar to those of LPA 1 KO mice. Because ATX +/− mice as well as mice with other genotypes did not show obvious abnormality at all, it was speculated that the significant difference between ATX +/− and ATX flox/− (~50% minus ~35%, i.e., ~15%) was important for normal cartilage formation. Alternatively, it is possible that the insertion of flox allele impairs the transcription or stability of ATX pre-mRNA in a specific cell type, which affect the normal cartilage formation. From P0 to the adult stage, the ATX flox/− mice displayed obvious abnormalities in craniofacial morphology and anteroposterior growth of the skull bone, and shorter femur, tibia and humerus (Fig. 2a-e). The intersphenoid synchondrosis ossified significantly earlier as well in the ATX flox/− mice (Fig. S2b). In addition, at the cellular level, the alignment of chondrocytes was significantly disturbed in ATX flox/− mice (Fig. 2f).
As was observed in zebrafish, loss of ATX-LPA 1 signaling did not affect the differentiation of chondrocytes, since LPA 1 KO did not affect the expressions at P0 of type II collagen (Col II), a marker of resting and proliferating chondrocytes or type X collagen (Col X), a marker of pre-hypertrophic and early hypertrophic chondrocytes (Fig. S2a,e). Taking account of the fact that both LPA 1 and ATX were highly expressed in the cartilage tissues in both zebrafish and mice ( Fig. 1c and Fig. S2f), we conclude that ATX-LPA 1 signaling functions after chondrocyte differentiation and that dyschondroplasia is the main defect of LPA 1 KO and ATX flox/− mice, and is the cause of craniofacial abnormalities and retardation of physical growth in these mice.

Inhibition of ATX-LPA 1 signaling delays S-phase entry in chondrocytes.
To examine the effect of LPA signaling on cellular functions of chondrocytes, we established primary cultured chondrocytes from rib cages. We found that proliferation of LPA 1 −/− chondrocytes was significantly slower than that of LPA 1 +/+ and LPA 1 +/− chondrocytes (Fig. 3a). In addition, the cell size of LPA 1 −/− chondrocytes was much smaller (Fig. S3a). We didn't observe any significant changes in the proliferative activity ( In serum-free medium, LPA significantly stimulated S phase entry of LPA 1 +/− chondrocytes whereas the effect was not observed in LPA 1 −/− chondrocytes or Ki16425-treated LPA 1 +/− chondrocytes (Fig. 3e,f). In vivo imaging of LPA 1 KO mice at P0 using 5-ethynyl-2′-deoxyuridine (EdU) showed that the proliferation of chondrocytes in the intersphenoid synchondrosis was significantly decreased (Fig. 3g,h). These in vivo and in vitro findings indicate that in the absence of LPA 1 signaling, the G 0 /G 1 -to-S phase transition is prolonged, which explains the reduction of cell proliferation and decreased cell number of chondrocytes in the intersphenoid synchondrosis ( Fig. 2g and 3g, h).  LPA enhances fibronectin assembly through LPA 1 . FN plays an important role in cell adhesion, which, in turn, affects the proliferation and survival of many cell types 23 . In LPA 1 +/− chondrocytes at 24 hours after LPA stimulation, FN was distributed in filamentous structures, which overlapped with F-actin and β 1-integrin (Fig. 5a). In contrast, in LPA 1 −/− chondrocytes and LPA 1 +/− chondrocytes treated with Ki16425, Y27632 or PTX, such filamentous structures were less developed (Fig. 5a). In the presence of LPA, extracellularly added fluorescently labeled FN (Hilyte-488 FN) was readily incorporated into the filamentous structures and colocalized with F-actin and β 1-integrin in LPA 1 +/− but not LPA 1 −/− chondrocytes (Fig. 5b). It thus appears that FN, once secreted from chondrocytes, is incorporated and assembled into filamentous structures in an LPA 1and integrin-dependent manner. Consistent with this notion, LPA-induced FN assembly was prominently suppressed by Ki16425, Y27632 and PTX, and partially inhibited by GRGDSP peptide but not by GRGESP (Fig. 5c). Importantly, most of the EdU-positive cells showed the filamentous FN structures (Fig. 5d,e). On the basis of these observations, we hypothesized that LPA stimulates S-phase entry by enhancing FN assembly downstream of LPA 1, through β 1-integrin activation. To further confirm that the interaction of β 1 integrin and FN transmits the intracellular signaling, we examined the formation of focal adhesions which are known to be activated by the coordinated action of β 1-integrin and FN. We observed that LPA 1 signaling promoted focal adhesion assembly as judged by colocalization of vinculin and actin (Fig. S5). Focal adhesions distributed peripherally in non-stimulated cells even in LPA 1 −/− chondrocytes. But, only in LPA 1 +/− chondrocytes, LPA promoted focal adhesions larger and some of them existed under the cell bodies. and this conformational change was also inhibited by Ki16425, Y27632 and PTX (Fig. S5). These observations confirmed the idea that intracellular signaling via β 1-integrin was activated downstream of LPA 1 signaling.  (Fig. 6a). Furthermore, LPA 1 −/− chondrocytes had significantly less deoxycholate-insoluble FN protein (Fig. 6b) and significantly less total ECM proteins (Fig. 6c). Under these conditions, both types of chondrocytes showed equal cell numbers per well (Fig. S6a) and FN and Col II mRNA levels (Fig. 6d), suggesting that deposition of ECM components was attenuated in LPA 1 −/− chondrocytes. Transmission electron micrographs of LPA 1 +/− and LPA 1 −/− chondrocytes confirmed that thick filamentous bundles were better developed in LPA 1 +/− chondrocytes than in LPA 1 −/− chondrocytes (Fig. S6b). We next compared the abilities of decellularized ECMs (Fig. S6c) formed by LPA 1 +/− and LPA 1 −/− chondrocytes to support cell proliferation of LPA 1 +/− chondrocytes. After 5 days of culture, ECMs formed by LPA 1 +/− chondrocytes had significantly more cells than ECMs formed by LPA 1 −/− chondrocytes (Fig. 6e). Thus LPA 1 signaling regulates the nature of the ECM and provides the proper external milieu for cell proliferation.

Discussion
LPA 1 KO mice show obvious craniofacial abnormalities, retardation of physical growth and a low bone mass 6,7 . Previous reports indicated that loss of LPA 1 signaling resulted in impaired differentiation of osteoclasts and osteoblasts in vitro 7,24,25 . However, the fundamental causes of the defects of LPA 1 KO mice has remained unclear. In this study we showed that LPA 1 is highly expressed in chondrocytes in both fish (Fig. 1c) and mice (Fig. S2f), and has a critical role in stimulating the proliferation and positioning of chondrocytes ( Fig. 1e and 2f), thereby promoting proper cartilage formation. Bones, especially long limb bones, are principally formed by a process called endochondral ossification, in which cartilage is replaced by bone and chondrocytes serve as a center of bone growth 26 . Because the cartilage phenotype was observed before bone tissues had formed, we conclude dyschondroplasia is the primary cause of impaired bone development and retarded physical growth of LPA 1 KO mice. We also showed that ATX is highly expressed in chondrocytes of both zebrafish and mice, and that inactivation of ATX in both species phenocopied dyschondroplasia as observed when LPA 1 was attenuated or deleted. We therefore conclude that ATX is the major LPA-producing enzyme in cartilage tissues.
The dyschondroplasia of LPA 1 KO and ATX flox/− mice can be attributed to dysfunction of chondrocytes. In vitro experiments revealed that the ATX-LPA 1 axis has a pivotal role in chondrocyte proliferation. We also examined the effect of an ATX inhibitor and LPA 1 antagonist on cell proliferation of various cell types and found that chondrocytes are more sensitive to LPA 1 signaling than other cell types (data not shown). The ATX-LPA 1 axis promoted cell cycle progression of chondrocytes, in agreement with LPA having growth factor-like activities 3,4,27,28 and anti-apoptoc effects as observed in the nervous system 5,29,30 . Pathophysiological effects of LPA as a growth factor remain less clear, however it can disrupt normal processes relevant to neurodevelopmental disorders like hydrocephalus and schizophrenia 31,32 . To the best of our knowledge, the present study is the first to demonstrate a physiological meaning for LPA-induced cell proliferation through cell cycle progression.
Our present results raise a new question: how does LPA support the proliferation of chondrocytes downstream of LPA 1 ? Kingsbury et al. reported that LPA-induced cell growth of neural progenitor cells is not due to increased proliferation but rather to reduced cell death via LPA 1/2 5 . However, this is not the case, because we did not observe any sign of cell death when LPA 1 signaling was attenuated (Videos 1-4). These observations also exclude the possibility that loss of LPA 1 signal induced anoikis in chondrocytes. Clues to answering this question are that LPA 1 −/− chondrocytes were less adhesive to ECM than LPA 1 +/− chondrocytes (Fig. 4d,e), and that LPA-induced chondrocyte proliferation was dramatically enhanced by the presence of FN and suppressed by an integrin-blocking RGD peptide (Fig. 4b,f). These findings suggest that LPA supports chondrocyte proliferation by upregulating β 1-integrin-mediated cell adhesion to FN. Indeed, mice in which cartilage-specific β 1-integrin is conditionally knocked out exhibited dyschondroplasia similar to that of LPA 1 KO mice 33 . Furthermore, topical administration of anti-α5β 1-integrin antibody or RGD-containing peptide to the upper limbs of mouse fetuses suppressed chondrocyte proliferation and shortened the upper limbs 34 . Together, these findings indicate that β 1-integrin and LPA 1 have closely related roles in the formation of cartilage.
Another important observation for understanding the mechanism of LPA-induced proliferation is that FN was assembled to form filamentous structures (Fig. 5a-c and Fig. S5). We presumed that the filamentous FN structure is a multimeric FN, i.e., FN fibril. FN fibrils within the ECM play central roles in both physiological and pathological processes during development and tissue regeneration by coordinating cell adhesion, growth, migration and differentiation. FN is assembled to form FN fibrils in an integrin-and actin stress fiber-dependent manner [35][36][37][38] . In vitro studies using fibroblasts and platelets have suggested that integrins and some agonists for G protein-coupled receptors including LPA are involved in the assembly of FN fibrils [39][40][41] . The present results clearly demonstrate that LPA 1 signaling induced assembly of the filamentous FN structures undersurface of chondrocytes; the FN structures then modified the proliferative and adhesive property of the cells. The structures were co-localized with actin stress fiber and β 1-integrin (Fig. 5a,b) and their formation was canceled by Y27632 and PTX (Fig. 5a,c). The coordinated changes in integrins and the actin cytoskeleton observed here parallel cadherin and focal adhesion assembly associated with actin changes observed in Schwann cells of the nervous system 42 . All together, these findings support the idea that the filamentous FN structure is an FN fibril and that the LPA-LPA 1 axis regulates actin stress fiber formation and then β 1-integrin translocation. These changes stimulate the FN assembly to form FN fibrils, which enhances the proliferation of chondrocytes (Fig. 7). This model is further

axis-induced cell proliferation via integrin-dependent fibronectin assembly in an inside-outside-in manner.
In cartilage tissues, LPA produced by ATX activates LPA 1 in chondrocytes (i); LPA 1 signaling induces actin stress fiber formation possibly through G α12/13 and G αi (ii); β 1-integrin is translocated along with actin stress fibers (iii); the translocated β 1-integrin then induces fibronectin assembly (fibronectin fibrils) (iv); fibronectin fibrils promote ECM deposition (v); ECM thus forms and then supports the proliferation of chondrocytes by promoting cell adhesion (vi). All these processes contribute to normal cartilage formation.
Scientific RepoRts | 6:23433 | DOI: 10.1038/srep23433 supported by the observation that the ECM formed by chondrocytes in the presence of LPA 1 signaling supports cell proliferation more efficiently than ECM formed in the absence of LPA 1 signaling (Fig. 6b). The observation also implicates that integrin activity itself can be regulated in both an inside-out and outside-in manner 43,44 . The present results indicate that the ATX-LPA 1 axis regulates integrins in an inside-out manner (from step (i) to (v) in Fig. 7), which results in the formation of FN fibrils and the subsequent organization of other ECM components such as Col II. The ECM thus formed then regulates integrins in an outside-in manner (from step (v) to (vi) in Fig. 7), which contributes to proper cell adhesion and proliferation of chondrocytes, leading to normal cartilage development. Thus, in addition to the previously indicated LPA 1 -G αi -Akt pathway 45 , we propose a new model of ATX-LPA-LPA 1 axis-induced cell proliferation that is mediated by integrin-dependent fibronectin assembly, occurring in an "inside-outside-in" manner (Fig. 7).
A key question is whether cells other than chondrocytes utilize this system for their proliferative activity. Chondrocytes and fibroblasts are both derived from mesenchymal stem cells. LPA 1 KO mice were resistant to bleomycin-induced lung fibrosis 46 and unilateral ureteral obstruction-induced renal fibrosis 47 . Thus it is possible that LPA-LPA 1 signaling upregulates proliferation of fibroblasts in a manner similar to what we propose in chondrocytes. In addition, because LPA 1 and ATX are overexpressed in several cancers 8,48,49 , the enhanced proliferation of certain LPA 1 -positive cancer cells, such as glioblastoma cells 50 , may reflect a similar mechanism, which is currently being examined.

Zebrafish lines. Zebrafish were maintained according to the Guidelines for Animal Experimentation
of Tohoku University and the protocol was approved by the Institutional Animal Care and Use Committee at Tohoku University. Wild-type AB zebrafish were obtained from Zebrafish International Resource Center (University of Oregon, Eugene, OR). The LPA 1 mutant was generated by target-selected mutational inactivation of genes (TILLING) according to standard methods 53 . Fish were maintained at 27-28 °C under a controlled 13.5 h light/10.5 h dark cycle. Embryos were obtained from natural spawning and staged according to morphology. The standard staging of zebrafish embryos is used and determined in hpf (hour post fertilization) or dpf (day post fertilization) at 28 °C 54 .

Generation of col2a1:egfp transgenic lines.
To generate a col2a1:egfp transgenic line, the col2a1 promoter 18 was introduced upstream of egfp using the Tol2kit system 55 . Briefly, the col2a1 promoter was introduced into p5E-MCS vector, and then p5E-col2a1, pME-EGFP, p3E-polyA and pDestTol2pA2 were combined with LR clonase II plus (Life Technologies). Twenty-five ng of the DNA plasmid was injected into the embryos at the 1-cell stage to establish a col2a1:egfp transgenic line that selectively expresses EGFP in cartilage.
Microscopic Analysis. Zebrafish larvae were anesthetized with 0.016% tricaine methanesulfonate solution (Sigma-Aldrich) and positioned in 3% methylcellulose (Sigma-Aldrich) on a slide glass. Images were captured with a Leica M80 stereomicroscope equipped with a Leica DFC425 digital camera (Leica Microsystems). Col2a1:egfp transgenic fish larvae were positioned in 1.5% agarose (Sigma-Aldrich) on a glass-bottom well (MatTek, MA) and imaged with a LSM 700 confocal laser-scanning microscope (Carl Zeiss).
Alcian blue staining. Zebrafish cartilage and bone were double stained with alcian blue and alizarin red as described previously 56 . At 4 or 5 dpf, larvae were fixed with 4% PFA in PBS, rocked at room temperature for 2 h, washed and dehydrated with 50% EtOH at room temperature for 10 min. After removing the EtOH, the larvae were double stained with acid-free stain solution (0.02% alcian blue, 0.005% alizarin red, 150 mM MgCl 2 in 70% EtOH), rocked overnight at room temperature, washed with water, depigmented with bleaching solution (1.5% H 2 O 2 in 1% KOH), rocked at room temperature for 20 min, and cleared with successive changes of a solution of glycerol and KOH. The lengths of Meckel's and ceratohyal cartilages were measured with a Zeiss Axio Imager (Carl Zeiss MicroImaging). Larvae with smaller or abnormally bent cartilage compared with control larvae were classified as 'malformed' larvae. Ki16425 treatment for zebrafish embryos. Embryos were treated with 120 μM of Ki16425 in the embryo medium with 1% DMSO 57 from 48 hpf. Embryo medium containing Ki16425 was replaced approximately every 24 hours.
In the 198-bp PCR products derived from the wild-type allele with LPA 1 -180 and LPA 1 -182, an ScrFI restriction site was introduced, and therefore the ScrFI treatment degraded the 198-bp PCR product from the wild-type allele into a 174-bp fragment. The PCR products were cleaved with ScrFI and resolved on a 4% agarose gel. Similarly, the 198-bp PCR products from the mutant alleles with LPA 1 -181 and LPA 1 -182, HphI restriction site were introduced. The HphI restriction enzyme degrades the PCR product from mutant allele into a 163-bp DNA fragment.
Whole mount in situ hybridization. Antisense RNA probes labeled with digoxigenin (DIG) for lpa1, atx, slug, sox10 and sox9a were prepared with an RNA labeling kit (Roche Applied Science). Whole-mount in situ hybridization was performed as previously described 59 . Mice. Mice were maintained according to the Guidelines for Animal Experimentation of Tohoku University and the protocol was approved by the Institutional Animal Care and Use Committee at Tohoku University. The LPA 1 KO mice generated by Contos et al. 6 were transferred and maintained on a mixed 129SvJ/C57BL/6J background. Experiments comparing LPA 1 KO and HT mice used offspring of mice heterozygous for the LPA 1 mutant allele. The ATX conditional KO mice with two floxP alleles (ATX flox/flox ) were generated by van Meeteren et al. 60 . ATX heterozygous KO mice (ATX +/− ) were generated by Tanaka et al. 9 . Mice with a floxP allele and ATX null allele (ATX flox/− ) were generated by crossing ATX +/− and ATX flox/flox mice. Skeletal analysis. Plain radiographs were taken using a soft X-ray apparatus (LaTheta LCT-200, Hitachi-Aloka). The size of the head and the lengths of the femur, tibia and humerus were determined with ORS Visual (Nihon Binary) software. Whole-mount skeletal staining was conducted as described previously, with slight modifications 61 . Briefly, mice were skinned, eviscerated and dehydrated in 95% (v/v) ethanol overnight. Skeletons were stained with 0.3% (w/v) alcian blue (Sigma-Aldrich) for 24 hours, stained with 1.5% (w/v) alizarin red (Wako) for 2 hours, treated with 1% (w/v) potassium hydrate and stored in glycerol/EtOH.

Isolation of primary mouse rib chondrocytes.
Chondrocytes were isolated as previously described 62 .
In short, chondrocytes from rib were isolated from newborn wt, LPA 1 HT or KO mice. Rib cages were dissected, rinsed in PBS, incubated at 37 °C for 45 min in 3 mg/ml collagenase type II (Worthington), cleaned of adherent tissues and digested with 0.5 mg/ml collagenase type II at 37 °C overnight. The cell suspension was filtered through a 40 μm cell strainer. The cells were washed, counted and plated. For all experiments, chondrocytes were plated 48 hr before any treatment.
Evaluation of cell proliferation. Cell proliferation was determined with a Cell Counting Kit-8 (Dojindo).
Cells were plated in 96-well plates at 1 × 10 4 cells per well and cultured in the growth medium. Sixty min before the indicated time points, CCK-8 solution was added to each well and the cell numbers were measured by reading the absorbance (450 nm). To evaluate DNA synthesis, cells were seeded on cell culture-treated or Col II-or FN-coated (Sigma Aldrich, 5 μg/mL each) wells in 96-well plates. After 24 hr starvation, the cells were treated with 10 μM 5′-BrdU (Sigma-Aldrich) and 10% FCS or 10 μM LPA for the indicated times. When an inhibitor was used, cells were treated with Ki16425 (5 μM), ONO-8430506 (10 μM), Y27632 (10 μM), GRGDSP (500 μM) or GRGESP (500 μM) for 30 min before the addition of FCS or LPA. PTX (200 ng/mL) was treated at the same time as starvation. 5′-BrdU incorporation was quantified by counting 5′-BrdU-positive cells. 5′-BrdU was detected by anti-BrdU antibody conjugated to fluorescein (Roche, 1:1000). For EdU (Life Technologies) incorporation in vitro, 10 μM EdU was used instead of 5′-BrdU. To examine the incorporation of EdU in vivo, pups (P0-1) were subcutaneously injected with 50 mg/kg EdU and were sacrificed 2 hours after the injection. Heads of the embryos were dissected and fixed in 4% paraformaldehyde overnight at 4 °C. Tissues were embedded in Optimal Cutting Temperature (OCT) compound and stained according to the manufacturer's instructions. EdU-positive cells were counted in a resting zone within a unit area. All images were captured with the LSM700 confocal microscope equipped with 10×/0.45 M27 Plan-Apochromat.
Tissue and cell staining. For histological analysis, newborn mice were fixed in 4% fresh paraformaldehyde in PBS, pH 7.2, overnight, dehydrated in a graded alcohol series (50,70,90,95, and 99.5%), and embedded in paraffin. Sections were cut at 10 μm and stained with hematoxylin and eosin (Mutoh Industries).

Time-lapse.
In live-cell imaging of chondrocytes, phase-contrast images were taken (LD plan-NEOFLUAR, 20x/0.4, Ph2, 6frames/hr) with a Zeiss inverted microscope (Axio Observer.Z1) equipped with a heated chamber (37 °C) and CO 2 controller (5.0%) over a period of 48 hr. Primary chondrocytes were seeded onto 12-well plates (Greiner), incubated for 2 days at 37 °C and treated with an LPA 1 antagonist (Ki16425, 5 μM) or an ATX inhibitor (ONO-8430506, 10 μM). Live-cell images were taken every 5 or 10 min for 48 hr. We chased cells that divided twice. The doubling time was taken as the time between the two divisions. The duration of the M phase was evaluated by cell morphology. Supplementary Videos 1-4 were simplified as 1.5 frames/hr for downsizing.
Evaluation of cell spreading. Cells were plated on glass coverslips coated with Col II or FN and were stained with phalloidin. Cell images were captured with the LSM700 confocal microscope equipped with 63×/1.40 Oil DIC M27 Plan-Apochromat. and the cell spreading areas were calculated using the Zeiss Efficient Navigation system (Carl Zeiss).
Quantitative RT-PCR. To prepare total RNA from tissues, tissues were first embedded in OCT compound and frozen sections were cut at 25 μm thickness and mounted on poly-l-lysin-coated LCM transfer film (LEICA-BEST) on glass slides. The chondrocytes in the tissue sections were dissected with a Leica LMD7000 Laser Microdissection System. RNA of the harvested chondrocytes was extracted with an RNAqueous Micro Kit (Ambion). Total RNA from cultured cells was isolated using a GenElute Mammalian Total RNA Miniprep Kit (Sigma-Aldrich). Total RNA samples were reverse-transcribed using High-Capacity cDNA RT Kits (Applied Biosystems) according to the manufacturer's instructions. PCR reactions were performed with SYBR Premix Ex Taq (Takara Bio) on an ABI Prism 7300 thermocycler (Applied Biosystems). Standard plasmids ranging from 10 2 to 10 8 copies per well were used to quantify the absolute number of transcripts of cDNA samples. The numbers of transcripts were normalized to the number of transcripts of a house-keeping gene (GAPDH) in the same sample.
Evaluation of ECM amount. After decellularization, ECMs were solubilized with 5.0 M Urea, 2.0 M Thiourea, 50 mM DTT and 0.1% SDS in PBS and scraped with a rubber policeman 63 . The collected lysates were placed in 95 °C water for 5 min and centrifuged at 12000 × g for 10 min at 4 °C. Protein concentration was evaluated by the Bradford method.
Transmission electron microscopy (TEM). Samples were fixed with 2% paraformaldehyde and 2% glutaraldehyde in 0.1M cacodylate buffer pH 7.4 at 4 °C overnight, washed 3 times with 0.1 M cacodylate buffer for 30 min each, postfixed with 2% osmium tetroxide in 0.1M cacodylate buffer at 4 °C for 3hr, dehydrated in graded ethanol solutions (50%, 70%, 90% and 100%), infiltrated with propylene oxide (PO) 2 times for 30 min each, transferred to a 70:30 mixture of PO and resin (Quetol-812, Nisshin EM Co.) for 1h, allowed to stand overnight to volatize the PO, transferred to fresh 100% resin and heated at 60 °C for 48 hr to polymerize the resin. Seventy-nm sections were cut from the polymerized resins (Ultracut UCT, Leica), mounted on copper grids, stained with 2% uranyl acetate at room temperature for 15 min, washed with distilled water, secondary-stained with lead stain solution (Sigma-Aldrich) at room temperature for 3 min and observed with a transmission electron microscope (JEM-1400Plus, JEOL) at an acceleration voltage of 80 kV. Determination of lysophospholipase D activity. Lysophospholipase D activity of mice plasma was determined as described using 14:0 lysophosphatidylcholine as substrate 8 .

In situ
Statistics. All statistical analyses were carried out using EXSAS. Differences were considered significant at P < 0.05. All data are expressed as means ± s.d.