Loss of Mob1a/b in mice results in chondrodysplasia due to YAP1/TAZ-TEAD-dependent repression of SOX9

Chondrogenesis and endochondral ossification are key contributors to the development of the vertebral skeleton. During chondrogenesis, mesenchymal stem cells (MSCs) undergo condensation and differentiation into resting chondrocytes, followed by the proliferation of the chondrocytes and their maturation into prehyperplastic and finally hypertrophic chondrocytes (Kronenberg, 2003; Michigami, 2013). Eventually, these terminally differentiated chondrocytes undergo apoptosis during endochondral ossification, leaving a cartilaginous matrix that becomes mineralized and replaced with bone. The processes of chondrogenesis and endochondral ossification are tightly regulated by multiple entities, including transcription factors, growth factors, morphogens and hormones (Melrose et al., 2016).

Among the transcription factors involved in chondrogenesis and endochondral ossification is SOX9. In fact, SOX9, which is a member of the Sry-related high mobility group box (SOX) family, is an indispensable master regulator of chondrogenesis (Bi et al., 1999; Akiyama, 2008; Lefebvre et al., 1998). In humans, heterozygous mutations of the SOX9 gene lead to campomelic dysplasia, which is characterized by severe skeletal malformation (Wagner et al., 1994). Supporting evidence provided by mouse models has revealed that loss of Sox9 results in hypoplastic cartilage (Akiyama et al., 2002). At the molecular level, SOX9 interacts cooperatively with SOX5 and SOX6 to drive chondrocyte proliferation and differentiation (Lefebvre et al., 1998; Akiyama et al., 2002; Smits et al., 2001; Ikeda et al., 2004). Other signaling pathways involving fibroblast growth factors (FGFs) (Ornitz, 2005), bone morphogenetic proteins (BMPs) (Tsuji et al., 2006), parathyroid hormone (Ellegaard et al., 2010), Indian hedgehog (IHH) (Vortkamp et al., 1996) and WNT/β-catenin (Huang et al., 2012) are also key players in chondrocyte differentiation during skeletal development.

Hippo signaling is modulated in response to cell density, external mechanical forces, and rigidity of the extracellular matrix (ECM) (Edgar, 2006; Nishio et al., 2013). The core components of the Hippo pathway are the mammalian STE20-like protein (MST) kinases (Creasy and Chernoff, 1995), the large tumor suppressor homolog (LATS) kinases (Tao et al., 1999), and the adaptor proteins salvador homolog 1 (SAV1) (Valverde, 2000) and Mps one binder kinase activator 1 (MOB1) (Moreno et al., 2001). MOB1A/B are the adaptor proteins for the LATS kinases. By binding to LATS kinases, MOB1A/B strongly increase the kinase activities of these enzymes (Moreno et al., 2001). Activated LATS kinases in turn phosphorylate Yes-associated protein 1 (YAP1) and transcriptional co-activator with PDZ-binding motif (TAZ; also known as WWTR1) (Sudol, 1994; Kanai et al., 2000). YAP1/TAZ are key downstream transcriptional co-factors that act mainly on TEA domain transcription factors (TEADs) to regulate numerous target genes involved in cell growth and differentiation (Zhao et al., 2008). After phosphorylation by LATS kinases, YAP1/TAZ are excluded from the nucleus and retained in the cytoplasm, where they are ubiquitylated by E3-ubiquitin ligase SCFβ^TRCP (also known as BTRC) and subjected to proteasome-mediated degradation (Zhao et al., 2010). Thus, in most cell types, YAP1/TAZ are essentially positive regulators of cell proliferation that are negatively controlled by upstream Hippo core components.

YAP1/TAZ are considered to be key factors in the regulation of MSC lineage commitment. Under the control of SOX2, YAP1 maintains MSC self-renewal and inhibits osteogenic differentiation (Seo et al., 2013). However, some studies have reported that low TAZ expression promotes adipogenesis, whereas high TAZ levels drive osteogenesis (Hong et al., 2005; Cui et al., 2003; Yang et al., 2013). Thus, the functions and molecular mechanisms by which YAP1/TAZ influence mesenchymal cells are complicated and remain largely unknown. It is clear that cartilage is rich in ECM and also subject to strong external forces, both of which are important upstream regulators of Hippo-YAP1/TAZ signaling. In addition, Hippo-YAP1/TAZ signaling has been shown to engage in crosstalk with FGFs (Rizvi et al., 2016), BMPs (Alarcón et al., 2009), IHH (Wang et al., 2016), WNT/β-catenin (Varelas et al., 2010), SOX2 (Lian et al., 2010) and SOX9 (Song et al., 2014), all of which are crucial for chondrogenesis. Nevertheless, the molecular mechanisms by which Hippo-YAP1/TAZ signaling controls chondrocyte generation and homeostasis remain unclear.

Col2a1-Yap1 transgenic mice were recently reported to display increased early chondrocyte proliferation driven by YAP1/TEAD-dependent SOX6 activation, but also exhibited YAP1/RUNX2-dependent COL10A1 inhibition, impaired chondrocyte maturation and reduced skeleton size (Deng et al., 2016). However, chondrocyte-specific Yap1-deficient mice are slightly larger than age-matched wild-type (WT) littermates (Deng et al., 2016). In contrast to YAP1, TAZ overexpression was found to accelerate chondrocyte maturation and promote RUNX2-dependent COL10A1 expression (Deng et al., 2016). It appears that TAZ competes with YAP1 for interaction with RUNX2 in order to modulate COL10A1 expression and control chondrocyte maturation.

We previously reported that Mob1a/b null mutant mice succumb to embryonic lethality at embryonic day (E) 6.5 (Nishio et al., 2013). We have also demonstrated that Mob1a/b loss induces extreme hyperactivation of endogenous YAP1/TAZ, resulting in the most severe phenotypes reported among mice mutated in Hippo core components in various tissues (Nishio et al., 2017). Thus, MOB1A/B is a crucial hub in the Hippo signaling pathway. In this study, we generated chondrocyte-specific Mob1a/b-deficient mice and found that hyperactivation of endogenous YAP1/TAZ induced by loss of Mob1a/b impaired chondrocyte proliferation and differentiation and led to the onset of chondrodysplasia. Our data indicate that these phenotypes occur because a YAP1/TAZ-TEAD complex functions as a transcriptional repressor of SOX9, a master regulator of chondrogenesis.

To analyze the functions of endogenous YAP1/TAZ in chondrogenesis in vivo, we generated chondrocyte-specific Mob1a/b double-knockout mice (Col2a1-CreERT; Mob1a^flox/flox; Mob1b^−/−; hereafter cMob1 DKO) by mating Col2a1-CreERT transgenic mice with Mob1a^flox/flox and Mob1b^−/− mice. Administration of 4-hydroxytamoxifen (tamoxifen) at P0 activates Cre expression, deleting the floxed Mob1a gene. We confirmed efficient Mob1a deletion in Mob1b null chondrocytes by PCR analysis of DNA from chondrocytes isolated from control and cMob1 DKO mice (Fig. S1). cMob1 DKO mice were born at the expected Mendelian ratio but developed slowly, showing an overall reduction in body size at postnatal day (P) 84 compared with littermate controls (Mob1a^flox/flox; Mob1b^−/−) (Fig. 1A).

To study the roles of MOB1A/B during postnatal chondrogenesis, we measured the lengths of the long bones and the size of the cartilaginous growth plates in control and cMob1 DKO mice at P84. Compared with controls, mutants with Mob1a/b deficiency in chondrocytes showed significant decreases in total body length as well as in the length of the femur, tibia, humerus and forelimb (Fig. 1B). The size of the articular cartilage layer was also decreased in the mutants at P12 (Fig. 1C). Close histological examination of growth plates at P21 revealed that each chondrocyte zone (resting, proliferative, and hypertrophic) was present in cMob1 DKO mice but proportionally reduced in size compared with that in control animals (Fig. 1D). Thus, loss of Mob1a/b in chondrocytes results in chondrodysplasia.

Because cMob1 DKO mice exhibited abnormal histology in their growth plate cartilage, we analyzed the proliferation and differentiation of chondrocytes. Histological examination of control and mutant growth plates at P21 using PCNA staining to identify proliferating cells revealed that many PCNA-positive cells were present in control growth plates (as expected), especially in the proliferative zone. However, numbers of PCNA-positive cells in growth plates of P21 cMob1 DKO mice were significantly decreased compared with controls (Fig. 2A). Notably, the percentage of TUNEL-positive apoptotic cells in control and cMob1 DKO growth plates was not significantly different (Fig. S2). To confirm our observations at the molecular level in vitro, we performed siRNA-mediated knockdown of MOB1A/B in the human chondrocyte cell line H-EMC-SS. Depletion of MOB1A/B significantly reduced the proliferation of these cells in vitro (Fig. 2B), indicating that loss of Mob1a/b negatively affects chondrocyte proliferation in vivo and in vitro.

We next examined the expression levels of several genes required for the establishment and maintenance of ECM. We isolated chondrocytes from control and cMob1 DKO mice and used qRT-PCR to assess mRNA levels of Col2a1, Col9a1, Col9a2, Comp, Col11a1 and aggrecan (Acan). In all cases, relative mRNA expression was significantly downregulated in cMob1 DKO chondrocytes compared with controls (Fig. 2C). To evaluate chondrocyte maturation, we used immunohistochemistry to detect stage-specific markers in chondrocytes in growth plates of control and cMob1 DKO mice. Type 2 collagen (Col II) is expressed in all chondrocyte layers, whereas osterix (also known as SP7) and IHH are markers of the prehypertrophic layer, and Col X is specific to the hypertrophic chondrocyte layer. Numbers of chondrocytes expressing these markers were significantly decreased in cMob1 DKO mice (Fig. 2D, Fig. S3), indicating that MOB1 plays a fundamental role in supporting chondrocyte differentiation/maturation.

Lastly, we examined endochondral ossification by assessing the bone volume, osteoid volume, trabecular thickness, trabecular number and osteoblast number of the proximal tibia of control and cMob1 DKO mice (Fig. S4A) at P12, as well as the longitudinal growth rate in the primary spongiosa of this limb (Fig. 2E). We also subjected proximal tibia tissue to Col I staining (Fig. S4B). All of these properties were significantly reduced in the mutant mice compared with controls. As a further corroborating approach, we crossed cMob1 DKO mice with Rosa26-LSL-YFP reporter mice to generate Col2a1-CreERT; Mob1a^flox/flox; Mob1b^−/−; Rosa26-LSL-YFP (mutant) reporter animals. Examination of YFP expression (green) by osteoblasts around bone ECM containing Col I (red) beneath the growth plate from P21 mice confirmed that none of the osteoblasts in the mutant tissue was YFP positive, suggesting that MOB1-deficient chondrocytes did not differentiate into osteoblasts (Fig. S5). Thus, loss of Mob1a/b in chondrocytes inhibits their proliferation, differentiation/maturation, and endochondral ossification, resulting in the chondrodysplasia phenotype observed in cMob1 DKO mice.

To investigate the effects of chondrocyte-specific Mob1a/b loss on Hippo pathway components and downstream effectors, primary chondrocytes isolated from Col2a1-CreERT; Mob1a^+/+; Mob1b^−/−; Rosa26-LSL-YFP (control) and Col2a1-CreERT; Mob1a^flox/flox; Mob1b^−/−; Rosa26-LSL-YFP (mutant) reporter mice were treated with/without 0.1 μM tamoxifen for 96 h in vitro, and YFP⁺ chondrocytes were sorted and used as control and cMob1 DKO reporter chondrocytes. Chondrocytes lacking MOB1A/B showed reduced YAP1 (Ser127) and LATS1 (Thr1079) phosphorylation, and a modest increase in total YAP1 and TAZ proteins (Fig. 3A). No differences were detected in phosphorylated (T138/T180) MST1/2 (also known as STK4/3), total MST1, SAV1 or LATS1. Notably, loss of Mob1a/b significantly downregulated mRNA levels of Sox9, Sox5 and Sox6 compared with controls (Fig. 3B), and SOX9 protein was markedly reduced in Mob1a/b-deficient chondrocytes (Fig. 3C). This latter result was confirmed by immunohistochemical examination of the growth plates from control versus mutant mice at P21 (Fig. 3D). As noted above, SOX9 is a master transcription factor that acts on genes involved in cartilage development and cooperates with SOX5 and SOX6 to regulate chondrogenesis (Akiyama, 2008; Ikeda et al., 2004). In control mice at P21, YAP1 was strongly expressed in the nucleus of prehypertrophic chondrocytes but less so in hypertrophic chondrocytes (Fig. 3D). However, proliferative and hypertrophic chondrocytes from cMob1 DKO mice showed both enhanced YAP1 activation and reduced levels of SOX5, SOX6 and SOX9 (Fig. 3D). Thus, the defect in chondrodysplasia caused by loss of Mob1a/b is very likely to be due (at least in part) to a decrease in expression of SOX9.

To investigate the role of YAP1 in chondrocyte proliferation and differentiation, we generated H-EMC-SS cells that conditionally expressed YAP1(5SA), a constitutively active form controlled using doxycycline (Dox) and a Tet-On system. Overexpression of YAP1(5SA) significantly decreased the proliferation of chondrocytes as determined by the MTS assay (Fig. 4A). In accordance with this result, qRT-PCR revealed that mRNA levels of five genes involved in the establishment and maintenance of ECM were also reduced (Fig. 4B).

To determine the dependence of the chondrodysplasia phenotype of cMob1 DKO mice on YAP1/TAZ, we generated two triple knockout (TKO) mouse strains: TKO(YAP) mice, which were Mob1a/b homozygous deficient plus Yap1 homozygous deficient (Col2a1-CreERT; Mob1a^flox/flox; Mob1b^−/−; Yap1^flox/flox); and TKO(TAZ) mice, which were Mob1a/b homozygous deficient plus Taz homozygous deficient (Col2a1-CreERT; Mob1a^flox/flox; Mob1b^−/−; Taz^flox/flox). The defects in the lengths of the growth plates and the long bones and overall body size were all significantly rescued by either Yap1 or Taz deletion (Fig. 4C,D). These results indicate that, in WT mice, YAP1 functions to inhibit chondrocyte proliferation and maturation, and that the cartilage abnormalities observed in cMob1 DKO mice result from hyperactivation of YAP and/or TAZ activity.

To clarify the links between MOB1A/B, YAP1/TAZ and SOX9, we carried out siRNA-mediated knockdown of MOB1A/B proteins in human H-EMC-SS or mouse ATDC5 cells (Fig. 5A, Fig. S6A). Levels of phosphoYAP1 (Ser127) and SOX9 were decreased whereas total YAP1 and TAZ proteins were increased in these MOB1A/B-depleted cells, as compared with cells transfected with si-scramble control. Transfection of H-EMC-SS chondrocytes with vectors overexpressing human WT YAP [YAP(WT)] or constitutively active YAP1(5SA), or with human WT TAZ [TAZ(WT)] or constitutively active TAZ [TAZ(SA)], showed that overexpression of YAP1 (WT or 5SA) or TAZ (WT or SA) significantly decreased SOX9 protein (Fig. 5B,C). Similarly, knockdown of MOB1A/B (Fig. 5D, Fig. S6B) or overexpression of either YAP1(5SA) (Fig. 5E) or TAZ(SA) (Fig. 5F) resulted in reduced SOX9, SOX5 and SOX6 mRNA levels (Fig. 5E,F, Fig. S6B). Thus, the Hippo-YAP1/TAZ pathway regulates the expression of SOX9, SOX5 and SOX6 mRNAs and thus influences their protein levels.

It is primarily the TEAD family of transcription factors that is modulated by binding to the YAP1/TAZ co-factors. To clarify whether the decrease in the expression of SOX mRNAs depended on TEADs, we analyzed mRNA levels of SOX9, SOX5 and SOX6 after siRNA-mediated knockdown of TEAD1-4. Application of siTEAD1-4 to H-EMC-SS chondrocytes efficiently inhibited the expression of TEAD proteins (Fig. 6A) and led to upregulated production of SOX9, SOX5 and SOX6 mRNAs (Fig. 6B). These data implied that a YAP1/TEAD complex might function as a transcriptional repressor tasked with controlling SOX mRNA expression. To investigate this hypothesis, we engineered H-EMC-SS chondrocytes to express Dox-inducible YAP1(5SA/S94A), which has an additional mutation (S94A) in the TEAD-binding domain that prevents binding to TEADs (Zhao et al., 2008; Shimomura et al., 2014). Expression of YAP1(5SA/S94A) tended to bolster SOX9 and SOX6 mRNA expression and induced a modest but statistically significant increase in SOX5 mRNA (Fig. 6C,D).

We next applied ChIP assays to nuclear extracts of H-EMC-SS cells to determine YAP1(5SA) binding to the SOX9 promoter. Indeed, YAP1(5SA) was strongly recruited to the TEAD binding site of the SOX9 promoter in these H-EMC-SS cells (Fig. 6E,F). These results show that a YAP1/TEAD complex directly binds to the SOX9 promoter, allowing this complex to act as a transcriptional repressor of the SOX9 gene and so exert a profound negative regulatory effect on chondrogenesis.

In humans, achondroplasia (ACH), which is classified as short-limbed dwarfism, is frequently caused by a hereditary autosomal dominant mutation in the proximal tyrosine kinase domain of FGFR3 (Shiang et al., 1994). In mice overexpressing Fgfr3 with the corresponding ACH mutation (Fgfr3^ach), the growth plate cartilage shows reductions in height of both the proliferative chondrocyte zone and the hypertrophic chondrocyte zone (Naski et al., 1998). Because Hippo-YAP1/TAZ signaling has been shown to engage in crosstalk with FGFs (Rizvi et al., 2016), we investigated whether Hippo-YAP1/TAZ signaling might be triggered downstream of FGFR3 engagement. We overexpressed FGFR3 G380R, a constitutively active form of the human mutant protein (Bellus et al., 1995), in H-EMC-SS cells and confirmed high levels of FGFR3 mRNA in the altered cells (Fig. 7A). However, examination of these cells by western blotting showed that this increase in FGFR3 G380R expression did not activate YAP1 or TAZ in either H-EMC-SS or ATDC5 cells (Fig. 7B,C). This was confirmed by immunohistochemical assays to detect YAP1 in the proximal tibia of P14 control mice compared with Fgfr3^ach mice (Naski et al., 1998) (Fig. 7D). Finally, constitutive FGFR3 activation in H-EMC-SS and ATDC5 cells did not increase the mRNA levels of YAP1/TAZ target genes such as CTGF and CYR61 (Fig. 7E,F). These data suggest that Hippo-YAP1/TAZ signaling may contribute to the onset of dwarfism in humans via an independent pathway that does not function downstream of FGFR3.

Although the regulation of YAP1/TAZ is crucial for the commitment of MSCs to differentiate into osteoblasts or adipocytes (Seo et al., 2013; Hong et al., 2005), there have been few studies analyzing the effects of the Hippo signaling pathway on chondrocytes. A recent report showed that homozygous Col2a1-Yap1^Tg/Tg transgenic mice exhibit a chondrodysplasia phenotype (Deng et al., 2016), and we have demonstrated here that MOB1 deletion also results in YAP1/TAZ-dependent chondrodysplasia. Thus, the Hippo-YAP1/TAZ axis must be important for chondrogenesis. Considering that the chondrocyte defects observed in MOB1-deficient mice are more pronounced than those of heterozygous Col2a1-Yap1^Tg/+ transgenic mice, and that MOB1 deletion in other organs consistently results in the most severe phenotypes reported among animals with conditional deletions of Hippo core components (Nishio et al., 2016), the MOB1 adaptors must constitute the most important hub in Hippo signaling.

With respect to the mechanism driving chondrodysplasia in the absence of MOB1, in vitro studies have shown that increased YAP1 activity under conditions of elevated matrix rigidity or high fluid-flow shear stress leads to impaired chondrocyte maturation, whereas downregulation of YAP1 in response to a soft substrate maintains chondrogenic marker expression (Zhong et al., 2013a,b). Other work has demonstrated that overexpression of YAP1 in murine C3H10T1/2 mesenchymal-like cells can inhibit chondrogenic differentiation in vitro (Karystinou et al., 2015). Similarly, in vivo, overexpression of YAP1 in mice attenuates endochondral maturation and inhibits the formation of cartilaginous callus tissue after bone fracture (Deng et al., 2016). We have shown here that MOB1 deletion leading to hyperactive YAP1 expression impairs ECM production (Fig. 2C). All these reports are consistent in their conclusion that increased YAP1 activation blocks chondrocyte differentiation/maturation. However, the function of YAP1 in chondrocyte proliferation, as well as the base functions of TAZ in chondrocytes, remain controversial. Although YAP1 overexpression reportedly increases the proliferation both of the ATDC5 chondrocyte cell line in vitro and murine chondrocytes in vivo (Deng et al., 2016), the expected effects on immature chondrocytes have been difficult to document. In contrast to Deng et al. (2016), our data show that MOB1 deletion leading to YAP1/TAZ activation decreases chondrocyte proliferation, as established by the MTS assay in vitro (Fig. 4A) and by PCNA staining in vivo (Fig. 2A). The fact that both the proliferative and hypertrophic chondrocyte layers in our mutant mice were decreased in size supports our contention that chondrocyte proliferation is impaired when YAP1 signaling is excessive. Deng et al. (2016) reported that TAZ competes with YAP1 for RUNX2 activation and promotes chondrocyte maturation. However, we show here that the phenotypes of MOB1-deficient chondrocytes can be mostly rescued by additional deficiency of YAP1 or TAZ (Fig. 4C,D). In addition, we demonstrate that either constitutively activated YAP1 or activated TAZ can suppress the expression of SOX factors (Fig. 5B,C,E,F), leading us to conclude that there are no functional differences between YAP1 and TAZ in this context. The reasons underlying the discrepancies between our work and that of Deng et al. (2016) are currently unknown.

SOX9 is an indispensable initiator and master regulator of chondrogenesis (Bi et al., 1999; Akiyama, 2008; Lefebvre et al., 1998). As noted above, SOX9 interacts cooperatively with SOX5 and SOX6 to regulate cartilage matrix genes, including Col2a1, Col9a1, Col11a1 and Acan, to drive chondrocyte proliferation and differentiation (Oh et al., 2014). Accordingly, neither Sox5 nor Sox6 expression can be detected in Col2a1-Cre; Sox9^flox/flox conditional knockout mice (Lefebvre et al., 1998; Akiyama et al., 2002; Ikeda et al., 2004). Loss of Sox9 specifically in murine chondrocytes results in severely hypoplastic cartilage (Akiyama et al., 2002), and heterozygous mutations in the human SOX9 gene cause hereditary campomelic dysplasia (Wagner et al., 1994). We found that loss of MOB1 in murine chondrocytes significantly suppressed both protein and mRNA expression of Sox9, Sox5 and Sox6 in a YAP1/TAZ-TEAD-dependent manner (Fig. 3B-D), and that overexpression of YAP1 or TAZ also downregulated SOX9, SOX5 and SOX6 expression levels (Fig. 5B,C,E,F). Deletion of SOX9 in mouse limb buds reportedly abolishes the expression of both SOX5 and SOX6, confirming that SOX9 is the master regulator of chondrogenesis and is necessary for SOX5 and SOX6 generation during chondrocyte differentiation (Akiyama et al., 2002). Within enhancer regions, SOX5 and SOX6 bind to recognition sites near that bound by SOX9, thereby consolidating SOX9 binding to DNA and potentiating SOX9 activity (Liu and Lefebvre, 2015). These observations imply that the reduced expression of SOX5 and SOX6 that accompanies the decrease in SOX9 caused by MOB1 deletion contributes to the chondrodysplasia observed in our mutant mice.

Song et al. (2014) reported that YAP1 directly regulates SOX9 transcription through a conserved TEAD binding site in the SOX9 promoter, and that YAP1 maintains the cancer stem cell properties of esophageal tumor cells by upregulating SOX9 expression (Song et al., 2014). However, we show here that, although YAP1 is indeed recruited to the SOX9 promoter, it induces repression of SOX9 expression in a TEAD-dependent manner (Fig. 6A-F). This situation of context-dependent opposing effects has been frequently observed for various transcription factors (Fry and Farnham, 1999; Kim et al., 2015). In addition, although YAP1/TAZ most often form complexes with TEADs that enhance their activity, recent studies have revealed that YAP1/TAZ can also act as transcriptional co-repressors (Kim et al., 2015; Zaidi et al., 2004; Valencia-Sama et al., 2015). Thus, it is not unreasonable to conclude that a YAP1/TAZ-TEAD complex may function as either a transcriptional activator or repressor of SOX9 expression, depending on the tissue-specific context.

As noted above, ACH is caused by a hereditary autosomal dominant mutation in FGFR3 (Shiang et al., 1994). However, we show here that overexpression of the constitutively activated FGFR3 G380R mutant protein does not activate YAP1/TAZ in vitro (Fig. 7B,C) or in vivo (Fig. 7D). Furthermore, mRNA levels of YAP1/TAZ target genes such as CTGF and CYR61 were not altered in two chondrocyte cell lines (H-EMC-SS and ATDC5) overexpressing FGFR3 G380R (Fig. 7E,F). Our finding that FGF signaling does not activate YAP1/TAZ is in line with a report by another group (Yu et al., 2012) that examined human embryonic kidney cells (HEK293A). Additional studies are required to determine the nature and extent of the crosstalk between the Hippo and FGF signaling pathways.

In conclusion, we have clarified the physiological functions of MOB1-YAP1/TAZ signaling in chondrocytes, and have shown that inappropriate hyperactivation of a YAP1/TAZ-TEAD complex that functions as a transcriptional repressor of SOX9 can lead to the onset of chondrodysplasia in mice. In this light, it would be interesting to analyze the frequency of YAP1/TAZ hyperactivation in dwarfism patients. Our results increase our molecular understanding of the effects of the Hippo signaling pathway in vivo, and might provide new insights into potential therapeutic strategies for dwarfism patients.

Mouse strains used in this study were Col2a1-CreERT Tg (The Jackson Laboratory), Mob1a^flox/flox; Mob1b^−/− (Nishio et al., 2012, 2016), Rosa26-LSL-YFP reporter (Srinivas et al., 2001), Yap1^flox/flox (Knockout Mouse Project Repository, UC Davis, CA, USA), Taz^flox/flox (kindly provided by Dr J. Wrana), and Fgfr3^ach (kindly provided by Dr H. Akiyama).

Mice were kept in pathogen-free facilities at Kyushu and Kobe Universities. Protocols for animal experiments were approved by the Animal Research Committees of Kyushu and Kobe Universities.

Chondrocyte-specific Mob1a/b homozygous double-mutant mice (Col2a1-CreERT; Mob1a^flox/flox; Mob1b^−/−) were generated by mating Col2a1-CreERT Tg with Mob1a^flox/flox; Mob1b^−/− mice. Col2a1-CreERT Tg mice were of the C57BL/6 background, and Mob1a^flox/flox; Mob1b^−/− mice were backcrossed to C57BL/6 for more than six generations. Mob1a^flox/flox; Mob1b^−/− mice without the Col2a1-CreERT transgene were usually chosen to serve as controls because no significant differences in total body length or lengths of long bones and cartilaginous zones were observed between Mob1a^flox/flox; Mob1b^−/− and Col2a1-CreERT; Mob1a^flox/flox; Mob1b^−/− mice that were injected with 4-hydroxytamoxifen (Sigma-Aldrich). To delete the floxed Mob1a gene, a single dose of tamoxifen (0.1 mg) was injected into Col2a1-CreERT; Mob1a^flox/flox; Mob1b^−/− and control pups at P0. Primers used for genotyping PCR are listed in Table S1.

The human chondrocyte cell line H-EMC-SS (Riken Cell Bank, Tsukuba, Japan) was maintained in MEMα medium (Wako) supplemented with 10% heat-inactivated fetal calf serum (FCS), penicillin (100 U/ml) and streptomycin (100 μg/ml) and cultured in a humidified incubator at 37°C and 5% CO₂. The mouse embryonal carcinoma-derived chondrogenic cell line ATDC5 (Riken Cell Bank) was maintained in a 1:1 mixture of DMEM and Ham's F-12 medium (Wako) containing 5% FCS in a humidified atmosphere at 37°C and 5% CO₂.

The ventral parts of rib cages of P2 mice were digested with collagenase D and chondrocytes were isolated as described previously (Beier et al., 1999). Chondrocytes isolated from control mice (Col2a1-CreERT; Mob1a^+/+; Mob1b^−/−), or from Col2a1-CreERT; Mob1a^flox/flox; Mob1b^−/− mice carrying the Rosa26-LSL-YFP reporter allele, were plated in 6-well dishes at 5000 cells/cm² and grown to confluence in DMEM containing 10% FCS. Plated chondrocytes were treated with 0.1 μM tamoxifen for 96 h. YFP⁺ cells were collected using an SH800 cell sorter (Sony).

Cell proliferation was measured by the MTS method (Cory et al., 1991). MTS assays were performed using the CellTiter 96 assay (Promega) according to the manufacturer's instructions.

Mouse tissues were fixed in 4% paraformaldehyde in PBS, decalcified in 10% EDTA, embedded in paraffin, and sectioned. Deparaffinized sections were antigen-retrieved using Immunosaver (Nissin EM, Tokyo, Japan), and then incubated with primary antibodies at 4°C overnight. Primary antibodies were against SOX5 (ab94396, Abcam; 1:200), SOX6 (sc-393314, Santa Cruz Biotechnology; 1:100), SOX9 (sc-20095, Santa Cruz Biotechnology; 1:100), PCNA (610664, BD Transduction Laboratories; 1:200), Col I (ab34710, Abcam; 1:100), Col II (LB-1297, LSL, Tokyo, Japan; 1:400), Col X (LB-0092, LSL; 1:200), IHH (ab39634, Abcam; 1:100), osterix (SP7) (ab22552, Abcam; 1:100), YFP/GFP (ab6673, Abcam; 1:500) or YAP1 (WH0010413M1, Sigma-Aldrich; 1:500). Anti-rabbit/mouse-HRP (Dako) was used for DAB staining. Secondary antibodies were tagged with Alexa Fluor 488 or Alexa Fluor 568 (Molecular Probes). In some slides, nuclei were visualized using Hematoxylin and Eosin (H&E) or DAPI.

Apoptosis of chondrocytes was analyzed by TUNEL staining using the In Situ Cell Death Detection Kit (Roche) according to the manufacturer's instructions. Nuclei were visualized with DAPI.

Ethanol-fixed tibiae from P12 mice were fixed in 70% ethanol, embedded in glycol methacrylate resin, and sectioned into 5 μm slices. For histomorphometric analyses, an area (1.62-2.34 mm²) 1.2 mm below the growth plate in the proximal tibia was evaluated. Histomorphometric parameters, such as trabecular bone volume/tissue volume (BV/TV), osteoid bone volume/tissue volume (OV/TV), trabecular thickness (Tb.Th), trabecular number (Tb.N), and osteoblast number/bone surface (N.Ob/BS), were calculated.

Mice received subcutaneous administration of 20 mg/kg calcein. The longitudinal growth rate of the primary spongiosa was measured 24 h later in histological tissue sections as previously described (Pass et al., 2012).

Total RNA was isolated from cells using RNAiso Plus (Takara Bio) according to the manufacturer's instructions. Real-time qRT-PCR analysis was carried out with THUNDERBIRD SYBR qPCR Mix (Toyobo) following the manufacturer's instructions, and with the primers listed in Table S2. PCR amplifications were performed using the StepOne real-time PCR system (Applied Biosystems). Ct values for each gene amplification were normalized by subtracting the Ct value calculated for Gapdh/GAPDH. Normalized gene expression values report the relative quantity of mRNA.

Western blotting was carried out using a standard protocol and primary antibodies recognizing MOB1 (3863, Cell Signaling; 1:1000), MST1 (3682, Cell Signaling; 1:1000), phosphoMST1/2 (3681, Cell Signaling; 1:1000), SAV1 (13301, Cell Signaling; 1:1000), LATS1 (3477, Cell Signaling; 1:1000), phosphoLATS1 (9159, Cell Signaling; 1:1000), YAP1 (4912, Cell Signaling; 1:1000), phosphoYAP1 (4911, Cell Signaling; 1:1000), TAZ (V386) (4883, Cell Signaling; 1:1000), SOX9 (sc-20095, Santa Cruz Biotechnology; 1:500), FGFR3 (sc-13121, Santa Cruz Biotechnology; 1:500), GAPDH (sc-25778, Santa Cruz Biotechnology; 1:1000) and actin (A2066, Sigma-Aldrich; 1:1000). Primary antibodies were detected using HRP-conjugated secondary antibodies (Cell Signaling).

siRNAs targeting MOB1A/B or TEAD1-4 expression are listed in Table S3. Transfection of siRNA oligonucleotides (30 nM) into H-EMC-SS cells was performed using Lipofectamine RNAiMAX (Invitrogen) following the manufacturer's protocol. Transfection of empty pcDNA3.1 vector, or pcDNA3.1 vector expressing FGFR3 G380R (Bellus et al., 1995), into H-EMC-SS cells was performed using Lipofectamine 2000 (Invitrogen) following the manufacturer's protocol. Transfection of empty pcDNA3.1 vector, pcDNA3.1 vector expressing FGFR3 G380R, or siRNA oligonucleotides (30 nM) into ATDC5 cells was performed using FuGENE HD Transfection Reagent (Promega) following the manufacturer's protocol. After 24 h growth to achieve confluence, siRNA-transfected or cDNA-transfected ATDC5 cells were cultured for 6 days in a 1:1 mixture of DMEM and Ham's F-12 medium containing 5% FCS and 10 µg/ml insulin.

Lentiviruses expressing YAP1(WT), YAP1(5SA), YAP1(5SA/S94A), TAZ(WT) or TAZ(SA) were produced by transient transfection of HEK293T cells with pMDLg/pRRE, pRSV-Rev, pMD2.G, and either pSLIK-Flag-Myc-YAP1(WT), pSLIK-Flag-Myc-YAP1(5SA), pSLIK-Flag-Myc-YAP1(5SA/S94A), pSLIK-Flag-His-TAZ(WT) or pSLIK-Flag-His-TAZ(SA) using Lipofectamine 2000 (Invitrogen) (Otsubo et al., 2017). At 48 h post-transfection, lentivirus-containing supernatant was collected. Cultured H-EMC-SS cells were incubated with lentivirus supernatant for 24 h and then transferred to growth medium containing G418 to select for stable transfectants.

ChIP assays were performed as described (Goto et al., 2015). Briefly, cells were cross-linked with formaldehyde and homogenized by sonication. Precleared chromatin was incubated with either anti-DYKDDDDK (FLAG) tag antibody beads (Wako) or mouse IgG, followed by precipitation with Protein G Sepharose 4 Fast Flow resin (Amersham Biosciences). Semi-quantitative PCR analysis was performed using KAPA Taq polymerase (Kapa Biosystems). Quantitative PCR analysis was carried out using THUNDERBIRD SYBR qPCR Mix (Toyobo). Primers used for PCR in ChIP assays are listed in Table S4.

Data are presented as mean±s.d. Statistical significance of differences between experimental groups was determined using Student's t-test. P<0.05 was considered statistically significant.

We thank H. Akiyama (Gifu University) and J. Wrana (Lunenfeld-Tanenbaum Research Institute) for the Fgfr3^ach mutant and Taz^flox/flox mice, respectively; A. Ito (Ito Bone Histomorphometry Institute), A. Fujimoto, M. Kamihashi and M. Suzuki (all of Kyushu University) for expert technical assistance; and K. Nakao (Kyoto University) for critical discussions.