TGFβ and FGF promote tendon progenitor fate and act downstream of muscle contraction to regulate tendon differentiation during chick limb development

Tendon is a connective tissue that transmits the forces generated by muscle to bone and allows body motion. Type I collagen is the main structural and functional component of tendons. The signals regulating the production and the spatial organisation of type I collagen in developing tendons have not been fully identified. Moreover, because type I collagen is not specific to tendons, it is not possible to follow tendon development by mapping collagen expression (reviewed by Gaut and Duprez, 2016). The basic helix-loop-helix (bHLH) transcription factor Scleraxis (Scx) has been identified as an early tendon marker during development. Scx is expressed in tendon progenitors, developing tendons and adult tendons (Schweitzer et al., 2001; Pryce et al., 2007; Mendias et al., 2012). Scx is not the unique master gene driving tendon development, as tendons are formed in Scx^−/− mice, albeit displaying differentiation defects (Murchison et al., 2007). Moreover, Col1a1 expression is downregulated in tendons of Scx^−/− mutant mice (Murchison et al., 2007), consistent with transcriptional regulation of the mouse Col1a1 gene by Scx, via direct binding on the Col1a1 promoter (Lejard et al., 2007). Tenomodulin (Tnmd) and Col14a1 expression is lost in developing limb tendons of Scx^−/− mice (Murchison et al., 2007). The transmembrane glycoprotein Tnmd is considered as a late tendon marker (Jelinsky et al., 2010; Sugimoto et al., 2013) and is highly expressed in embryonic day (E) 14.5 mouse limb tendon cells (Havis et al., 2014). Thrombospondin 2 and 4 (THBS2 and THBS4) were also identified in the transcriptome of mouse limb tendon cells (Havis et al., 2014) and have been shown to be involved in tendon development in mouse, Drosophila and zebrafish (Kyriakides et al., 1998; Subramanian et al., 2007; Subramanian and Schilling, 2014). Two other transcription factors are involved in tendon development: the homeobox protein Mkx (mohawk) (Ito et al., 2010; Liu et al., 2010; Kimura et al., 2011) and the zinc finger transcription factor EGR1 (early growth response factor 1) (Lejard et al., 2011). In contrast to Scx, Mkx and Egr1 are not expressed during early tendon limb development and are not specific to tendons (Anderson et al., 2006; Liu et al., 2006; Ito et al., 2010; Lejard et al., 2011), but they activate Scx and Tnmd expression in various stem cell types and positively regulate type I collagen production in vivo (Ito et al., 2010; Lejard et al., 2011; Guerquin et al., 2013; Liu et al., 2015). In addition to the tendon-related transcription factors, two signalling pathways have been identified as being involved in tendon development: the transforming growth factor-beta (TGFβ) and fibroblast growth factor (FGF) signalling pathways (reviewed by Huang et al., 2015a; Gaut and Duprez, 2016). The TGFβ signalling pathway positively regulates Scx expression in early E9/E10 mouse limb explants (Pryce et al., 2009; Havis et al., 2014). TGFβ function in chick tendon development is less understood. Although TGFβ1 and 2 have been shown to increase SCX and TNMD expression in high-density cultures of chick limb cells (Lorda-Diez et al., 2009), TGFβ1 failed to activate SCX expression in Hamburger Hamilton stage (HH) 20/21 chick limb explants (Lorda-Diez et al., 2010). FGF positively regulates SCX expression in axial and foetal limb tendons during chick development (Edom-Vovard et al., 2002; Brent et al., 2003; Brent and Tabin, 2004). In contrast to the chick model, FGF has an anti-tenogenic effect in mouse embryonic tendon cells (Brown et al., 2014) and inhibition of the ERK MAPK pathway is sufficient to increase Scx expression in early mouse limb explants (Havis et al., 2014). Although the experimental situations in the reports described above differ between the chick and mouse models, they nevertheless suggest a differential regulation of Scx by FGF in the chick and mouse models.

Another important aspect of tendon development is its dependency on muscles. Axial, limb and head tendons require the presence of muscles for full development in chick, mouse and zebrafish embryos (reviewed by Gaut and Duprez, 2016). However, in the absence of muscle, Scx expression is normally initiated (and then lost) in limb and head regions of mouse, chick and zebrafish embryos (Schweitzer et al., 2001; Edom-Vovard et al., 2002; Bonnin et al., 2005; Grenier et al., 2009; Chen and Galloway, 2014; Huang et al., 2015b). The muscle dependency of Scx expression defines two phases for limb tendon formation: a progenitor, muscle-independent phase and a differentiation, muscle-dependent phase. This muscle dependency applies only to stylopod (arm) and zeugopod (forearm) limb tendons, as autopod (digit) tendons are dependent on cartilage in mouse embryos (Huang et al., 2015b). The molecular mechanisms underlying the muscle dependency of chick tendon development remain elusive. Although one can assume that the muscle dependency of Scx expression is related to muscle activity, the requirement of mechanical forces for chick limb tendon development has not been addressed and the molecular signals downstream of muscle contraction involved in tendon differentiation have not been identified.

FGF positively regulates SCX expression via the ERK MAPK signalling pathway in chick somites (Brent and Tabin, 2004; Smith et al., 2005) and in foetal chick limbs (Edom-Vovard et al., 2002; Eloy-Trinquet et al., 2009). However, the role of FGF on SCX expression was not determined in chick limb undifferentiated cells during the muscle-independent phase of limb tendon development. SCX expression is initiated in E3 (HH20) chick limb buds (Brent and Tabin, 2004). At these stages, a source of FGF is observed in the apical ectodermal ridge (Niswander et al., 1994). We implanted FGF4 beads in early chick limb buds (E3 to E4) and analysed SCX expression by RT-q-PCR and in situ hybridisation experiments (Fig. 1A-C). SCX and COL1A2 expression was upregulated as soon as 4 h after FGF4 bead implantation; this upregulation was maintained 24 h after FGF4 bead implantation (Fig. 1A). We used ETV4 (also known as PEA3) and SPRY2 as transcriptional readouts of ERK MAPK activity (O'Hagan et al., 1996; Mason et al., 2006; Havis et al., 2014). The mRNA levels of ETV4 and SPRY2 were increased in FGF4-implanted limbs 4 h and 24 h after grafting (Fig. 1A) and SPRY2 expression was activated around FGF4 beads 24 h after grafting (Fig. 1C). TNMD is not expressed before E5 in chick limbs (Shukunami et al., 2006) and FGF4 was not able to activate TNMD prematurely (data not shown). This FGF4 tenogenic effect in chick limb buds contrasted with the previously demonstrated anti-tenogenic effect of FGF4 in mouse limb explants (Havis et al., 2014). We next performed chick limb bud explants in order to exclude differences due to different experimental designs and allow comparison between the chick and mouse models. Consistent with the in vivo FGF4 bead experiments (Fig. 1A-C), we observed that FGF4 increased the mRNA levels of SCX and SPRY2 (Fig. 1D), whereas blockade of ERK MAPK with the inhibitor PD18 decreased SCX, ETV4 and SPRY2 expression in chick limb bud explants 6 h after treatment (Fig. 1D). In order to allow comparison between species, we performed equivalent mouse limb bud explant experiments and found that FGF4 significantly decreased Scx expression, whereas PD18 increased Scx expression in mouse limbs, 6 h after treatment (Fig. 1E), consistent with previously published effects of 24 h FGF4 and PD18 treatments in mouse limbs (Havis et al., 2014). We conclude that the FGF tenogenic effect observed in chick limb cells is opposite to the anti-tenogenic FGF effect observed in mouse limb cells.

We next tested whether the FGF4 effect on SCX in chick cells involved the SMAD2/3 pathway. We applied the SMAD2/3 inhibitor SIS3 in FGF4 gain-of-function experiments in chick limb buds (Fig. S1). Blockade of SMAD2/3 did not block the positive effect of FGF4 on SCX expression (Fig. S1). This result is consistent with the absence of modification of SMAD7/Smad7 expression upon FGF/ERK MAPK manipulations in both chick and mouse limb explants (Fig. 1D,E). Smad7 is a negative-feedback regulator that is considered to be a general TGFβ/SMAD2/3 transcriptional target gene (Massague, 2012). We conclude that FGF4 positively regulates SCX independently of the SMAD2/3 intracellular pathway in chick limb cells.

TGFβ2 induces Scx expression in E10.5 mouse limb explants (Pryce et al., 2009), but TGFβ1 does not modify SCX expression in E3.5 (HH20/21) chick limb explants (Lorda-Diez et al., 2010). We found TGFB2 to be expressed in ventral parts of E3 chick limb buds (Fig. 2A), as previously described (Lorda-Diez et al., 2010). Application of TGFβ2 beads in E3/E4 (HH19/21) chick limb buds increased SCX expression 6 h after grafting (Fig. 2B) and the mRNA levels of SCX and COL1A2 were increased in grafted limbs compared with control limbs (Fig. 2C). TGFβ2 application on chick limb bud explants also increased SCX expression in addition to increasing THBS2 (Fig. 2D). TGFβ2 was not able to activate TNMD prematurely in chick limb undifferentiated cells (data not shown), as in mouse limb undifferentiated cells (Havis et al., 2014). Blockade of TGFβ receptors (SB43) and of the SMAD2/3 signalling pathway (SIS3) decreased SCX expression, in addition to that of CO1A2, THBS2 and THBS4 (Fig. 2D). Consistently, SMAD7 mRNA levels, the transcriptional readout of the SMAD2/3 intracellular pathway, were increased following TGFβ2 application and decreased with the inhibitors SB43 and SIS3 (Fig. 2D). These results show that TGFβ2 positively regulates SCX expression, and that the SMAD2/3 intracellular pathway is required for SCX expression in early chick limb undifferentiated cells.

TGFβ is known to activate the ERK MAPK pathway as a non-canonical signalling pathway (reviewed by Guo and Wang, 2009; Massague, 2012). The expression of ETV4 and SPRY2 was not modified upon TGFβ2 bead application (Fig. 2C). In order to confirm experimentally that the positive effect of TGFβ2 on SCX expression did not involve the ERK MAPK signalling pathway, we applied the inhibitor PD18 in TGFβ2 gain-of-function experiments in chick limb buds and limb explants. The blockade of the ERK MAPK pathway did not modify the positive effect of TGFβ2 on SCX expression in chick limbs (Fig. 2C) and in chick limb explants (Fig. 2D). We conclude that TGFβ2 activates SCX expression independently of the ERK MAPK signalling pathway in chick limb cells.

TNMD is considered as a late tendon marker in chick and mouse embryos, and is expressed during the differentiation and muscle-dependent phase of limb tendon development (reviewed by Dex et al., 2016). Tnmd mutant mice display an altered structure of collagen fibrils, and reduced self-renewal and increased senescence of tendon progenitors, in post-natal tendons (Docheva et al., 2005; Alberton et al., 2015). TNMD was expressed in SCX-positive tendons in E9 chick limbs (Fig. 3A,B, arrows), but also in dermal regions (Fig. 3B, arrowhead). Retroviral mouse Fgf4 (mFgf4/RCAS) induced ectopic TNMD expression in chick limbs (Fig. 3C-F), in addition to activating SCX expression (Edom-Vovard et al., 2002). Consistently, the relative mRNA levels of TNMD and SCX tendon genes and ETV4 were increased in mFgf4/RCAS-limbs compared with control limbs (Fig. 3D). THBS2, another late tendon marker (Havis et al., 2014) was also upregulated in chick limbs upon retroviral Fgf4 (Fig. 3D,G,H). We conclude that FGF4 positively regulates TNMD and THBS2 expression in chick limbs during the differentiation and muscle-dependent phase of limb tendon development.

Scx/SCX expression is lost in stylopod/zeugopod muscleless limbs of mutant mice or experimental chick embryos (Schweitzer et al., 2001; Edom-Vovard and Duprez, 2004), defining the muscle-dependent phase of limb tendon development. In the absence of muscle activity, Scx/GFP expression is diminished but not lost in zeugopod/stylopod regions of forelimbs of E18.5 paralysed mdg mice (Huang et al., 2015b). In order to determine the importance of mechanical signals for chick limb tendon development, we blocked muscle contraction in chick embryos using the drug decamethonium bromide (DMB). DMB acts as an acetylcholine agonist, induces depolarisation in skeletal muscles and ultimately leads to rigid muscle paralysis and to immobilised embryos (Nowlan et al., 2010). We applied DMB or control buffer in E4.5 chick embryos and analysed gene expression either by in situ hybridisation on sections and wholemounts or by RT-q-PCR (Fig. 4). In the absence of muscle contraction, muscles, visualised with MYOD expression, were present 2 days after DMB application, but displayed splitting delay 3 days after DMB application (Fig. S2). As previously described, limbs of immobilised embryos were smaller than control limbs (Nowlan et al., 2010). During the muscle-independent phase, SCX expression was not affected in chick limbs of immobilised embryos, 24 h after DMB application (Fig. 4A,B), consistent with normal SCX expression in muscleless limbs of E6 experimental chick embryos (Edom-Vovard et al., 2002) and E12.5 mouse Pax3 mutants (Schweitzer et al., 2001). SCX expression was decreased in limbs of immobilised embryos from E6.5 (Fig. 4C-F). In order to confirm the decrease of SCX expression observed by in situ hybridization, we compared SCX mRNA levels in paralysed limbs versus control limbs by RT-q-PCR (Fig. 4G). RT-q-PCR analyses of whole forelimbs, forelimbs without digits, or digits only indicated a decrease of SCX expression in the absence of muscle contraction (Fig. 4G). The expression of COL1A2 was slightly decreased in limbs of immobilised embryos (Fig. 4G), consistent with the general and non-tendon-specific expression of type I collagen. The decrease of SCX expression was more obvious in stylopod/zeugopod regions compared with digits (Fig. 4C-F), consistent with SCX expression pattern in muscleless limbs of chick and mouse embryos (Schweitzer et al., 2001; Edom-Vovard et al., 2002; Bonnin et al., 2005) and with the modular development of mouse limb tendons (Huang et al., 2013, 2015b). Similar SCX downregulation was observed in stylopod/zeugopod tendons of hindlimbs in immobilised chick embryos (Fig. S3). In situ hybridisation to forelimb sections at the levels of the zeugopod (Fig. 4H,I) and digits (Fig. 4J,K) confirmed the more pronounced decrease of SCX expression in zeugopod compared with digits. SCX was also decreased in stylopod/zeugopod tendons of forelimbs, 3 days after injection of pancuronium bromide (PB), an acetylcholine antagonist, which induced flaccid muscle paralysis (Nowlan et al., 2010) (Fig. S4). The expression of the late tendon markers TNMD and THBS2 was also lost in limb tendons of immobilised E7.5 embryos (Fig. 4L-O). We conclude that SCX, TNMD and THBS2 expression is sensitive to mechanical signals provided by muscle contraction in stylopod/zeugopod tendons, during chick limb development.

In order to determine whether the FGF signalling pathway is involved in the downregulation of tendon gene expression in the absence of muscle contraction, we analysed the expression of components of the FGF/ERK MAPK signalling pathway related to tendon development, in immobilised chick embryos. During the muscle-dependent phase of limb tendon development, ETV4, SPRY1 and SPRY2 are expressed ubiquitously in chick limbs but with a high expression at muscle and tendon interface (Eloy-Trinquet et al., 2009). FGF4 is expressed at muscle tips close to tendons (Edom-Vovard et al., 2002), whereas FGF8 is expressed in tendons close to muscles (Edom-Vovard et al., 2001). The expression of ETV4 and SPRY2 was dramatically decreased in limbs of immobilised chick embryos assessed by RT-q-PCR and in situ hybridisation analyses (Fig. 5A-E). The ETV4 and SPRY2 downregulation was more pronounced in forelimbs (digit excluded) compared with digits alone (Fig. 5A). In the absence of muscle contraction, the expression of FGF ligands related to tendon development, FGF4 and FGF8, was lost at muscle tips (Fig. 5F,G, arrows) and in tendons (Fig. 5H,I, arrows), respectively. These results showed that the expression of FGF ligands and transcriptional readouts of ERK MAPK activity was downregulated at the muscle/tendon interface in chick limbs, in the absence of muscle contraction.

In order to determine whether FGF would rescue tendon gene expression in the absence of mechanical signals, we applied mFgf4-expressing retroviruses in chick limbs (mFgf4/RCAS) and injected DMB in order to prevent muscle contraction (Fig. 6A). In the absence of muscle contraction, SCX expression was downregulated (Fig. 6C,D). mFgf4 was able to activate ectopic SCX expression in limbs of immobilised embryos (Fig. 6D-F). Consistent with this, the relative mRNA levels of SCX, ETV4 and SPRY2 were significantly upregulated in mFgf4-paralysed-limbs compared with paralysed limbs (Fig. 6B). Under these experimental conditions, TNMD and THBS2 expression was not changed (Fig. 6B). The relative mRNA levels of TGFB2, TGFB3 and SMAD7 were not changed in the presence of mFgf4 in immobilised embryos (Fig. 6B), indicating that TGFβ signalling was not modified under these experimental conditions. We performed a similar FGF rescue experiment in chick limb explants, in which we considered that the E5 limb explant culture system was devoid of mechanical movements. Analysis of the relative mRNA levels in chick limb explants compared with stage-matched limbs originating from in ovo embryos showed a significant diminution of SCX, TNMD, THBS2, ETV4, SPRY2 and FGF4 gene expression (Fig. 6G), similar to that observed in immobilised chick embryos (Figs 4 and 5). Consistent with the in ovo FGF rescue experiments (Fig. 6A-F), the application of recombinant FGF4 in limb explant cultures induced a significant increase of the mRNA levels of SCX, ETV4 and SPRY2, but did not affect COL1A2 and SMAD7 expression (Fig. 6H). The expression levels of TNMD and THBS2 genes were not increased upon FGF4 treatment and even displayed a decrease of expression (Fig. 6H). We conclude that FGF4 activates SCX but not TNMD or THBS2 expression in chick limbs in immobilisation conditions.

Next, we wanted to determine whether TGFβ was also sensitive to immobilisation. Both Tgfb2 and Tgfb3 have been shown to be involved in mouse limb tendon development (Pryce et al., 2009). In E7.5 limbs, TGFB2 was observed in tendons, in addition to displaying expression in muscles (Fig. S5). TGFB3 was mainly expressed in chick limb muscles, with faint expression in tendons (Fig. S5). In DMB-treated embryos, the mRNA levels of SMAD7 and TGFB2 were decreased in paralysed limbs compared with control limbs (Fig. 7A). TGFB2 expression was lost in limb tendons of the zeugopod regions (Fig. 7B-E, arrows), but not in digits (Fig. 7F-I) of immobilised embryos. The diminution of the relative mRNA levels of SMAD7 and TGFB2 was also observed in limb explants compared with stage-matched limbs originating from in ovo embryos (Fig. 7J). These results show that the TGFβ/SMAD2/3 signalling pathway was decreased in chick limb tendons under immobilisation conditions. Application of TGFβ2 to limb explants increased SCX, TNMD, THBS2 and SMAD7 expression compared with control limb explants (Fig. 7K). The expression levels of the transcriptional readouts of ERK MAPK activity were not modified upon exposure to TGFβ2. We conclude that TGFβ2 is sufficient to maintain the expression of SCX and the tendon differentiation markers TNMD and THBS2 in chick limbs under immobilisation conditions.

Our TGFβ2 gain- and loss-of-function experiments in early chick limbs and explants (Fig. 2) show that TGFβ2 is sufficient and the SMAD2/3 intracellular pathway is required for SCX expression in undifferentiated limb cells. These results are fully consistent with those obtained in early mouse limb explants (Pryce et al., 2009; Havis et al., 2014). These results highlight a universal role for TGFβ in initiating the commitment of undifferentiated limb mesodermal cells towards the tendon lineage during chick and mouse development (Fig. 8). In zebrafish embryos, blocking the TGFβ pathway (using the chemical drug SB431542) inhibits scxa expression (Chen and Galloway, 2014), suggesting that TGFβ is also important for the initiation of scxa expression in fish. The developmental TGFβ tenogenic effect is likely to be related to the recognized effect of TGFβ in positively regulating Scx expression in embryonic tendon progenitors (Brown et al., 2014) and stem cell culture systems (Pryce et al., 2009; Barsby and Guest, 2013; Goncalves et al., 2013; Guerquin et al., 2013; Havis et al., 2014). We find that the positive effect of TGFβ2 on chick limb SCX expression is independent of ERK MAPK signalling (Fig. 2C,D), as also observed in mouse limb explants (Pryce et al., 2009; Havis et al., 2014).

Tnmd/TNMD is one of the tendon markers displaying the highest expression levels in E14.5 mouse Scx⁺ cells but is not expressed in E11.5 mouse limb bud explants (Havis et al., 2014) or E4 chick limbs (Shukunami et al., 2006) or activated by TGFβ2 at these early stages (Havis et al., 2014). However, TNMD expression is activated upon TGFβ exposure in late chick (Fig. 8) and mouse (Havis et al., 2014) limb explants. This is consistent with previous reports showing TNMD upregulation by TGFβ ligands in 3D-culture systems of human tendon cells (Bayer et al., 2014) and of equine embryo-derived stem cells (Barsby et al., 2014), and in high-density cultures of chick limb cells (Lorda-Diez et al., 2009). It is worth mentioning that TGFβ dramatically decreases Tnmd expression (and activates Scx) in 2D-culture systems of embryonic or adult mouse tendon progenitors and in mouse mesenchymal stem cells (Guerquin et al., 2013; Brown et al., 2014; Liu et al., 2015). We believe that the opposite effects of TGFβ on Tnmd expression are due to the different cell contact environments in 2D-culture versus 3D-culture systems.

In vivo and ex vivo experiments demonstrated that FGF activates SCX expression in early chick limb buds. This is consistent with FGF function in somites of chick embryos (Brent and Tabin, 2004; Smith et al., 2005). This result observed in chick embryos is opposite to those obtained in mouse limb explants, in which FGF inhibits Scx expression and ERK MAPK inhibition activates Scx expression (Havis et al., 2014; Fig. 1E). Blockade of SMAD2/3 did not prevent the SCX activation by FGF4 in chick limbs (Fig. S1). Moreover, Smad7/SMAD7 expression was not modified in any of the FGF misexpression experiments (Fig. 1D,E) (Havis et al., 2014), indicating that the TGFβ pathway is not involved in the positive or negative effect of FGF on SCX/Scx expression in chick and mouse, respectively. We believe that FGF has a tenogenic effect in chick undifferentiated limb mesodermal cells, but has an anti-tenogenic effect on mouse undifferentiated limb mesodermal cells. The reasons for the opposite effect of FGF signalling on limb mesodermal cells between the chick and mouse models remain unclear. However, these results are consistent with an absence or deleterious effect of FGF on tendon marker expression in 2D-culture systems of various stem cells, including mouse embryonic tendon progenitors (Brown et al., 2014), mouse mesenchymal stem cells (Havis et al., 2014), canine tendon fibroblasts (Thomopoulos et al., 2010), human amniotic fluid stem cells or adipose-derived stem cells (Goncalves et al., 2013). Consistent with the FGF tenogenic function during chick tendon development, a clear beneficial effect of FGF has been described during tendon repair in a chick digital tendon injury model. The expression of the ligand FGFb is decreased in chick tendons during the process of tendon repair (Chen et al., 2008) and ectopic application of FGF has a beneficial effect on chick tendon repair (Tang et al., 2008, 2014). We conclude that FGF positively regulates SCX in chick limb undifferentiated cells (Fig. 8).

FGF4 and TGFβ2 activate SCX expression independently of each other in early chick limb buds (Fig. 2D; Fig. S1). Although intracellular crosstalk has been identified between the ERK MAPK and SMAD2/3 signalling pathways in many biological systems (reviewed by Massague, 2012), our results indicate that these signalling pathways do not interact in the activation of SCX in chick limb buds. The fact that two signalling pathways activate SCX independently of each other indicates the presence of a safety system for tendon specification in chick limbs. This safety system is classically observed during developmental processes. Another possible hypothesis could be that two pools of SCX-positive cells co-exist within the chick limb buds, one pool being sensitive to the TGFβ2/SMAD2/3 signalling pathway and another one being sensitive to the FGF ERK/MAPK signalling pathway.

Immobilisation following muscle paralysis induces a drastic diminution of SCX, TNMD and THBS2 gene expression in stylopod/zeugopod tendons of chick limbs (Fig. 4). This shows that tendon gene expression is sensitive to mechanical signals in chick limbs. However, in the mdg mouse, which is deprived of muscle activity, Scx/GFP-positive tendons are observed in stylopod/zeugopod limb regions, although they are reduced in size (Huang et al., 2015b). This difference could be due to the possibility that the GFP fluorescence can be still detected even if the Scx promoter is no longer active or the fact that the mouse embryos are still submitted to mechanical movements from maternal activity, whereas the pharmacologically induced immobilisation used in our experiments is more drastic. However, an alternative and plausible explanation is that these results indicate that tendon development in chick and mouse has differential requirements for mechanical movements. It has been demonstrated in mice that muscles are required for zeugopod tendon elongation, but only tendon size and individuation depend on mechanical forces in mouse limbs (Huang et al., 2015b). The complete loss of zeugopod tendons in chick immobilised embryos shows that mechanical signals are crucial for all the steps of chick zeugopod tendon differentiation. This can be correlated with the fact that tendon cells experience higher levels of mechanical signals in actively moving chick embryos in the egg compared with mouse embryos embedded in uterine membranes and with the faster development of the musculoskeletal system in chick versus mouse embryos.

Although differences exist between the mechanical signal requirement between chick and mouse tendon development, mechanical forces generated by muscle contraction are recognised as being required for skeletal system formation during chick and mouse development (reviewed by Shwartz et al., 2013). Immobilisation affects bone, cartilage and synovial joint morphogenesis, targeting general processes such as proliferation and differentiation leading to shape and size defects (Blitz et al., 2009; Kahn et al., 2009; Roddy et al., 2011a). A correlation has been established between biophysical stimuli patterns and skeletal regions affected upon immobilisation (Roddy et al., 2011b). Tendons that link muscles to bones are expected to experience high mechanical strains and are largely affected in immobilisation conditions. The mechanosensitivity of tendon development is maintained in adult life, as tendon cells are sensitive to mechanical signals generated by tendon loading (reviewed by Nourissat et al., 2015). Scx expression is downregulated in adult tendons in unloading conditions (Maeda et al., 2011), whereas Scx and Tnmd are activated under overloading conditions in mice (Mendias et al., 2012; Zhang and Wang, 2013). We conclude that tendon cells require appropriate mechanical signals during development and adult life.

In immobilised chick embryos, transcriptional readouts of both FGF/ERK MAPK and TGFβ/SMAD2/3 signalling pathways are downregulated in limb tendons. This shows that both pathways are sensitive to mechanical signals in chick limbs. These pathways are known to be sensitive to mechanical forces in other biological systems (Humphrey et al., 2014). Moreover, transcriptome profiling analyses have identified FGF and TGFβ signalling as being downregulated in developing humerus (limb bone) of immobilised mouse foetuses (Rolfe et al., 2014). In addition to being downregulated in tendons of immobilised embryos, both FGF4 and TGFβ2 prevent the decrease of SCX expression in chick limbs in immobilisation conditions. Rescue experiments with FGF4 or TGFβ2 in immobilised limbs do not activate TGFβ or FGF transcriptional readout, respectively, indicating that FGF4 and TGFβ2 act independently of each other to activate SCX expression during limb tendon differentiation. The ability of TGFβ2 to rescue SCX expression in chick limbs in immobilised embryos is reminiscent of the requirement of the SMAD2/3 pathway for Scx induction downstream of mechanical forces in tendon cells (Maeda et al., 2011). However, TNMD and THBS2 downregulation was only prevented by TGFβ2 and not by FGF4. We hypothesise that TGFβ2 activates another signal required for TNMD and THBS2 expression, which is not regulated by FGF4, in immobilisation conditions. In the presence of muscle contraction, FGF4 activates TNMD and THBS2, whereas in the absence of muscle contraction FGF4 is not able to activate TNMD or THBS2 expression, highlighting the independent effects of both pathways in tendon differentiation.

In summary, both FGF4 and TGFβ2 signalling molecules are involved in the commitment of undifferentiated chick limb mesodermal cells towards the tendon lineage and act downstream of mechanical forces to regulate tendon differentiation during chick limb development (Fig. 8). Both FGF4 and TGFβ2 have a tenogenic effect, independently of each other, during both muscle-independent and -dependent phases of chick limb tendon development.

Fertilised chick eggs (JA 57 strain) (EARL Morizeau, Dangers, France) were incubated at 38°C. Embryos were aged according to the number of days of incubation (embryonic day) or staged according to Hamburger and Hamilton (HH) stages (Hamburger and Hamilton, 1992). Swiss mouse embryos (Janvier Labs) were collected after natural overnight matings. For staging, fertilisation was considered to take place at midnight. The manipulation of chick and non-transgenic mouse embryos was performed in accordance with the guidelines of the French National Ethics Committee.

FGF4, TGFβ2, TGFβ2+PD18, FGF4+SIS3 beads or mFgf4/RCAS-expressing cells were grafted in limbs of E3/E4 chick embryos as described (Edom-Vovard et al., 2002). Embryos were harvested 4, 6 or 24 h after grafting. Further details are provided in supplementary Materials and Methods.

Chick limb explants were prepared as described by Placzek and Dale (1999) and treated with TGFβ2, FGF4, PD18, SB43, SIS3 or TGFβ2+PD18 as described by Havis et al. (2014). Further details are provided in supplementary Materials and Methods.

Decamethonium bromide (DMB) and pancuronium bromide (PB) were prepared in Hank's solution, at final concentrations of 12 mM. DMB or PB solution (100 µl) was injected daily using a Pipetman pipette (Gilson) into the amniotic fluid next to the embryos after vitelline membrane removal in E4.5, E5.5 and E6.5 chick embryos. Control embryos were injected with Hank's solution using the same daily protocol. Immobilised or control embryos were analysed at E5.5 (24 h), E6.5 (48 h) or E7.5 (72 h). Forelimbs or hindlimbs were isolated and analysed by in situ hybridisation on sections or wholemounts or by RT-q-PCR analysis. For RT-q-PCR analysis, RNAs were prepared from the whole limbs, the limbs without digits or the digits alone.

RT-q-PCR of experimental or control chick limbs, chick limb explants or mouse limb explants were performed as previously described (Havis et al., 2014). A detailed protocol is provided in supplementary Materials and Methods.

Control or manipulated chick limbs were fixed and processed for in situ hybridisation as previously described by Havis et al. (2014). The probes that were used are described in supplementary Materials and Methods. Differentiated muscle cells were detected after in situ hybridisation with the monoclonal antibody MF20 (non-diluted supernatant) developed by D. A. Fischman and obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa.

We thank laboratory members for comments on the manuscript and Sophie Gournet for illustrations.