Mechanical force regulates tendon extracellular matrix organization and tenocyte morphogenesis through TGFbeta signaling

Mechanical forces between cells and extracellular matrix (ECM) influence cell shape and function. Tendons are ECM-rich tissues connecting muscles with bones that bear extreme tensional force. Analysis of transgenic zebrafish expressing mCherry driven by the tendon determinant scleraxis reveals that tendon fibroblasts (tenocytes) extend arrays of microtubule-rich projections at the onset of muscle contraction. In the trunk, these form a dense curtain along the myotendinous junctions at somite boundaries, perpendicular to myofibers, suggesting a role as force sensors to control ECM production and tendon strength. Paralysis or destabilization of microtubules reduces projection length and surrounding ECM, both of which are rescued by muscle stimulation. Paralysis also reduces SMAD3 phosphorylation in tenocytes and chemical inhibition of TGFβ signaling shortens tenocyte projections. These results suggest that TGFβ, released in response to force, acts on tenocytes to alter their morphology and ECM production, revealing a feedback mechanism by which tendons adapt to tension.


Introduction
Cells in all multicellular organisms are exposed to mechanical forces through adhesions to neighboring cells and to the extracellular matrix (ECM), as well as the ebb and flow of the environment. Force has been shown to influence cellular processes such as cell division, survival, migration, and differentiation (Behrndt et al., 2012;Culver and Dickinson, 2010;Hamada, 2015;Keller et al., 2008;Roman and Pekkan, 2012). Cellular responses to force include the activation of cell surface receptors such as integrins (Itgs), G-protein-coupled receptors (GPCRs), transient receptor potential (TRP) ion channels, and Piezo channels (Busch et al., 2017;Chachisvilis et al., 2006;Maartens and Brown, 2015;Mederos y Schnitzler et al., 2008;Popov et al., 2015;Wu et al., 2017). Despite recent insights into the nature of such responses, few in vivo studies have investigated how cells adapt to force and alter the ECM landscape to strengthen or weaken it accordingly (Maeda et al., 2011;Ng et al., 2014).
The musculoskeletal system bears among the strongest forces experienced by any tissue, such as the tensional forces exerted upon tendons and ligaments (Heinemeier et al., 2013;Wang, 2006). Tendons can withstand such forces due to the specialized organization of collagen (Col) fibers and proteoglycans within each tendon fibril. Tendon injuries are extremely common and debilitating, especially in athletes, the elderly, and patients with neuromuscular diseases such as muscular dystrophy (Bö nnemann, 2011;Walden et al., 2017). Despite their prevalence, little is known about how tendon fibroblasts (tenocytes) respond in vivo to tensional force at muscle attachments, or how they adapt to changes in mechanical load. Tendons form a variety of attachment sites -connecting muscles to cartilages and bones as well as other muscles and soft tissues. A myotendinous junction (MTJ) is a specialized ECM-rich region at the interface of muscle-tendon attachment sites that functions as the primary sources of force transmission. Each type of attachment bears varying levels of force, which correlates with distinct composition and organization of its tendon ECM (Ker et al., 2000;Wang, 2006). While extensive research has been conducted to evaluate the effects of exercise on size and strength of muscle fibers, less is known about how it effects tendon morphology and function. Understanding this is key to gaining insights into the causes of tendon defects and developing new treatments for tendon injuries or atrophy.
Previous studies in vitro have suggested that tenocytes actively respond to changes in force in their environment by modulating ECM composition and organization (Maeda et al., 2010;Rullman et al., 2009). Excised tendons stretched in collagen gels, as well as tissue samples from chronic Achilles tendonitis patients, upregulate various collagens and ECM-modulating proteins, particularly Col3, Matrix Metalloproteinase 9 (MMP9) and MMP13 (Ireland et al., 2001;Pingel et al., 2014). In addition, collagen fibril size decreases and fibril packing increases in tendinopathies, likely due to increased ECM turnover (Pingel et al., 2014). These studies have suggested ECM modifications and morphological changes in tendinopathies but they have largely been limited to cultured tendons or tendon fragments. In vivo, several growth factor signaling pathways and transcription factors have been implicated downstream of mechanical force in tendon development and repair in mice. These include several members of the Transforming Growth Factor (TGF) superfamily, including TGFb and Bone Morphogenetic Proteins (BMPs), as well as Fibroblast Growth Factors (FGF) (Gumucio et al., 2015;Nourissat et al., 2015). Mice lacking the transcription factor Scleraxis (Scx) show severe defects in force-transmitting and load-bearing tendons, suggesting that Scx is essential for maintenance of tendon ECM in response to mechanical force . In addition, Scx directly regulates transcription of tendon ECM components, including Col1a1 (Havis et al., 2014;Subramanian and Schilling, 2015). Our studies of the ECM protein Thrombospondin 4b eLife digest Tendons -the fibrous structures that attach muscles to bones -must withstand some of the strongest forces in the body. Little is known about how tendons develop or adapt to withstand these forces. Studies have shown that muscles respond actively to force, as seen during exercise. Do tendons respond in similar ways?
Tendons consist of collagen fibers surrounded by a 'matrix' of proteins. Also embedded in the matrix are specialized cells called tenocytes, which regulate the production of the different components of the tendon. A genetic modification allows tenocytes to be tracked using a fluorescent gene product that can be viewed using a microscope. Subramanian et al. have now used this technique in zebrafish to watch how the behaviors of the tenocytes change in response to forces applied to the tendon. Subramanian et al. show that at the start of muscle contraction, tenocytes put forth long projections from their cell bodies that extend perpendicular to the muscle fibers. This suggests that the projections act as force sensors. Consistent with this idea, paralyzing the muscle causes the projections to shrink. This shrinkage correlates with changes in how the tendon matrix proteins are organized.
Further investigation reveals a force-responsive signaling pathway in the tenocytes that controls how these cells grow and produce key tendon matrix proteins. Subramanian et al. believe this pathway is central to how tendons adapt to the forces applied during muscle contraction.
A better knowledge of how force affects tendon structure could ultimately help to improve treatments for tendon injuries and tendon atrophy. In particular, understanding how force affects how tenocytes develop could help researchers to develop new ways to regenerate and repair tendons.
(Tsp4b) in zebrafish have shown that it is an essential scaffolding protein for tendon ECM assembly, required to maintain muscle attachments subjected to mechanical force via muscle contraction, and able to strengthen attachments when overexpressed (Subramanian and Schilling, 2014).
Here, we show that mechanical force causes remarkable morphological changes in tenocytes in zebrafish, which form a dense curtain of projections at MTJs, and in their surrounding ECM. Tenocyte projections have been reported in electron micrographs of mammalian tendon fascicles yet how they form and their functions in tendon development remain largely unexplored (Kalson et al., 2015;Knudsen et al., 2015). Our results suggest that tenocytes play an active role in sensing force and thereby regulating ECM composition and overall tendon strength. In addition, we show that the force of muscle contraction regulates the growth and branching of tenocyte projections via TGFb signaling. Such feedback between tenocytes and ECM may be a common mechanism for force adaptation within the musculoskeletal system.

Results
Tenocytes elongate with the onset of muscle contraction Tenocytes in zebrafish express two Scx orthologues, scxa and scxb (Chen and Galloway, 2014). Using a bacterial artificial chromosome (BAC) transgenic line that expresses mCherry under the control of regulatory elements for scxa, Tg(scxa:mCherry), we examined the morphogenesis of tenocytes during embryonic (20 hr post fertilization (hpf) to 72 hpf) and early larval (72 hpf to 5 dpf) zebrafish development. Expression of scxa:mCherry was first detected at 20 hpf in muscle and tendon progenitors of the somites. In a developing zebrafish embryo, muscles in the trunk establish attachments at bilateral, 'chevron' shaped somite boundaries that subdivide each muscle segment forming the vertical myoseptum (VMS). In addition, dorsal and ventral compartments within each somite are subdivided by a horizontal myoseptum (HMS), which extends laterally from the notochord (NC), along which oblique myofibers attach. By 24 hpf, as the first myofibers differentiated, scxa:mCherry expression in muscle progenitors diminished and became progressively restricted to scattered tendon progenitors along the HMS and VMS (~24 cells per VMS) ( Figure 1A, D,G) ( Figure 1-video 1). Cells with the highest levels of scxa:mCherry expression were located laterally, adjacent to the HMS, while more medial cells expressed lower levels ( Figure 1A', D', G'). By 36 hpf, scxa:mCherry+ cells doubled in number (~44/VMS) and became increasingly localized to the HMS and VMS at future MTJs ( Figure 1B,E,H) ( Figure 1-video 1). At this stage, cells with the highest scxa:mCherry expression that were located medially and in the ventral somites began to extend projections laterally along the VMS, perpendicular to the orientation of muscle fibers ( Figure 1B', E', H'). By 48 hpf these projections extended 70-80 mm ( Figure 1C,F,I). 3D-reconstructions of confocal stacks at 60 hpf revealed that this polarized network of tenocyte projections covered the entire VMS ( Figure 1C', F', I'; Figure 1-figure supplement 1). Time-lapsed videos of Tg(scxa:mCherry) embryos capturing images at 20 min intervals from 48 to 60 hpf showed that tenocyte projections are dynamic and constantly changing in length and branching pattern (Figure 1-video 2). Tenocytes along the HMS near the NC have shorter, more convoluted projections than tenocytes along the VMS ( Figure 1I'; Figure 1-figure supplement 1B). Thus, during the period in which axial muscles in the trunk begin to contract and embryos become motile, tenocytes align along future MTJs and undergo dramatic changes in cell shape that correlate with the establishment and strengthening of muscle attachments.
Cranial tendons also undergo dramatic morphological changes during the onset of muscle attachment and contractility. A cluster of scxa:mCherry+ tenocyte progenitors is first observed at 36 hpf in the ventral midline near the future attachment sites of the sternohyoideus (SH) and adductor mandibulae (AM), which are among the earliest muscles to differentiate at 53 hpf (Schilling and Kimmel, 1997). By 48 hpf, three major clusters of scxa:mCherry+ tenocytes are visible ventrally, one anterior that forms ventral mandibular, hyoid, and oculomotor muscle tendons and two posterior clusters associated with each SH (

Tenocyte elongation requires muscle contraction
Based on the close correlation between the onset of muscle contraction and tenocyte morphogenesis, we hypothesized that mechanical force serves as a cue for tenocytes to elongate and form projections. To test this idea, we first injected full-length mRNA encoding codon-optimized a-bungarotoxin (aBtx), a specific irreversible antagonist of acetylcholine receptors that blocks neuromuscular synapses and prevents skeletal muscle contractions (Swinburne et al., 2015;Westerfield et al., 1990). Embryos injected with aBtx mRNA at the one-cell stage were completely paralyzed until 60 hpf, after which they gradually recovered motility as aBtx activity declined. Depthcoded, 3D-reconstructed images of living trunk tenocytes along the MTJs of somites 16-17 at 48 hpf revealed an average reduction of 13 mm (18%) in axial tenocyte projection length in aBtxinjected embryos compared to uninjected controls (Figure 2A,B,E). Paralyzed embryos also showed reduced branching complexity in their projections ( Figure 2F) and projection density along the VMS (Figure 2-figure supplement 1). To restore mechanical force, we electrically stimulated aBtxinjected embryos to induce muscle contractions, as described previously (Subramanian and Schilling, 2014). Stimulation at 48 hpf for 2 min at 20V caused no visible muscle damage or significant change in tenocyte projection lengths compared to controls ( Figure 2C,E) while the same stimulation of aBtx-injected embryos rescued both tenocyte projection length and density along the VMS almost completely ( Figure 2D,E; Figure 2-figure supplement 1). The observed reductions in projection length and density were caused by paralysis rather than any unanticipated effect of aBtx, since homozygous mutants paralyzed due to lack of a functional voltage-dependent L-type calcium channel subtype beta-1 (Cacnb1), necessary for excitation-contraction coupling in muscle, showed similar (10-15 mm) reductions in projection length (Figure 2-figure supplement 2) (Zhou et al., 2006). Tenocytes in cacnb1 mutant embryos fail to compact and elongate. Since aBtx-injected embryos recover from paralysis at 65 hpf, prior to cranial muscle contractions, we compared cranial tenocyte patterning in immunostained 4 dpf cacnb1 mutant embryos with their siblings. We observed both a failure of cranial tenocytes to compact and elongate, as well as reduced projections and frayed myofibers ( Figure 2-figure supplement 3). These results indicate a strong correlation between mechanical force from muscle contraction and tenocyte morphogenesis, suggesting that force stimulates the dynamic growth and branching of tenocyte projections.

ECM organization at MTJs requires muscle contraction
We previously showed that Tsp4b secreted by tenocytes is essential for ECM organization at MTJs and strengthens muscle attachments (Subramanian and Schilling, 2014). We hypothesized that force stimulates tenocytes to secrete Tsp4b from the projections they extend into the tendon ECM. Consistent with this, injection of tsp4b-gfp full length mRNA into Tg(scxa:mCherry) embryos produced Tsp4b-GFP protein that localized to MTJs along the attachment sites at 48 hpf ( Figure 3A-C, I). This exogenous Tsp4b-GFP protein was dramatically reduced in aBtx-injected embryos, particularly around projections, and became diffuse compared to uninjected controls ( Figure 3D-F,I). Likewise, immunohistochemical staining for Tsp4b at 48 hpf in aBtx-injected embryos showed dramatic reductions along the attachment sites compared to controls ( Figure 3G,H,J). In contrast, other ECM proteins such as laminin (Lam) at 48 hpf and and fibronectin (Fn) at 24 hpf showed no significant changes at the MTJ in aBtx-injected embryos at 48 hpf (Figure 3-figure supplement 1). Defects in Tsp4b distribution were due to the lack of mechanical force, since restoring force in paralyzed embryos through electrical stimulation rescued both local levels and the overall area of Tsp4b protein localization along the VMS (Figure 3-figure supplement 2). To test the hypothesis that changes in Tsp4b localization were due to reduced tsp4b gene expression in response to lack of force, we performed real-time PCR and found a significant reduction in tsp4b expression at 48 hpf in aBtx-injected embryos, while no significant change in expression was observed at 24 hpf (Figure 3figure supplement 3). These results suggest a role for mechanical force in both assembly of tendon ECM and expression of key MTJ ECM genes during development and demonstrate that muscle contractions regulate the composition and organization of the tendon ECM.

Microtubules maintain tenocyte projections and their interactions with tendon ECM
Cellular projections in neurons, keratinocytes and pigment cells are rich in microtubules (MTs) and in some cases F-actin, while filopodial extensions of cells are typically more actin-based (Eom et al., 2015;Witte et al., 2008). To determine the cytoskeletal structure of tenocyte projections we injected full-length mRNA encoding eGFP-atubulin and found that this fusion protein localized to    (Rusan et al., 2001). Similar injections of plasmids encoding EGFP-Lifeact-7 failed to show labeled actin in the projections. To determine if MTs are critical for maintaining projections, we treated embryos with Nocodazole, which caused them to fragment ( Figure 4D,E). Immunohistochemical staining of Nocodazole-treated embryos for Tsp4b showed scattered Tsp4b + puncta localized at MTJs along the VMS and reduced Tsp4b protein levels in the VMS ( Figure 4F,G,H,I). These results suggest that MTs are the key structural components of tenocyte projections required to sustain the organization of tendon ECM.
TGFb signaling is required for tenocytes to extend projections in response to force Previous studies from primary cultures of tenocytes and stretch tests on isolated tendons in vitro have proposed a mechanoresponsive role for TGFb signaling (Gumucio et al., 2015;Havis et al., 2016;Maeda et al., 2011). TGFb secreted by muscles or latent in the ECM of the MTJ could be released in response to force and thereby regulate both tenocyte morphogenesis and ECM production. To address this hypothesis, we treated Tg(scxa:mCherry) embryos with a chemical inhibitor of TGFb signaling (SB431542 -which blocks TGFb receptors) for 12 hr from 24 to 36 hpf (Chen and Galloway, 2014). This treatment severely reduced signaling in both muscle fibers and tenocytes as confirmed by immunostaining for phosphorylated SMAD3 (pSMAD3) in SB431542-treated embryos compared to controls ( Figure 5A-C,E-G,I). In addition, tenocyte projections were reduced in length by an average of~20 mm in SB431542-treated embryos ( Figure 5D,H,J), similar to the effects of aBtx ( Figure 2B,E). However, unlike embryos injected with aBtx, SB431542-treated embryos continued to swim actively. These results suggest that TGFbsignaling acts downstream of muscle contraction to stimulate growth and branching of tenocyte projections. To confirm if muscle contraction is essential for activation of TGFb signaling, we stained control and aBtx-injected, Tg(scx:mCherry) embryos with anti-pSMAD3. While control embryos showed strong pSMAD3 localization in the nuclei of muscles and tenocytes, pSMAD3 staining was strongly reduced in the nuclei of tenocytes in aBtx-injected embryos ( Figure 6A-G). Here, in contrast to embryos treated with SB431542 ( Figure 5C,G), paralysis specifically reduced pSMAD3 in tenocytes and not in muscle nuclei. This correlated with the reduction in length of tenocyte projections ( Figure 6H). These results suggest that mechanical force from muscle contraction serves as a cue for TGFb mediated signaling in tenocytes to control their morphogenesis and differentiation.
To further confirm that mechanical force has a role in induction of TGF-b responses in tenocytes, we stained control and aBtx-injected embryos with or without electrical stimulation, with anti-pSMAD3 antibody to verify if localization of pSMAD3 in nuclei of tenocytes could be rescued. aBtxinjected embryos stimulated with mild electric current showed increased pSMAD3 localization in tenocyte nuclei strongly suggesting that mechanical force from muscle contraction can rescue TGFb signaling in tenocytes ( Figure 6-figure supplement 1).

Tenocyte projections regulate force-dependent gene expression
Previous studies have linked mechanical force with the expression of tenogenic and myogenic genes (Chen et al., 2012;Maeda et al., 2011). Our results showing similar tenocyte projection defects in Nocodazole-treated, aBtx-injected and SB431542 treated embryos suggest that they induce similar changes in expression of force-responsive genes. Real-time PCR analysis on cDNA prepared from 48 hpf control and aBtx-injected embryos revealed that paralysis led to an almost complete loss of expression of tsp4b, as well as TGFb-induced protein (tgfbip) and, connective tissue growth factor a (ctgfa), while expression levels of other tendon genes, such as scxa, were unaffected ( Figure 6-figure supplement 2A). All three genes (tsp4b, tgfbip and ctgfa) were restored to control levels of expression with electrical stimulation (Figure 6-figure supplement 2A). We further validated the results with digital droplet PCR (ddPCR) on cDNA prepared from FACS sorted tenocytes and whole  . Loss of tenocyte projections through destabilization of microtubules in nocodazole-treated embryos also led to reduced expression of tsp4b and tgfbip while expression of ctgfa and scxa were elevated ( Figure 6-figure supplement 3B). Taken together, these results are consistent with the hypothesis that mechanical force acts through TGFb signaling to regulate tenocyte-specific transcription including ECM components such as Tsp4b.  (H) Dot plot shows individual tenocyte projection lengths (p value was calculated by t-test **<0.00005). Representative muscle nuclei are marked by a blue continuous ROI. Scale bar = 10 microns. The measurements used for quantitative analysis and creation of the plots can be accessed from

Discussion
Mechanical forces generated by cells adhering to ECM alter their shapes and functions during development, but few studies have investigated the underlying mechanisms in vivo (Dan et al., 2015;Hamada, 2015;Ladoux et al., 2015). Here we show that early developing tenocytes in zebrafish express the tenogenic fate determinant, scxa, prior to the differentiation of muscle fibers and respond to the onset of muscle contraction by elongating and extending an array of polarized projections. These projections are disrupted by changes in force as is the corresponding organization of the tendon ECM, which is critical for MTJ function (Subramanian and Schilling, 2014). Our results show for the first time in vivo that TGFb signaling responses induced by mechanical force from muscle contraction correlate with changes in tenocyte morphogenesis and tendon ECM composition during tendon development. These results suggest a novel role for tenocyte projections as force sensors and responders in the feedback between tenocyte and ECM that physically balance responses to mechanical force (Figure 7).

Roles for tensional force in tendon morphogenesis
Tendons primarily experience tension from muscle contractions (Lavagnino et al., 2015;Wang, 2006). In contrast, skeletal cell types (e.g. osteocytes, osteoblasts, chondrocytes) are exposed to compressive forces (Klein-Nulend et al., 2012) or shear forces in the case of chondrocytes in joints exposed to fluid flow (Servin-Vences et al., 2017).We show that in the absence of tension during development, tenocytes reduce the extent and spread of their projections into the tendon ECM and this can be rescued by a short bout of contraction. Tendon defects and injuries result from dramatic changes in tension experienced either instantly or periodically over extended periods of muscle disuse or overuse (Franchi et al., 2013;Wang et al., 2012). Embryonic tenocyte progenitors experience muscle contractions at early stages and must continuously adapt to changes in muscle strength. Our results support the idea that the establishment and adaptation of MTJs occurs in response to mechanical force from muscle contraction and involves both changes in tenocyte morphogenesis and ECM production.
We show that paralysis reduces tenocyte branching and tendon ECM, which can be rescued by restoring muscle contractions through electrical stimulation. Early experiments on developing chick embryos have shown that induced lack of muscle activity (either by lack of neuronal innervation or by injecting paralysis-inducing drugs) negatively affects the growth of associated skeletal structures, suggesting a role for force from muscle contraction as an essential cue for proper growth and differentiation of the skeleton (Hall and Herring, 1990;Hamburger and Waugh, 1940). During development, the skeleton is exposed to two major types of force -contractile (tension) force from muscles and compressional force (e.g. gravity). A contractile force from muscles has a greater impact on the growth of bones when compared to compression, indicating a primary role for muscle function in guiding the growth of associated skeletal tissues (Ellman et al., 2014;Warden et al., 2013). Recent studies in paralyzed limbs have shown that the development of a tendon-bone attachment unit, the enthesis, is affected by lack of muscle contraction (Schwartz et al., 2013;Tatara et al., 2014). Our studies suggest that muscle contraction has a similar role in the development of tendons. Immobilization experiments performed on canine models have shown that mechanical force is required for repair of tendon injuries (Gelberman et al., 1982). More recent studies using paralysis and restricted movement have shown that mechanical force has multiple roles in the maintenance of tenogenic gene expression, secretion of tendon ECM, and tenocyte survival Hettrich et al., 2011;Maeda et al., 2011).

Microtubules are essential for tenocyte projection stability and function
Cellular filopodia and neuronal axons require either F-actin and MTs or both in the formation and maintenance of projections, and new classes of cellular projections are emerging from recent studies such as cytonemes and airinemes (Bornschlö gl, 2013;Eom et al., 2015;Huang and Kornberg, 2015;Witte et al., 2008). MTs also serve as pathways for trafficking various proteins, RNA, and other intracellular components along projections. Tenocytes in Drosophila rely on a network of polarized MTs for the maintenance of cellular structure and function (Subramanian et al., 2003), but similar requirements for cytoskeletal components have not been investigated in vertebrate tendons. Here, we show that zebrafish tenocytes are rich in MTs, which are required to maintain projections. Pharmacological disruption of MTs destabilized the projections without affecting tenocyte cell bodies. This reduced Tsp4b localization suggesting that tenocyte projections both sense force and respond to it by altering ECM organization in an MT-dependent manner. A caveat to this result is that treatment of embryos with Nocodazole causes global destabilization of MT in the entire embryo. Hence, the effects on tenocyte projections and tendon ECM organization could also arise in response to MT destabilization in neighboring muscle fibers, axons and other cells. Similar roles for cellular projections have been observed in pigment cells, where airinemes composed of both F-actin and MTs play a role in longrange signaling by secreting signaling ligands at the tips of their projections (Eom et al., 2015). We find that loss of tenocyte projections leads to upregulated expression of scxa and ctgfa in MT-deficient embryos, suggesting that they revert to a more dedifferentiated state.

Tenocyte projections are force sensors in the tendon ECM
Previous EM studies of human and rat tendons have described tenocytes projecting into the tendon matrix, but their functional significance has remained unclear (McNeilly et al., 1996;Pingel et al., 2014). 3D reconstructions from EM studies suggest a role for these projections, referred to as 'fibropositors' in one study, in secreting collagen fibrils (Canty et al., 2004). Analysis of Scx expression in chick and mouse using immunostaining and transgenic reporter lines, respectively, have shown that limb tendons elongate as the musculoskeletal system matures (Brent et al., 2003;Kardon, 1998;Pryce et al., 2007). Our results in zebrafish reveal that such elongated projections are conserved, but quite distinct in different classes of tenocytes. While cranial tenocytes resemble those in the limb in that they extend in parallel to the direction of force, axial tenocytes extend their projections perpendicular to the plane of muscle contraction (with opposing directions of contractile force) (Figure 1 and Figure 1-figure supplement 2). Based on our results, we propose that these distinct morphologies reflect a more structural, load-bearing role for cranial (and limb) tenocytes, while early larval axial tenocytes in zebrafish function as tension sensors in the myoseptum. Many important questions remain and form the basis of future studies, including why these cells are so polarized and how this mediolateral polarity develops. Consistent with the tension-sensor hypothesis, the timing of the outgrowth of tenocyte projections tightly correlates with the onset of muscle contraction. Tension sensing projections are observed in other musculoskeletal tissues as osteocytes extend projections into the bone matrix where they are thought to form a network of force sensors (Cowin et al., 1991) that modulate bone formation and resorption (Klein-Nulend et al., 2012;Schaffler et al., 2014). Likewise, the cues that cause osteoblasts to form these projections as they differentiate into osteocytes remain unknown (Franz-Odendaal et al., 2006). Similar to our results with zebrafish tenocytes, mammalian osteocyte projections increase in density in response to force, consistent with a role as force sensors and responders in both cases.
In both bone and tendon, the ECM undergoes dynamic changes in expression of collagens, fibronectin, laminin and MMPs, and this is also the case in the developing somites of zebrafish embryos (Jenkins et al., 2016;Snow and Henry, 2009). Tenocyte projection formation also correlates with the establishment of tendon ECM. Previous studies have shown remodeling of MTJ ECM between 24 and 48 hpf with a progressive reduction of Fn, which is replaced by increased in levels of Lam at the MTJ (Jenkins et al., 2016). Our results suggest that initial production and accumulation of Fn is independent of force at 24 hpf, when tsp4b also shows force-independent expression. The later forcedependent expression and localization of Tsp4b at 48 hpf indicates that dynamic regulation of tendon ECM occurs after the onset of muscle contraction, which suggests a role for mechanical force in the process. Mammalian tenocytes actively sense mechanical force in vitro, resulting in changes in gene expression, cytoskeletal organization and ECM secretion (Banos et al., 2008;Havis et al., 2016;Maeda et al., 2010;Maeda et al., 2013;Maeda et al., 2011). This depends, at least in part, on gap junctional complexes that localize to tenocyte projections (Maeda et al., 2012). Exercise induces Tenomodulin (Tnmd) and Col1a1 expression and tenocyte proliferation in rats (Eliasson et al., 2009;Zhang and Wang, 2013) and stress induces COL4A1 and COL6A1 expression in chick tenocytes (Marturano et al., 2014). Despite these changes in gene expression, the molecular mechanisms underlying these cellular signaling responses to force are unclear. Embryonic tenocyte projections in zebrafish end in bouton-like structures close to the dermis (Figure 1-figure supplement 1A,B), which may act as signaling beacons and ECM secreting centers.
The strong correlation between onset of muscle function, changing myotendinous ECM and tenocyte morphogenesis suggests a model in which force is transduced through cues from the ECM that induce the formation of projections ( Figure 7A). Similar processes may underlie the projections of osteocytes and other mesenchymal cell types. Such feedback likely allows tendons to adapt to changing mechanical force during normal development and exercise, as well as in healing and repair of tendon injuries.

Mechanical forces and signaling
Our results show for the first time that activation of TGFb signaling in response to mechanical force is required for tenocyte morphogenesis, in particular the growth and branching of tenocyte projections. Paralyzed embryos (aBTX-injected) lose pSMAD3 expression in tenocytes and projections shorten, which is rescued by restoring muscle contraction. Similarly, pharmacological inhibition of TGFb receptors reduces pSMAD3 expression and shortens tenocyte projections. Studies of mechanotransduction have identified several putative signaling pathways involved, depending on the tissue, including TGFb, YAP/TAZ, and Integrins, as well as membrane channels such as TrpV4 and Piezo receptors (Busch et al., 2017;Gumbiner and Kim, 2014;Lavagnino et al., 2015;Servin-Vences et al., 2017). Some of these pathways such as TGFb and YAP/TAZ share intermediate signaling components and targets, which complicates our understanding of their role in mechanotransduction in specific tissues (Qin et al., 2018;Szeto et al., 2016). This could help explain the modest reduction in expression of tgfbip and ctgfa2 in SB431542-treated embryos, as other mechanotransduction signaling pathways may still function in these embryos to partially maintain expression levels of these genes ( Figure 6-figure supplement 3) In vitro studies of tenocyte primary cultures and excised tendon tissue have shown elevated TGFb signaling in response to mechanical load (Heinemeier et al., 2003;Heinemeier et al., 2007;Maeda et al., 2013;Maeda et al., 2011;Yang et al., 2004) and mice show elevated TGFb signaling in muscles and tendons following exercise (Maeda et al., 2011). These studies suggest that TGFb signaling, in addition to its earlier role in tenocyte specification (Havis et al., 2016), is involved in mechanotransduction in these cells after they differentiate. TGFb signaling is activated by many factors, including integrins, BMP1 and MMPs which can act on the large latent complex (LLC), to release active TGFb ligand from the ECM (Horiguchi et al., 2012;Keski-Oja et al., 2004;Todorovic et al., 2005). This could be the critical cue from the ECM that induces and modulates the formation of tenocyte projections. One candidate for initiating these events is Tsp4, since Tsps can activate TGFb signaling by destabilizing latencyassociated peptide (LAP) (Bailey Dubose et al., 2012). The tendon matrix is rich in MMPs and Tsps, which dynamically change in composition and activity depending on mechanical force (Jenkins et al., 2016;Popov et al., 2015;Subramanian and Schilling, 2014). The dynamic reductions in Tsp4b that we have shown in response to paralysis could fail to activate latent TGFb in the tendon ECM. Furthermore, because we see reductions in Tsp4b expression in paralyzed embryos, our results support a model where force triggers TGFb signaling leading to increased expression of Tsp4b, which in turn activates TGFb expression, creating a positive feedback loop ( Figure 7A,B). Transection of tendons or injection of botulinum toxin (Botox) to induce paralysis in mice causes tenocyte death and reduced expression of tenogenic genes (Maeda et al., 2011). In contrast, we observe neither cell death nor significant changes in tenogenic gene expression in paralyzed (cacnb1 mutant) zebrafish embryos until 5 dpf, several days after tenocyte differentiation. We interpret such a response as a separate response to prolonged disuse rather than an adaptation to force.
These studies have shown a direct relationship between mechanical force and tendon development through TGFb signaling in tenocytes. How do tenocyte progenitors begin the process of elongation and growth of projections? What are the roles of these projections during tendon embryonic development and in adult tendons? Osteocytes are known to induce repair pathways in bone when cracks or stress damage their processes (Dooley et al., 2014;Mulcahy et al., 2011). Do tenocyte projections perform a similar role in tendon repair? These are some of the questions that need to be addressed in the field of tendon biology. Understanding the relationship between force and tendon development is essential for developing effective treatment strategies that include engineering tendons to treat tendon injuries. Force sensing projections that allow cells to adjust their surrounding ECM, such as those we have described in tenocytes, may also be a more general feature of cells, particularly within the musculoskeletal system.

Zebrafish transgenics and mutants
Tg(scx:mCherry) transgenics were generated by injecting a BAC construct (CH211-251g8) containing mCherry ORF inserted in frame after the start codon of the scxa gene (McGurk et al., 2017). A new mutant allele of cacnb1 was identified in a forward genetic screen and outcrossed with Tg(scxa: mCherry) to create a cacnb1;Tg(scxa:mCherry) line. All embryos were raised in embryo medium at 28.5˚C (Westerfield, 2007), and staged as described previously (Kimmel et al., 1995). Craniofacial muscles and cartilages were labeled as described previously (Schilling and Kimmel, 1997). Adult fish and embryos were collected and processed in accordance with approved UCI-IACUC guidelines.
A stock solution of 33 mM Nocodazole (Sigma M1404, SID: 336851328), an inhibitor of tubulin polymerization, was prepared in DMSO and diluted to a final working concentration of 0.33 mM in embryo medium. Embryos were incubated in 0.33 mM Nocodazole for 12 hr at 28.5˚C. Treated embryos were either mounted for live imaging or fixed for immunostaining.

RT-PCR
Whole embryo RNA was extracted from control and paralyzed embryos collected at 48 hpf according to standard protocols using Trizol (Invitrogen 15596018) and Direct-zol RNA MinipPrep kits (Zymo Research R2061). RNA concentration was normalized between samples and used as a template for cDNA synthesis. cDNA was synthesized with oligodT primers using the standard protocol of ProtoScript II First Strand cDNA Synthesis Kit (NEB E6560). The synthesized cDNA was diluted to 1:20 and used as a template for RT-PCR using the protocol for the Luna Universal qPCR master mix (NEB M3003S). The primers used for RT-PCR are listed in Table 1. The reaction was run on a Light-Cycler 480 II Real time-PCR Instrument (Roche) and analyzed using LightCycler 480 Software. Each qPCR experiment was designed with triplicates of reactions for every biological sample and two biological samples were used for each analysis (Subramanian and Schilling, 2014). ddPCR cDNA was prepared from whole embryo RNA using the standard protocol of ProtoScript II First Strand cDNA Synthesis Kit (NEB E6560). The cDNA concentration was determined following standard protocol and reagents from the Qubit SSDNA assay kit (Invitrogen Q10212) and fluorescence was read on a Qubit 2.0 fluorometer (Invitrogen Q32866). A total concentration of 1 ng was used from each sample to prepare 20 ml of ddPCR reaction following the instructions and reagents from QX200 EvaGreen ddPCR Supermix (Bio-Rad 186-4033).Primers for the PCR are listed in Table 1. The droplets were generated using QX200 Droplet Generation Oil for EvaGreen (Bio-Rad 1864005) on a QX200 Droplet Generator (Bio-Rad 1864002). The PCR reaction was run on a standard thermocycler under standard cycling conditions. Following the PCR the droplets were analyzed using QX200 Droplet Reader (Bio-Rad 1864003). The data were analyzed using QuantaSoft Analysis Pro Software.

Muscle stimulation
Electrical stimulation was used to induce muscle contraction, as previously described (Subramanian and Schilling, 2014). Both aBtx injected and control embryos or larvae were anaesthetized with Tricaine (ethyl 3-aminobenzoate methanesulfonate, Sigma A5040, SID: 329770864), placed on a silicone plate in embryo medium and stimulated for 2 min at 20V, 6 msec duration, 4 pulses/sec frequency and 6 msec delay between successive pulses. With these settings neither control nor paralyzed embryos showed any muscle detachment. Embryos were allowed to recover in embryo medium for 12 hr and further processed for immunostaining or RT-PCR.

Microscopy and image analysis
Embryos processed for fluorescent immunohistochemistry were imaged using a Nikon A1 confocal system with an Nikon Eclipse Ti inverted microscope using a CFI Plan Apochromat VC 60XC (water immersion) objective. Confocal stacks were analyzed using Image J software. The depth-coded 3D reconstructions were created using Nikon software (NIS-Elements AR 4.60.00 64-bit). To better visualize tenocyte projections along the Z-axis, the 3D reconstructed image was rotated to about 45 o . The length of projections was measured using the Neurite Tracer plugin on Image J.

Statistical analysis
Sample size and number of data points required for each experiment were determined using a power analysis calculator (www.powerandsamplesize.com). The embryos were collected from a single tank of fish and processed for injection and downstream stimulation together to minimize variation introduced during handling. Fixation and staining of embryos were also performed together for all samples in a given experiment. Imaging of embryos within each experiment was performed with identical parameters. In order to control for variation in position of tenocyte cell bodies and antibody penetrance variation, projection length, fluorescence intensity of ECM proteins and pSMAD3 were always measured in the ventral half of the VMS in somites 16-19. In experiments where a normal distribution was not present, we analysed the significance using a Wilcoxon Rank Sum test. In datasets involving two samples of unequal variance, a t-test was used. In experiments with more than two experimental conditions, an ANOVA single-factor analysis was performed with posthoc multiple comparisons using Tukey method on R. Data were also quantified and analyzed separately by two of the authors to account for user bias and they obtained similar results. Fluorescence Intensity (FI) to quantify protein localization was measured as described previously (Subramanian and Schilling, 2014).