Mechanical force regulates Sox9 expression at the developing enthesis Open Access

Mechanical force plays crucial roles in development and maintenance of the musculoskeletal system. Connective tissues such as tendons transmit the forces of muscle contraction to skeletal elements. Effective transmission requires transferring a range of physical forces from soft muscle tissue to much stiffer cartilage or bone (Shaw and Benjamin, 2007; Thomopoulos et al., 2010; Zelzer et al., 2014). Specialized areas at the tendon-bone interface called entheses accommodate these physical constraints using a unique fibrocartilaginous extracellular matrix (ECM) that varies in composition along the length of the enthesis (Cury et al., 2016; Rufai et al., 1995; Apostolakos et al., 2014). Osteo-tendinous junction (OTJ) entheses are often sites of tendon-associated injuries such as avulsion fractures, rotator cuff tears and chronic age-related diseases called entheseopathies, which are characterized by increased calcification (Shaw and Benjamin, 2007; Tadros et al., 2018). Treatments of either acute or chronic enthesis injuries are challenging because the ECM is never fully restored after healing. Thus, it is important to understand the mechanisms by which entheses form during development and respond to force to acquire their strength.

Entheses are associated with many different types of tendon and ligament attachments to bone and vary in ECM composition and organization depending on the force associated with the attachment (Benjamin et al., 2002; Benjamin and Ralphs, 1998; Hems and Tillmann, 2000; Apostolakos et al., 2014). Two crucial transcription factors specify cells and regulate expression of ECM components in a developing enthesis: the tenogenic fate determinant scleraxis, Scx, and the chondrogenic fate determinant Sox9 (Blitz et al., 2013; Sugimoto et al., 2013). Expression of Scx in mesenchymal stem cells (MSCs) drives them toward a tendon fibroblast (tenocyte) fate, whereas expression of Sox9 promotes cartilage differentiation (Akiyama et al., 2002, 2005). Co-expression of Scx and Sox9 has been described in cells of the OTJ enthesis of the Achilles tendon with the calcaneus as well as tendons attached to the mandible and supraspinatus-humerus (Blitz et al., 2013; Ideo et al., 2020; Roberts et al., 2019; Sugimoto et al., 2013; Kult et al., 2021). This is thought to be important for the establishment of the specialized ECM structure of these entheses, as both Scx and Sox9 directly regulate transcription of several collagens and other ECM components (Subramanian and Schilling, 2015). As a result, embryonic entheses have been defined by the co-expression of Scx and Sox9. In mice, Scx and Sox9 exhibit graded expression at fibrocartilaginous enthesis attachment sites, with cells closer to the bone expressing higher levels of Sox9 and cells closer to the tendon midbody and muscle expressing higher levels of Scx (Blitz et al., 2013; Sugimoto et al., 2013).This graded expression correlates with the gradient in stiffness and elasticity of enthesis tissue from bone to tendon, suggesting that it may play a causal role in tuning tendon ECM composition (Thomopoulos et al., 2010).

We have previously shown that mechanical force from muscle contraction controls transcriptional programs in tenocytes and the organization of tendon ECM in developing zebrafish (Subramanian et al., 2018). Loss of muscle in Myod1^−/−;Myf5^−/− mutant mice as well as induced paralysis in chick or mouse embryos disrupts the growth of bones and formation of bony eminences associated with entheses (Hall and Herring, 1990; Hosseini and Hogg, 1991; Rot-Nikcevic et al., 2006). For example, injection of alpha-Bungarotoxin (aBgtx) protein into the neonatal mouse brachial plexus to reversibly paralyze muscles disrupts collagen (Col) composition and other aspects of enthesis ECM organization, thereby compromising its function (Schwartz et al., 2013). However, it is not known how mechanical force regulates transcriptional dynamics during tenocyte differentiation at an enthesis or how this in turn affects interactions between tenocytes and the enthesis ECM.

Here, we describe the development of embryonic entheses during craniofacial development in zebrafish, and show that cells associated with muscle attachment sites on developing cartilages co-express scxa and sox9a transcripts, consistent with previously described presumptive enthesis progenitor identity (Sugimoto et al., 2013; Kult et al., 2021). We also describe the ECM organization at these and other developing cranial entheses, including the sternohyoideus enthesis, which elongates upon the onset of muscle contraction to resemble commonly studied entheses in long bones of mammals. Using both irreversible and reversible methods of paralysis, as well as a quantitative in situ hybridization method for measuring transcript levels, we show that mechanical force from muscle contraction regulates the establishment of graded scxa and sox9a expression at developing entheses (Subramanian et al., 2018; Niu et al., 2020). This is, in part, through TGFβ signaling, as pharmacological inhibition of TGFβ signaling also disrupts the ratio of scxa/sox9a expression along these developing entheses, suggesting that it plays an important mechanotransductive role in regulating the initial formation of entheses during embryonic musculoskeletal development.

Early cartilages of the craniofacial skeleton and associated tendons/ligaments in zebrafish embryos differentiate between 48 and 72 hours postfertilization (hpf) coinciding with the onset of head and jaw movements (Schilling and Kimmel, 1997; Subramanian et al., 2018). We used in situ hybridization chain reaction (isHCR) to assess spatial patterns of sox9a and scxa expression at developing cranial muscle attachments and sites of co-expression (presumptive entheses) (Fig. 1). We performed three-color isHCR using sets of probes to detect the cartilage marker sox9a (Fig. 1A), and the tenocyte markers scxa (Fig. 1B) and col1a1a (Fig. 1C) in 72 hpf wild-type (WT) embryos. These were compared with 72 hpf WT Tg(BAC scx:mCherry^fb301; sox10:gfp^ir937) double transgenic embryos stained with antibodies against GFP to detect sox10 in cartilage cells (Fig. 1E), mCherry to detect scxa in tenocytes (Fig. 1F) and myosin heavy chain (MHC) to mark jaw muscle fibers (Fig. 1G). In the ventral cartilages and muscles of the first and second pharyngeal arches, sox9a and scxa expression coincided with the GFP (chondrocytes) and mCherry (tenocytes) antibody signals, validating the isHCR probe set specificity.

In addition, we identified overlapping regions of expression of scxa and sox9a at many tendon-cartilage attachment sites that we hypothesized to be developing entheses (Fig. 1H). To aid in detecting enthesis cells and their 3D organization, we used DAPI to stain nuclei. By 72 hpf cranial tendons are at least partially functional and developing scxa⁺ tenocytes have largely migrated to sites of tendon attachments to cartilage (Fig. 1H). To aid in visualizing ECM organization at entheses, we also stained 4 day postfertilization (dpf) and 5 dpf embryos with antibodies against Col2a to mark cartilage/fibrocartilage (Fig. 1I,K), Col1a to mark the tendon midbody (Fig. 1L,N) and Tsp4b to mark myotendinous junctions (MTJ) (J,K,M,N). We could not perform dual staining for Col2a and Col1a as their respective antibodies were raised in the same host species. In elongated tendons, such as those that connect the sternohyoideus (SH) muscle to the ceratohyal (ch) cartilage – hereafter referred to as the ch-SH attachment – Col2a labeling extended from ch posteriorly toward SH (Fig. 1I-K, yellow asterisk). This coincided with localization of Col1a in ch-SH (Fig. 1L-N, yellow asterisk), expression of which partially overlapped posteriorly with Tsp4b (Fig. 1K,N, insets). This colocalization of Col2a and Col1a at the anterior end of ch-SH is similar to fibrocartilage described in tendon insertions and entheses in mice and humans (Fig. 1O,P) (Thomopoulos et al., 2010; Apostolakos et al., 2014).

As isHCR signal amplification depends on enzyme-free isothermal hybridization of labeled amplifiers with linkers on the probes, the fluorescent signal remains restricted to the vicinity of the expressed mRNA molecules, allowing a semi-quantitative estimate of expression levels (Choi et al., 2016). We focused on the overlapping regions of expression of scxa and sox9a in putative entheseal tenocytes at the ch-interhyal (IH) muscle attachment, hereafter referred to as ch-IH tendon (Fig. 1H), and Meckel's (mc)-intermadibularis anterior (IMA) muscle, hereafter referred to as mc-IMA, as these attachments initially form between 48 and 72 hpf and become functional in jaw movements during this period. We performed isHCR hybridization for sox9a and scxa in WT embryos at 48, 55 and 60 hpf, then quantified expression in tenocytes as defined by expression of scxa at the mc-IMA attachment by measuring mean total tenocyte fluorescence (Fig. 2A-O) in 3D regions of interest (ROIs) created per cell using the DAPI nuclear signal. Developing cranial tendons are populated by ∼30-50 scxa-expressing cells and only a small fraction of these cells lies at the interface of the tendon and cartilage – the future enthesis. After examining the tendon-cartilage junction in a z-stack, we selected eight cells per embryo (∼four cells for each attachment) to be quantified in five embryos, which was ∼70% of the tenocytes at the enthesis that co-express scxa and sox9a. Our analysis showed a decrease in sox9a expression and an increase in col1a1a expression between 48 and 55 hpf, followed by a decrease in sox9a expression between 55 and 60 hpf and increase in scxa expression from 48-60 hpf (Fig. 2P), As a result, the ratio of scxa versus sox9a increased, particularly from 55 to 60 hpf (Fig. 2Q). To determine ratios of scxa and sox9a in tenocytes based on their relative positions within putative entheses (i.e. distance from the cartilage attachment), we measured scxa and sox9a expression at positions 1-4, with ‘1’ being a cell located closest to the cartilage and ‘4’ a cell further from the cartilage, at 48, 55, 60 and 72 hpf in the mc-IMA enthesis (Fig. S1A-K). We found no significant positional variation in scxa expression except for a decrease at 72 hpf between cells in positions 2 and 3 (Fig. S1C). In contrast, sox9a expression was highest in tenocytes closest to the OTJ (positions 1 and 2) from 48 to 55 hpf, with a gradual decrease progressively further away. (Fig. S1F). At 72 hpf sox9a expression was low in cells at all positions, with lowest levels at position 4. These dynamics are reflected in the much higher ratio of scxa versus sox9a expression in cells further from the cartilage attachment. (Fig. S1L). Correspondingly we observed higher expression of col1a1a at 60 hpf and 72 hpf in tenocytes adjacent to the cartilage, with steep decreases in expression at positions 3 and 4. These results suggest that the differential expression of scxa and sox9a seen at mature entheses arises at embryonic stages of cartilage and muscle differentiation and correlates with cell position within an enthesis primordium.

The temporal correlation between the differential expression of scxa and sox9a and the onset of coordinated jaw muscle contraction (60-72 hpf) suggests that force from muscle contraction regulates these expression levels at embryonic entheses. To test this hypothesis, we quantified expression in tenocytes of the mc-IMA in paralyzed embryos. Homozygous mutants lacking the function of a calcium channel, voltage-dependent, beta 1 subunit (cacnb1^−/−), required for myofiber contraction, show no movement as embryos but develop with no obvious morphological differences from WT siblings for the first 4 dpf (Subramanian et al., 2018). We quantified expression of scxa and sox9a in tenocytes at mc-IMA attachments in cacnb1^−/− embryos at 48, 55, 60 and 72 hpf and changes in expression levels from WT for scxa (decreases of 39% at 60 hpf and 26% at 72 hpf), sox9a (increases of 1.2-fold at 60 hpf and 1.5-fold at 72 hpf) and col1a1a (decreases of 52% at 60 hpf and 59% at 72 hpf) (Fig. 3A-O; Fig. S2A-C). A lack of the active jaw movements in cacnb1^−/− embryos correlated with a failure to upregulate scxa and downregulate sox9a expression to the same extent seen in WT in tenocytes of the mc-IMA attachment, resulting in a 53% and 40% reduction in the ratio of scxa to sox9a at 60 and 72 hpf, respectively (Fig. 3T; Fig. S2D). We saw no difference in this ratio at 48 hpf, before any cranial muscle movements, but cacnb1^−/− embryos showed a slightly reduced ratio of scxa to sox9a starting at 55 hpf, before there were coordinated cranial muscle contractions but when individual muscle fibers still exhibited random sporadic contractions.

To determine whether muscle contraction has similar effects on gene expression in other tendons besides mc-IMA and their developing entheses, we quantified expression of scxa, sox9a and col1a1a in tenocytes along the length of the ch-SH tendon (see Fig. 1). Cells at positions 1-5, which are closest to the ch cartilage, showed strong changes in expression of sox9a, scxa and col1a1a similar to our observations in the mc-IMA enthesis. sox9a expression was higher in cells in position 1 in cacnb1^−/− embryos compared with cells at position 1 in WT controls. sox9a levels gradually decreased from anterior to posterior (1-5) along the tendon, which correlated with a decrease in the ratio of scxa to sox9a at the tenocytes closer to the enthesis in cacnb1^−/− embryos (Fig. S3A-J). These effects of muscle activity, although stronger in tenocytes at the enthesis, were also noticed in chondrocytes, as we observed a modest change in sox9a expression in adjacent chondrocytes in cacnb1^−/− embryos when compared with WT (Fig. S4A-E).

To confirm that the loss of muscle contraction and not some other role for Cacnb1 was responsible for these changes in scxa and sox9a expression, we injected aBgtx RNA into one-cell-stage embryos to block acetyl choline receptors in neuromuscular synapses and result in complete paralysis of zebrafish embryos for 3-4 days without noticeably disrupting other aspects of development (Subramanian et al., 2018). Similar to our results in cacnb1^−/− mutant embryos, the ratio of scxa to sox9a expression was significantly reduced (65%) in aBgtx-injected (aBgtx-inj) embryos at 72 hpf (Fig. 4A,B,E,F,K). These results further support that these changes in gene expression at developing entheses are due to mechanical force from muscle contraction.

If the hypothesis that force from muscle contraction drives the changes in gene expression holds true, then restoring muscle contraction in paralyzed embryos should rescue the relative expression levels of scxa and sox9a at the enthesis. To test this, we induced muscle contractions in aBgtx-inj paralyzed embryos at 72 hpf by electrical stimulation using a Grass simulator (20 V for 2 min) as described previously (Subramanian et al., 2018; Subramanian and Schilling, 2014). Quantification of scxa and sox9a expression using isHCR showed that, whereas stimulation upregulated expression of both genes in WT embryos (Fig. 4A-D,I,J), there was no significant change in scxa expression between aBgtx-inj and stimulated aBgtx-inj embryos (Fig. 4E-H,I). In contrast, sox9a expression decreased significantly in stimulated aBgtx-inj embryos when compared with non-stimulated aBgtx-inj animals (Fig. 4E-H,J). This resulted in a marked increase in the ratio of scxa to sox9a expression in stimulated aBgtx-inj embryos to levels similar to non-stimulated WT (Fig. 4K). These results suggest that mechanical force from muscle activity maintains a balance of scxa to sox9a expression at the developing enthesis primarily through repression of sox9a in tenocytes.

Several studies, including ours, have shown that mechanical force regulates TGFβ signaling in the musculoskeletal system (Havis et al., 2016; Maeda et al., 2011; Subramanian et al., 2018; Liu et al., 2015). TGFβ signaling also regulates Scx expression and tendon development in mammalian limbs (Pryce et al., 2009; Tan et al., 2020). TGFβ signaling actively regulates tenocyte morphogenesis in the zebrafish trunk (Subramanian et al., 2018). To test whether TGFβ similarly controls transcriptional regulation in developing cranial tendons, we stained 60 hpf embryos with antibodies against phosphorylated SMAD3 (pSMAD3) and quantified expression of pSMAD3 in entheseal tenocytes. We observed significant reductions in pSMAD3 localization in the nuclei of the tenocytes at the developing enthesis (Fig. S5).

As tools for tissue-specific ablation of TGFβ signaling are currently unavailable in zebrafish, to test whether the balance of scxa and sox9a expression that we observed at mc-IMA, ch-IH and ch-SH attachments of zebrafish requires TGFβ signaling, we used small molecule TGFβ signaling antagonists. We have previously shown that SB431542, a potent selective inhibitor of ALK5 receptors, effectively reduces TGFβ responses in tenocytes as assayed by immunostaining for pSMAD3 (Subramanian et al., 2018). We incubated embryos in either 50 µM SB431542 or another selective inhibitor of ALK5, SB525334 (similarly at 50 µM), at 60 hpf in DMSO containing embryo medium for 12 h. At 72 hpf, the embryos were fixed and divided into two batches. One batch was used to perform immunohistochemistry to assess the localization of pSMAD3. The other batch was used to perform isHCR and quantify and compare scxa and sox9a expression at the enthesis between DMSO-treated controls and inhibitor-treated embryos. Immunostaining of the embryos with anti-pSMAD3 antibody showed reduced nuclear pSMAD3 in tenocytes (SB431542-treated) at the mc-IMA attachment interface (∼15% lower than DMSO) with a stronger effect in SB525334-treated embryos (∼27% lower than DMSO) (Fig. S6). The remaining embryos used for isHCR showed significantly reduced scxa expression (∼14%) and a corresponding increase (∼40%) in sox9a expression in the tenocytes of both SB431542- and SB525334-treated embryos compared with the DMSO-treated control embryos, which leads to a lowered ratio (∼37% lower) of scxa to sox9a in the inhibitor-treated embryos (Fig. 5). These results, along with the reduction of pSMAD3 localization observed in paralyzed embryos, suggest that TGFβ signaling is required for maintaining the relative expression of scxa and sox9a at developing entheses.

The classical view of an enthesis is a region of specialized ECM and cells (tenocytes, fibrocartilage) at the skeletal insertion site of a tendon or ligament that helps transfer mechanical load from soft to hard tissue. Cells of entheses in developing tetrapod limbs co-express Scx and Sox9, the balance of which has been hypothesized to regulate ECM composition along the length of a tendon in response to force (Felsenthal and Zelzer, 2017; Saito et al., 2021). How entheses first form in the embryo, their initial expression of Scx and Sox9, and their responses to mechanical force at the level of individual cells remain unclear. Here, we show that small groups of tenocytes co-express scxa and sox9a in craniofacial tendon attachments to cartilage in embryonic zebrafish that prefigure future entheses on later endochondral/perichondral bones. We find that tenocytes at these early entheses exhibit heterogeneity in levels of sox9a transcripts that correlate with their positions i.e. sox9a levels decrease further from the cartilage attachment. This transcriptional heterogeneity occurs in tendons with different types of attachments – including flattened sheet-like attachments such as mc-IMA and long extended tendons such as ch-SH. Similar to the ECM organization described in mouse and human entheses, zebrafish enthesis also show graded Col2a and Col1a expression, highest in fibrocartilage adjacent to skeletal attachments, and higher Tsp4b closer to the MTJ. We show that these position-specific differences in expression are amplified at the onset of active muscle contraction and jaw movements in embryos. We further show that mechanical force from muscle contraction regulates the scxa/sox9a ratio in individual tenocytes in vivo, primarily through repression of sox9a, which inversely correlates with col1a1a expression in the tenocytes at the entheseal and tendon ECM and helps maintain this ratio downstream of force to promote enthesis development.

We also provide evidence that mechanotransduction through TGFβ signaling regulates embryonic entheses. Based on these results, we propose a model for entheseal development in zebrafish in which the transitional nature of enthesis ECM is organized and maintained through an optimal balance of Scx and Sox9 co-expression, which is regulated by mechanical force from muscle contraction through TGFβ signaling in entheseal tenocytes (Fig. 6). Loss of muscle function leads to a loss of force-dependent, TGFβ-mediated repression of Sox9, thereby reducing the Scx/Sox9 ratio. Our model agrees with evidence suggesting that Scx and Sox9 are expressed in opposing gradients along the length of more mature entheses in mice, with much higher levels of Sox9 closer to the bony insertion (Blitz et al., 2013; Kult et al., 2021). It suggests that these patterns of expression arise in embryonic enthesis progenitors and reflect co-expression in individual tenocytes. It further suggests that entheses are specified much earlier than previously reported, with establishment of the graded expression of Sox9a versus Scxa from stiffer to softer tissue domains coinciding with the onset of embryonic muscle contractions. This view also provides a mechanism of feedback between the ECM and the tenocytes that secrete it, as both Scx and Sox9 directly regulate transcription of key ECM components (Bell et al., 1997; Espira et al., 2009; Furumatsu et al., 2010; Leéjard et al., 2007; Subramanian and Schilling, 2015).

Both Scx and Sox9 directly regulate transcription of collagens, proteoglycans and other ECM factors necessary for the organization and function of cartilage and tendon tissues. Whereas Scx drives expression of Col1a, Sox9 directly activates expression of Col2a (Bell et al., 1997; Espira et al., 2009; Leéjard et al., 2007). Col1a and other components of the tendon ECM enable proper transfer of tensile force to the more fibrous, load bearing Col2a-rich cartilage ECM (Schwartz et al., 2013; Zelzer et al., 2014). Hence, by fine-tuning levels of Scx and Sox9 and their ECM targets, the enthesis tissue is engineered to transition from a soft elastic Col1-rich tissue to a strong fibrous Col2-rich tissue at the cartilage side (Blitz et al., 2013; Sugimoto et al., 2013). Our results demonstrate that in zebrafish this occurs, at least in part, through co-expression of Scx and Sox9 in the same cells, as tendons first become functional in the developing embryo, and that this requires force (Fig. 3). We show that graded assembly of the ECM also involves higher Col1a near the enthesis and tendon mid-body versus higher Tsp4b near the embryonic MTJ. Similarly, Tsp4 expression localizes to tendons and the MTJs of mice (Frolova et al., 2014). Thus our results expand the spectrum of graded ECM components that define different tendon regions (Fig. 1O,P).

Requirements for muscle function in enthesis development in developing limb tendons have been reported in studies of drug-induced paralysis in chick and Myod1;Myf5 mutant mice that lack muscles (Hall and Herring, 1990; Rot-Nikcevic et al., 2006). However, in both cases defects in long bones and loss of muscles could secondarily disrupt enthesis development. In contrast, paralysis induced by injection of αBgtx or mutation of cacnb1 in zebrafish embryos does not alter the morphologies of the craniofacial cartilage elements or the expression of sox9a in chondrocytes, consistent with a direct effect of force on tenocytes (Subramanian et al., 2018; Hoyle et al., 2022). By comparing paralyzed and WT embryos at key stages of zebrafish embryonic craniofacial development, we have identified the stage at which the ratio of scxa to sox9a responds to force in an enthesis. Coordinated contractions of cranial muscles begin at 60-72 hpf and our data suggest that differences in the ratio of scxa and sox9a expression between WT and cacnb1^−/− embryos arise due to downregulation of sox9a levels in the absence of muscle contraction in cacnb1^−/− embryos. This finding further strengthens our model that force regulates the optimum ratio of scxa to sox9a in developing entheses by downregulation of sox9a expression. Similar to other in vivo studies on the roles of muscle contractile force in tendon enthesis development and function (Eliasson et al., 2009; Maeda et al., 2011; Sharir et al., 2011; Shwartz et al., 2012, 2013; Felsenthal and Zelzer, 2017), our studies do not isolate the cells from the surrounding cellular environment and, as such, do not directly measure the effects of force on the cells (Chiquet et al., 2003).

Central to our model is the regulation of TGFβ signaling by force and corresponding changes in the tendon ECM. Similar to our previous studies on trunk tendons, we show that paralysis through injection of aBgtx mRNA reduces phosphorylation of SMAD3 in entheseal tenocytes, suggesting that TGFβ signaling at the enthesis in response to mechanical force regulates tenocyte gene expression (Subramanian et al., 2018). Further, we have also shown that stimulation of aBgtx-inj embryos rescued the ratio of scxa to sox9a to WT levels by restoring the repression of sox9a expression. Although this stimulation of 2 min is sufficient to rescue the ratio of scxa to sox9a, it does not appear to be sufficient to change scxa expression. Future studies are needed to study the differences in responsivity of scxa and sox9a transcription to magnitude and duration of muscle activity. We have also shown that force regulates TGFβ signaling in embryonic trunk tenocyte morphogenesis and ECM production (Subramanian et al., 2018). We further show that pharmacological inhibition of TGFβ signaling reduces the ratio of scxa to sox9a expression at 72 hpf, similar to paralyzed embryos. (Arimura et al., 2017; |Blitz et al., 2013; 2009; Roberts et al., 2019).

Although non-specific and global effects of TGFβ inhibitors are not completely ruled out, we are limited to this approach due to the lack of necessary genetic tools in zebrafish to specifically downregulate or upregulate TGFβ in craniofacial development. We have chosen two different inhibitors and have titrated a dose that does not cause other musculoskeletal defects. Mechanical force activates TGFβ signaling in mammalian tendons and skeletal joints (Arimura et al., 2017; van der Kraan et al., 2012). Lack of joint movement can lead to reduced TGFβ signaling, causing upregulation of proteolytic enzymes in the joint and an increased risk of osteoarthritis. TGFβ can be released from latent TGFβ-binding proteins (LTBPs) in the ECM in response to force (Maeda et al., 2011). Our results are consistent with previous studies of entheses in mice and rats, where both TGFβ signaling and mechanical force regulate the ratio of Scx to Sox9 expression, likely altering ECM composition (Blitz et al., 2013; Saito et al., 2021). However, our results go beyond these studies by investigating transcriptional dynamics at the level of individual cells of the enthesis to show: (1) influences of force on TGFβ signaling in cranial tendons at embryonic stages when entheses first form; (2) that the altered ratios primarily reflect changes in Sox9 expression; and (3) that the effect of force on entheseal expression of sox9a is consistent across different types of entheses. These expression dynamics could reflect interactions between Scx and Sox9 themselves or the ECM proteins they regulate and feedback through TGFβ and other ECM-associated signals. In vitro studies have hinted towards a role for Scx in controlling Sox9 expression in association with E-box transcription factors (Furumatsu et al., 2010). Future development of transgenic lines expressing components of TGFβ signaling is needed to gain better insights into the role of mechanical force on tendon and enthesis development.

In addition to TGFβ, several other signaling pathways have been implicated in enthesis development in mammals, including BMP, FGF, Notch and Shh (Arimura et al., 2017; Blitz et al., 2013; 2009; Roberts et al., 2019). At craniofacial osteo-tendinous entheses, FGF and Notch signaling control the balance of Scx and Sox9 expression, thereby influencing enthesis ECM composition and organization (Roberts et al., 2019). Mechanical force can impact these developmental regulatory signals in tendons. For example, mechanical force from muscle activity regulates the distribution of primary cilia on tenocytes in embryonic and adult entheses that regulate Shh signaling (Fang et al., 2020). In addition to these pathways, mechanotransduction in tendon and cartilage development could be activated by different types and magnitude of forces.

Recent work has shown that the adult enthesis is populated by Gli1⁺ cells, which appear to replace the Scx/Sox9-expressing embryonic enthesis cells (Fang et al., 2020; Felsenthal et al., 2018). This raises several questions relating to the transition from embryonic to adult enthesis development and maintenance. What is the role of the Scx/Sox9-expressing enthesis tenocytes? Perhaps they function at earlier stages to fine-tune the tendon ECM to match the force required for each muscle attachment and thereby establish an ECM template for tendon maturation. If so, what are the cues responsible for their replacement by Gli1⁺ adult enthesis cells? Does TGFβ signaling and the embryonic tendon ECM aid in establishment of the adult enthesis? Future studies, such as spatial transcriptomics to resolve cellular heterogeneity as tendons mature, could provide answers to these fascinating questions.

TgBAC(scx:mCherry)^fb301, Tg(-7.2sox10:gfp)^ir937 transgenic and cacnb1^{−/−ir1092};TgBAC(scxa:mCherry)^fb301 lines have been described previously (Hoffman et al., 2007; Subramanian et al., 2018). All embryos were raised in embryo medium (EM) at 28.5°C and staged as described previously (Kimmel et al., 1995; Westerfield, 2007). 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 University of California, Irvine Institutional Animal Care and Use Committee guidelines.

Full-length aBgtx mRNA was synthesized from Pmtb-t7-alpha-bungarotoxin (aBgtx) vector (Megason Lab, Addgene plasmid #69542) following a previously published protocol and injected into Tg(scx:mCherry) embryos at the one- to two-cell stage (Swinburne et al., 2015). Stock solution of 10 mM SB431542 (Selleckchem S1067, CID: 4521392) and SB525334 (Selleckchem S1476, CID: 9967941), selective inhibitors of the TGFβ type I (ALK5) receptor were prepared in DMSO and used to obtain a final working concentration of 50 µM in EM. Embryos were incubated in 50 µM SB431542, 50 µM SB525334 or equivalent volume of DMSO for 12 h. Treated embryos were rinsed in pre-warmed (28.5°C) EM before fixation for immunostaining or isHCR protocol.

Electrical stimulation was used to induce muscle contraction, as previously described (Subramanian et al., 2018; Subramanian and Schilling, 2014). Both aBgtx-inj and control embryos or larvae were anesthetized with Tricaine (ethyl 3-aminobenzoate methanesulfonate, Sigma-Aldrich, A5040, SID: 329770864), placed on a silicone plate in EM and stimulated for 2 min at 20 V, 6 msec duration, 4 pulses/sec frequency and 6 msec delay between successive pulses. Stimulated embryos were allowed to recover in EM for 12 h and further processed for immunostaining or isHCR protocol.

All embryos used for immunofluorescence experiments were fixed in 4% neutral pH buffered paraformaldehyde (PFA, Thermo Fisher Scientific, T353-500, CID: 712) for 2 h at room temperature (25°C) or overnight at 4°C. The embryos were washed with 1× Phosphate Buffered Saline (PBS, CID: 24978514) and permeabilized with cold acetone (Thermo Fisher Scientific, A94, CID: 349996362) for 15 min at −20°C. Following permeabilization, they were rehydrated in PBDT (1×PBS with 2% DMSO, CID 679, Alfa Aesar, 42780) and 1% Triton X-100 (Sigma-Aldrich, T9284, CID: 5590) and processed according to standard antibody staining protocol. Primary antibodies used: rat monoclonal anti-mCherry at 1:500 (Molecular Probes, M11217, RRID: AB_2536611), rabbit anti-pSMAD3 at 1:500 (Antibodies-online, ABIN1043888, RRID: AB_2725792), mouse monoclonal anti-myosin heavy chain (MHC) at 1:250 [Developmental Studies Hybridoma Bank (DSHB), A4.1025], mouse monoclonal anti-Collagen type I (Col1a) at 1:200 (DSHB, 8-3A5, RRID: AB_2877070), mouse monoclonal anti-chicken Collagen type II (Col2a) at 1:200 (DSHB, II-II6B3, RRID: AB_528165) and rabbit anti-Tsp4b at 1:1000 (Subramanian and Schilling, 2014; Webster et al., 1988; De Blas et al., 1984; Linsenmayer and Hendrix, 1980; RRID: AB_2725793). DAPI (Invitrogen, D1306, RRID: AB_2629482) was used at 1:1000 to mark cell nuclei (Subramanian et al., 2018). Preabsorbed secondary antibodies were all obtained from Jackson ImmunoResearch and used for indirect immunofluorescence at 1:1000: Alexa Fluor 488 conjugated donkey anti-rabbit IgG (711-545-152, RRID: 2313584) and Alexa Fluor 594 conjugated donkey anti-rat IgG (712-586-153, RRID: AB_2340691). After staining, embryos were mounted in 1% low melt agarose (CID: 168241) in 1×PBS and imaged.

Embryos were fixed in 4% neutral pH buffered PFA for 2 h at room temperature (25°C) and used for isHCR protocol as previously published (Trivedi et al., 2018). Scxa, sox9a and col1a1a probes used in this study were custom designed and synthesized by Molecular Instruments, Inc. Following discussion with the technical support team, we increased the concentration of the scxa probe set to 3× to improve consistency in signal. isHCR hairpin amplifiers used in this assay were: B1-Alexafluor 488, B2-Alexafluor 546 and B3-Alexafluor 647 (Molecular Instruments Inc.). Following signal development, embryos were mounted in 1% low melt agarose in 5× SSC and imaged.

Embryos processed for fluorescent immunohistochemistry were imaged using a Nikon A1 confocal system (RRID: SCR_020318) with a Nikon Eclipse Ti inverted microscope using a CFI Plan Apochromat VC 60XC (water immersion) objective. Embryos processed for isHCR were imaged using Leica TCS SP8 confocal system (RRID: SCR_018169) with a Leica DMi8 inverted microscope using a Plan Apochromat HC 40× (water immersion) objective. Confocal stacks were collected with system optimized spacing of 0.45 microns and were analyzed using ImageJ (RRID: SCR_003070) and Imaris 10 (RRID: SCR_007370) softwares. Tenocyte progenitors at prospective entheses were identified by their proximity to cartilage chondrocytes. To quantify expression of either scxa or sox9a in the selected cell, using ImageJ a projection of a z-stack comprising the entire cell was created (∼8-10 slices) using DAPI signal to identify the edges of the cell. Quantification of fluorescence in the selected cells was performed by creating ROIs around the DAPI signal corresponding to the tenocyte as previously described (Subramanian et al., 2018). In experiments where Imaris 10 was used, the ROI was created using DAPI as a reference for each tenocyte at each z-section. These ROI surfaces were combined and the mean intensity of scxa, sox9a and col1a was measured in these cellular volumes.

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 on the same coverslip in a single imaging session. To account for variation between measurements from individual embryos in each sample, we performed linear mixed effects modeling followed by Tukey post hoc pair wise using lmer and lmtest packages in R (Oberpriller et al., 2022). Graphical plots were created in SPSS, Excel and R (RRID: SCR_001905) (using ggplot2 package). Fluorescence intensity to quantify Smad3 phosphorylation was measured as previously described (Subramanian et al., 2018). Measurement of fluorescence intensity was performed for 8-10 cells per embryo (4-6 embryos per condition).

We thank Dr Jaylen Michael Lee from the UCI Biostatistics unit at the Institute for Clinical and Translational Sciences for advising us on conducting linear mixed effects modeling with our datasets. We also thank Dr Danny Dranow and Pavan Nayak for their help in revising and improving the manuscript.