Introduction

The extracellular matrix (ECM) is a three-dimensional (3D) network that provides cells with biochemical and structural cues1. In adult skeletal muscle, the ECM is organized into three layers, the endomysium, perimysium, and epimysium. Each layer has a different ECM composition, ranging from the basement membrane-rich endomysium to the fibrillar collagen-rich epimysium2. The epimysium is thought to be continuous with the tendon to facilitate the transmission of muscle-generated contractile forces to bone for motion3,4. Tendons are predominantly comprised of type I collagen fibers oriented along the loading direction, providing tensile strength, with a distinct pericellular ECM that surrounds linear arrays of tenocytes5,6. Linking these tissues is the myotendinous junction (MTJ), a specialized interface marked by a localized enrichment of ECM, including type XXII collagen (COL22A1) and periostin (POSTN)5,7. While the structure and composition of adult muscle, tendon, and MTJ ECM are well described, how the ECM components of these disparate tissues are deposited and integrated during development is less clear.

The cells of the musculoskeletal system have different embryological origins, yet seamlessly integrate during morphogenesis8. In the mouse, the limb bud develops at embryonic day (E)9.5 as an outgrowth of the lateral plate mesoderm and includes tenocyte precursors9. Scleraxis (Scx), a transcription factor expressed in developing tendons, is detected in limb buds at E109. Muscle progenitors migrate into the forelimb from the somite by E10.510, which requires the transcription factor Paired box gene 3 (Pax3)11. Muscle precursors then differentiate and fuse into myofibers12. By E12.5, tendon progenitors are aligned between muscle and cartilage, and the basic forelimb pattern is complete by E13.513,14. MTJ development is hypothesized to occur simultaneously with the condensation of tendon and muscle progenitors15,16. The COL22A1+ MTJ was observed as early as E13.5 in the forelimb with the onset of muscle contraction7. However, fibrillar collagen deposition is not thought to take place until after the progenitors reach their final destination, reportedly starting around E14.517.

Notably, interactions between tendon and muscle are not required for initial musculoskeletal patterning. Tendons form in the appropriate position in muscle-less limbs18,19,20, and muscle cells aggregate in the correct location even when tendon development is disrupted21, suggesting that the patterning of the musculoskeletal system and subsequent integration is regulated by factors autonomous of muscle and tendon progenitors. Specifically, tendon patterning factors are likely produced independently of the presence of muscle cells by either tendon progenitors themselves or by non-myogenic tissue types such as other connective tissue cells and the ECM.

Perturbation of transcription factors that regulate connective tissue cell behavior (e.g. Tbx3, Tbx5, Osr1, Osr2) disrupts musculoskeletal patterning, and it has been hypothesized that the ECM plays an instructive role in muscle-tendon integration22,23. However, the distribution of the ECM and how it relates to muscle-tendon development remains poorly defined. To determine the spatio-temporal distribution of the ECM during limb development and the influence of skeletal muscle on matrix deposition, we used tissue clearing, 3D imaging, and proteomics to investigate wildtype (WT) and muscle-less Pax3Cre/Cre limbs. Our data revealed that an aligned ECM forms between these tissues, independent of muscle cell migration, but the specification of MTJ-specific ECM is dependent on muscle.

Results

ECM alignment preceded MTJ formation

We previously demonstrated that MTJ-specific ECMs are deposited after E12.5 and are dependent on static and cyclic loading7. To investigate the organization of ECM fibers prior to the establishment of the final muscle and tendon pattern, WT murine forelimbs were analyzed at E12. The resolution of the ECM in 3D was enhanced by decellularization or SeeDB clearing (Fig. 1A-D’; see Methods). Wheat germ agglutinin (WGA), which binds to proteoglycans containing sialic acid and N-acetylglucosamine, was used as a global label for musculoskeletal tissue ECM24. In decellularized E12 forelimbs, type I collagen (COL1) and WGA+ proteoglycans were assembled in a linear, fibrillar structure that emanated from the ulna where the presumptive long triceps will insert (Fig. 1E-E”’). To visualize the distribution of myogenic progenitors in the context of the ECM, Pax3Cre/ZsGreen1+ (Pax3GFP+) limbs were cleared using SeeDB. These tissues were also counterstained for ECM highly expressed in the developing limb: tenascin-C (TNC), fibrillin-2 (FBN2), and type V collagen (COL5: COL5A1, COL5A2, COL5A3)25. Proximally, Pax3GFP+ cells were loosely aggregated (Fig. 1F-H; *), whereas they appeared to be oriented along the proximal-distal (p – d) axis of the limb towards the distal tip of the limb (arrowheads). FBN2+ and COL5+ fibers were also aligned along the p – d axis, around and distal to the Pax3GFP+ cells (Fig. 1G-H; arrowheads). Aligned COL1+ fibers were even more prevalent after tendon and muscle integration at E13.75 (Fig. 1I-I”; arrowheads).

Fig. 1: ECM alignment during murine triceps development.
figure 1

A–D’ To visualize the 3D ECM structure of the triceps (green, A), murine forelimbs (B) were decellularized (decell) in 0.05% sodium dodecyl sulfate (SDS; C) or optically cleared using SeeDB (D). Samples were imaged using confocal microscopy and the organization of the ECM (COL5; magenta) was consistent after both decell (C’) and SeeDB (D’). Representative images from E13.5-E13.75 forelimbs. E–E”’ Aligned COL1+ (magenta) and WGA+ fibers (green) inserted into the lateral triceps near the ulna (U) in decellularized wildtype (WT) E12 forelimbs. F–H In SeeDB-cleared E11.5-12 forelimbs, Pax3GFP+ cells (green, *) aligned along the proximal (p) - distal (d) axis (arrow) with TNC (magenta), FBN2 (magenta), and COL5 (grey). Insets show aligned ECM networks distal to Pax3GFP+ cells (G-d, H-d) and around Pax3GFP+ cells (F-p, G-p, H-p) (arrowheads). I–I” COL1+ fibers (magenta, arrowheads) extended between ScxGFP+ tendon (green) and MY32+ muscle (blue) in a wholemount E13.75 zeugopod. Magnification 10× (C’), 20× (E–E”), 25× (D’, F–H), 63× (I–I”), z-projection: z = 40 µm (B’, C’) and 11.4 µm (F–H-d), 3D rendering: z = 12.7 µm, scale bars = 1 mm (B–E); 100 µm (C’–I”). Representative images from N = 3 independent biological replicates.

Full size image

Forelimb ECM composition was partially disrupted by the absence of muscle in mice

To investigate the contribution of muscle progenitors in depositing the ECM, the proteomes of WT and muscle-less (Pax3Cre/Cre) forelimbs were compared. Pax3 is functionally knocked out in mice homozygous for Pax3Cre, preventing the migration of muscle progenitors into the forelimb26. The ECM protein composition, or matrisome, of E13.5 limbs was determined by tissue fractionation and liquid chromatography-tandem mass spectrometry (LC-MS/MS)27,28 (Fig. 2A, B). Previously, we only focused on the proteins identified in the insoluble, ECM-rich fraction29; however, matrisome components thought to be important for musculoskeletal development, including glycoproteins (e.g. THBS4, NID1)30,31 and matrisome-associated proteins (e.g. IGF2, GPC4)32,33, were extracted in other buffers while attempting to enrich for the ECM following25,27. Therefore, we combined the intensities of all five fractions, which were normalized to the amount of protein extracted in each buffer, similar to our previous work25.

Fig. 2: Forelimb ECM composition was partially disrupted by the absence of muscle.
figure 2

A Left: log10(combined LFQ intensity value) heat map of matrisome components manually grouped via ECM protein classification. Combined LFQ intensity values for different fractions were generated following25 and proteins identified in n ≥ 2 biological replicates were included. White boxes signify zero intensity values. Right: C1 Fluidigm single-cell RNA-sequencing (scRNA-seq) data of the matrisome identified in E10-E15.5 limbs generated by He et al., as previously reported in ref. 25. Color indicates log2(average expression +1) and circle size represents percentage of cells expressing that specific transcript. Cell types: 1 mesenchymal; 2 muscle 1; 3 muscle 2; 4 muscle 3; 5 chondrocyte; 6 perichondrial; 7 neural crest; 8 endothelial; 9 erythro-myeloid progenitors; 10 macrophage; 11 epithelial. Heat map and scRNA-seq data of affiliated proteins can be found in Fig S2. B Pax3Cre/Cre forelimb lacked limb musculature (MY32, myosin; magenta) at E13.5. The shoulder girdle begins below the line. Magnification 10×. Scale bar = 200 μm. C Two-way ANOVA revealed there were no significant differences in matrisome composition between genotypes at E13.5 (p > 0.05, Supplementary Data 1). D Volcano plot comparison of log2(LFQ) values for Pax3Cre/Cre and WT E13.5 forelimbs, significance based on p < 0.05. E Matrisome components exclusively identified in either WT or Pax3Cre/Cre embryos for all tissue fractions. F Select GO terms identified as significant when comparing ECM components that were exclusive to or enriched in Pax3Cre/Cre or WT E13.5 forelimbs. N = 3 independent biological replicates were analyzed.

Full size image

There was no significant difference in the overall abundance of proteins between WT and Pax3Cre/Cre limbs as a function of Cellular Compartment or Matrisome Classification27,28 (Fig. 2C). Many ECM that are more abundant in adult tendon compared to muscle (e.g. COL1A1&2, COL5A1, EMILIN1, FBN2, TNC)5, were not altered in the Pax3Cre/Cre limb (Figs. 2A, 3A, B, S1). Additionally, ScxGFP+ cells were still present in the absence of muscle (Fig. 3, S1), consistent with the lack of overall change in the tendon-related matrisome. ECM proteins exclusively identified, or significantly more abundant, in WT limbs (Fig. 2D, E; 3E, F, S1, S2), included THBS4, COL6A1&2, IGF2, and GPC4. Basement membrane-associated proteins were also more abundant in controls (LAMA5, LAMC1, COL6A6, NID1, COL4A1)34. In contrast, many cartilage-related proteins were significantly elevated (ACAN, COL2A1, COL9A1,2&3, COL11A1&2, VCAN), or exclusively identified (COL8A2, COL27A1), in the Pax3Cre/Cre limb35,36,37.

Fig. 3: Matrisome distribution is altered in developing forelimbs lacking muscle.
figure 3

A, B’ EMILIN1+ and WGA+ fibers (arrowheads) were observed in the tendons of both WT and Pax3Cre/Cre embryos at E13.75. C, D’ Forelimb tendons from WT, but not Pax3Cre/Cre, forelimbs were POSTN+ (arrow). E, F”’ FBN2 was still present in the ScxGFP+ area in Pax3Cre/Cre forelimbs, whereas THBS4 was not. Dotted line delineates region of triceps muscle. Magnification 25× (A–D), 10× (E, F), 3D rendering of decellularized limbs z = 110 µm (A–D), cryosections (E–F”’), scale bars = 100 µm. U ulna, Hu humerus. Representative images from N = 3 independent biological replicates.

Full size image

To assess which cells may be depositing different ECM components, we compared our LC-MS/MS results with scRNASeq data from forelimbs at E10.5-E15.038. Some cartilage-related genes were more highly expressed in the chondrocyte cell type compared to muscle cell types, consistent with the enrichment of ACAN, COL2A1, COL9A1,2&3, COL11A1&2, COL27A1 in the Pax3Cre/Cre limbs (Fig. 2A). Thbs4 transcripts were enriched in muscle subtypes (muscle 1 – 3 cell types), and the protein was exclusively found in the WT matrisome (Fig. 2). In contrast to the proteomic results, few basement membrane components were enriched in transcripts of muscle cells during this time window, aside from Col6a6 (Fig. 2), supporting the idea of a immature skeletal muscle basement membrane at this time25,39.

GO analysis of ECM exclusive to, or significantly more abundant in, either genotype generated terms specific to WT (laminin-1 binding, cell migration) and Pax3Cre/Cre limbs (glycosaminoglycan binding, chondrocyte differentiation, ossification, growth plate cartilage chondrocyte differentiation) reflective of the tissues more abundant in each (Fig. 2F). Similarly, when the whole proteome was analyzed for proteins exclusive to, or significantly more abundant in, different genotypes, cytoskeletal motor activity and musculoskeletal movement terms were specific to WT and extracellular matrix structural constituent, chondrocyte differentiation, and collagen type IX trimer were significant in Pax3Cre/Cre limbs (Fig. S2).

Tendon and epimysial-like ECM formed in the absence of muscle

The minimal effect of muscle knockout on the matrisome (Figs. 2, 3), combined with the presence of tendons in Pax3 mutant mice (Fig. S1)18,40, suggested that overall connective tissue patterning was maintained in the absence of muscle. Therefore, we next investigated how ECM patterning in E12.5-E14.5 forelimbs was affected in Pax3Cre/Cre mutants. In WT forelimbs, WGA+ ECM fibers were enriched where the long triceps attached to the ulna and scapula at E12.5 (Fig. 4A). At E13.5 and E14.5, fibers in the tendon bundles of WT forelimbs became more elongated and densely packed (Fig. 4C, E). In Pax3Cre/Cre forelimbs, WGA+ fibers had a similar distribution pattern as the WT at all timepoints, but were more loosely arranged in the tendon (Fig. 4B, D, F).

Fig. 4: Tendon and epimysial-like ECM formed in the absence of muscle.
figure 4

A–F” Decellularized E12.5-E14.5 WT and Pax3Cre/Cre forelimbs were rich in WGA+ fibers at the insertion of the long triceps into the ulna (U; arrowheads; A’–F’) and origin of the long triceps at the scapula (S; arrowhead; A”–F”). However, the fiber bundles were less dense in Pax3Cre/Cre compared to WT forelimbs. G, H For the long triceps of both E13.5 WT and Pax3Cre/Cre embryos (dotted outline), WGA+ fibers formed epimysial-like structures. I, J In Pax3Cre/Cre forelimbs, laminin (LAM+) blood vessels (magenta) were still present (arrowheads). Magnification 20×, z-projection: z = 230 µm (G, H), 138 µm (I), and 181 µm (J), scale bars = 1 mm (A–F), 100 µm (G–J). Representative images from N = 3 independent biological replicates.

Full size image

WGA+ fibril bundles formed epimysial-like compartments even in the absence of muscle, with fibers emanating from the tendon through the muscle belly in both phenotypes in triceps and biceps (dotted line, Fig. 4G, H, S3). Additionally, a 3D network of laminin (LAM)+/ CD31+ blood vessels (Fig. 4I, J, S1,S3) was still present in Pax3Cre/Cre forelimbs at E12.5-E13.5.

Fibrillar ECM persisted in the muscle belly of Pax3 Cre/Cre mutants but MTJ-specific ECM did not

To further characterize the ECM in developing muscle, the composition and mechanical integrity of decellularized E13.5 WT and Pax3Cre/Cre forelimbs were investigated. In the WT limb, COL1+, COL5+, and FBN2+ fibers were concentrated in the tendon and remained highly aligned crossing through the muscle belly. In the mutant limb, COL1 staining outlined an epimysial-like structure where the muscle should be and COL5+ fibers were more sparse compared to the WT limb. However, FBN2+ fibers were still present across the muscle belly region, similar to WT littermates (Fig. 5A–F, arrowheads).

Fig. 5: Fibrillar ECM persisted in the muscle belly of Pax3Cre/Cre mutants but MTJ-specific ECM did not.
figure 5

A–F WT or Pax3Cre/Cre decellularized E13.5 triceps tendon (arrowheads) and muscle belly fibers were COL1+, COL5+, and FBN2+. G–I’ In WT decellularized E13.75 forelimbs, the COL22A1+ MTJ (green) and WGA+ ECM fibers (magenta) elongated under tension and were restored to original configuration when tension was removed, as indicated by the fiducial markers (dotted outline). * denotes an unresolved fiducial marker. G’–I’ The MTJ exhibited elongation in under tension (dotted line outline). J Quantification of average COL22A1+ MTJ fiber length revealed MTJ length significantly increased when stretched under tension and returned to the relaxed configuration for N = 3 independent limbs *: P-value < 0.05 Tukey’s post-hoc test. K, L The MTJ of decellularized WT but not Pax3Cre/Cre forelimbs was marked by COL22A1+ (green) at the end of the WGA+ tendons (magenta, arrowheads). Magnification = 20× (A–F), 25× (G–K). Z-projection: z = 230 µm (A), 122 µm (B), 24 µm (C), 23.6 µm (D), 230 µm (E), 115 µm (F), 96.3 µm (G, G’, H, H’), 96.5 µm (I, I’), 3D rendering: z = 110 µm (J, K), scale bars = 100 µm (A–K), 25 µm (G’–I’). Representative images from N = 3 independent biological replicates.

Full size image

To assess if the ECM within the tendon and across the MTJ is mechanically intact, E13.75 forelimbs were decellularized, but not fixed with paraformaldehyde prior to staining for COL22A1, a matrisome component found only at the interface of muscle and tendon. The stylopod was loaded under tension using a custom tensile tester (see Methods and Fig. S5)41, and the COL22A1+ interface elongated. The original ECM configuration was restored when the tissue was returned to the resting position (Fig. 5G–J; see additional replicates in Fig. S6). Repeated measures one-way ANOVA found no significant difference in COL22A1+ MTJ fiber length between Relaxed, Stretched, and Return to Relaxed configurations (not significant (ns) p = 0.07). In contrast, Tukey’s post-hoc test showed a significant difference in COL22A1+ MTJ fiber length between Relaxed and Stretched configurations (*p = 0.03). There was no significant difference between Relaxed and Return to Relaxed configurations (ns; p = 0.66). We then assessed if COL22A1 was present in the Pax3Cre/Cre at the proximal end of the tendon; however, it was absent in the triceps as well as in the wrist (Fig. 5K, L, S4). Attempts were made to mechanically test the ECM still present in the triceps of Pax3Cre/Cre limbs; however, the tissue was too fragile. Significant disruption of the underlying tissue was caused by the removal of the skin precluding our ability to accurately characterize the mechanical properties of the underlying tissue. Overall, we demonstrate that the ECM in the developing limbs is aligned prior to MTJ formation, the overall patterning arises without the presence of Pax3+ cells, but MTJ-specific ECM requires muscle.

Discussion

While cellular interactions between muscle-tendon precursors in the formation of the limb are well described8, the role of the muscle in ECM distribution is unclear. Here, we found the overall organization of the ECM was not affected by the absence of Pax3+ myogenic progenitors, suggesting a critical role of non-myogenic cells, such as connective tissue cells, in musculoskeletal patterning. However, the formation of a cohesive MTJ was dependent on the presence of muscle.

We observed regions of ECM aligned early in forelimb development that were persistent after the integration of tendon and muscle (Fig. 1). Myogenic progenitors were previously reported to align prior to differentiating into myofibers42, and the ECM may enhance this alignment. For example, previous studies demonstrated aligned topographical cues43, as well as signals from specific ECM proteins (e.g. TNC, fibronectin, laminins), promoted muscle migration and differentiation44,45,46. The initial aligned networks may provide a template along which the ECM that ultimately links muscle and tendon is refined.

As distinct musculoskeletal system components developed, the ECM in the limb organized into the tendon, epimysium, muscle belly, and MTJ. The ECM identified in the tendon, epimysium, and muscle belly fibers (Figs. 4, 5) in the muscle-less Pax3Cre/Cre limb and similar fibrillar ECM composition (Fig. 2) suggest muscle cells do not substantially contribute to interstitial ECM deposition. The source of this ECM could potentially be derived from tendon and connective tissue cells.

In the Pax3Cre/Cre limb, there was an elevation of cartilage-related proteins, which can be explained by the relative increase of cartilage content in the absence of muscle (Fig. 2E). The abundance of many basement membrane proteins was higher in the WT limbs once the intensities of all fractions were summed. The extraction of basement membrane in the less stringent buffers (C, N, M, CS; Supplementary Data 1) could be due to the endomysium not being fully mature at E13.539. Indeed, the primary isoform of laminin α2 (LAMA2), which is deposited after E15.547, was not identified in our previous proteomic analysis until later in limb development25.

In the Pax3Cre/Cre limb, an epimysial-like structure appeared to connect to the tendon-like ECM similar to the WT limbs (Fig. 4, S3). While the epimysium has unique ECM components relative to the perimysium and endomysium in the adult2,48, whether distinct ECMs are enriched in the epimysium vs perimysium during development is not clear2,49,50,51. Nevertheless, the morphology and location of the epimysial-like structure persisted in the Pax3Cre/Cre limb (Fig. 5). These fibers could be synthesized by connective tissues cells marked by transcription factors Tcf4, Osr1, Osr2. Tcf4+ and Osr1+ connective tissue cells are known to regulate ECM deposition in developing limbs52,53. The less extensive ECM bundles observed in Pax3Cre/Cre tendons could result from the lack of muscle-derived signals (e.g. biochemical, mechanical) that are needed for tendon maturation, and this difference may have contributed to Pax3Cre/Cre forelimbs tearing during attempted mechanical testing18,40. Importantly, this epimysial-like structure appeared to correspond to connective tissue cell accumulation between the developing muscles22,54. While several mechanisms could regulate muscle splitting14, our data support the hypothesis that accumulations of connective tissue cells regulate muscle patterning in part through ECM fiber deposition22,54. The persistence of vasculature in Pax3Cre/Cre forelimbs may also contribute to patterning as platelet-derived growth factor B (PDGFB) from blood vessels mediated connective tissue cell compartmentalization by promoting secretion of ECM55. Nevertheless, muscle was needed for the maturation of the overall ECM. Notably, our work focuses on the forelimb. While Pax3 is required for muscle migration to the fore- and hindlimbs56, there are limb-specific differences in muscle57,58 and mesenchymal transcription factors59,60 and future work is needed to confirm similar conclusions can be made about lower limb muscles.

There were additional differences in matrisome distribution that were potentially influenced by muscle signaling and contraction. Proteomics indicated that THBS4 was significantly decreased or absent in Pax3Cre/Cre limbs. THBS4 was required for myoseptum stability, and the expression is mechanically regulated in zebrafish30 and we previously found that the absence of muscle contraction significantly decreased THSB4 in the mouse limb7. In addition, POSTN abundance decreased in Pax3Cre/Cre limbs, as observed in our prior study7. POSTN is a matricellular protein that is elevated in tendon and muscle during repair and can promote tissue regeneration61,62 and is required to maintain the structural integrity between the muscle fiber bundles and the myoseptum in zebrafish63. Spatial investigation of THBS4 and POSTN indicated that both were decreased in the tendons of Pax3Cre/Cre mice (Fig. S1), revealing the need for muscle interactions for the proper deposition of tendon ECM.

Additionally, COL22A1 was absent in the MTJ of the Pax3Cre/Cre limb. COL22A1 is necessary to maintain MTJ integrity in the zebrafish, where COL22A1 knockout or knockdown leads to disrupted muscle phenotypes64,65. COL22A1 deposition at the triceps MTJ begins around E13.5 with the onset of muscle contraction, and morphological maturation is dependent on cyclic and static loading7. Multiple cell types have been hypothesized to synthesize COL22A1 at the MTJ based on transcriptomic data, including muscle15,66,67, tendon/connective tissue68,69, or both70,71. Our data revealed that COL22A1 integration requires the presence of muscle in the limb. The results presented here suggest that muscle is required for localization of COL22A1/THBS4/POSTN to the MTJ and tendon and builds on zebrafish studies that demonstrated the role of COL22A164,65, POSTN63, and THSB415 in the structural integrity of the myoseptum. By investigating mammalian appendicular muscles where connective tissue-based structures form independent of muscle, we demonstrated the presence of an aligned ECM decoupled from muscle cell migration and differentiation, which cannot be assessed in the axillary myoseptum of the zebrafish.

To gain insight into what cells may be depositing different ECM, we compared our results with scRNASeq data from forelimbs at E10.5-E15.038. We observed some consistent trends with ECM highly expressed in cartilage between RNA and protein expression; however, the decrease in basement membrane expression in Pax3Cre/Cre limbs could not be explained by the differential distribution of RNA. Overall, current protein composition cannot be directly extrapolated from genetic assays, as many mechanisms prevent the translation of proteins from RNA prior to the synthesis of ECM proteins by connective tissue cells or myofibers72. In addition, there are likely reciprocal interactions between cell types that promote the expression and synthesis of key proteins.

In summary, we provide evidence that an aligned ECM is established as early as E12, which becomes integrated across the MTJ as the musculoskeletal system takes on the final pattern. This network is mechanically robust by E13.75 and may provide structural and biochemical cues to guide the formation of the muscle-tendon interface. The general pattern of the ECM in the long triceps muscle-tendon unit was not affected when muscle progenitors failed to migrate into the limbs; however, the expression of some matrisome components (e.g. COL22A1, THBS4, POSTN) was dependent on the presence of muscle. Future studies that specifically disrupt key ECM in distinct connective tissue cell populations will provide additional insight into the ECM-based mechanisms that orchestrate musculoskeletal assembly. Information gained from murine studies has the potential to increase the fidelity of tissue engineering models of the human musculoskeletal system70 by identifying developmentally regulated ECM that supports tissue growth, integrity, and recovery from injury.

Materials and Methods

Animal models and tissue collection

All murine experiments were approved by either the Purdue University or the University of Colorado Boulder Institutional Animal Care and Use Committee (PACUC or IACUC; protocol #1209000723 or #2705). PACUC and IACUC ensure that all animal programs, procedures, and facilities at the university adhere to the policies, recommendations, guidelines, and regulations of the USDA and the United States Public Health Service following the Animal Welfare Act and university Animal Welfare Assurance. We complied with all relevant ethical regulations for animal use.

Pax3Cre (#005549, cross between C57BL/6J and 129S1/SvImJ)26 and ROSA-ZsGreen1 (ZsGreen1; #007906, strain C57BL/6J)73 transgenic mice were obtained from the Jackson Laboratory. Pax3Cre/ZsGreen1+ were generated to label all Pax3-expressing cells and their progeny with GFP+, which also includes a small subset of endothelial cells74. Specifically, males heterozygous for the Pax3Cre transgene (Pax3Cre/+) were time-mated with females homozygous for ZsGreen1. Noon of the day a copulation plug was found was designated as E0.5. Pax3Cre/+ heterozygous mice were crossed to generate Pax3Cre/Cre embryos, which were identified by their neural tube and neural crest defects26 or determined by genotyping. WT controls were defined as Pax3Cre/+ or Pax3+/+. The sex of the embryo was not determined. To visualize tendon progenitors, mice containing the ScxGFP transgene75, kindly provided by Ronen Schweitzer, were time mated. E10.5-E14.5 embryos were harvested from dams euthanized via CO2 inhalation followed by cervical dislocation. The embryos were transferred to 1× phosphate-buffered saline (PBS) on ice. Removal of the yolk sac and amnion was performed under a dissecting microscope (DFC450, Leica Microsystems) to avoid damaging the embryos. After dissection, forelimbs were immediately either fixed in 4% paraformaldehyde (PFA; Fisher Scientific) at 4°C overnight then washed with PBS for the preparation of sample for vibratome sectioning or wholemount staining, or processed for decellularization or cryosectioning.

Forelimb decellularization

After dissection, freshly harvested forelimbs were mounted in 1% low gelling temperature agarose (Sigma-Aldrich) in a 10 mm × 10 mm× 5 mm biopsy cryomold (Tissue-Tek). The agarose cubes containing forelimbs were submerged in 1 mL solution containing 0.05% sodium dodecyl sulfate (SDS; VWR), with 2% penicillin-streptomycin or 0.02% sodium azide (Sigma-Aldrich), in PBS, and gently rocked at room temperature (RT). The SDS solution was replaced every 24-48 hours (h) until decellularization was complete, after 3-6 days. Upon decellularization, the agarose cubes were washed in 1× PBS buffer for 1 h, fixed with 4% PFA in PBS for 1 h, then washed with PBS for 1 h again with gentle rocking at RT. The decellularized forelimbs were carefully extracted from the agarose under a dissecting microscope, and stored in 1× PBS buffer at 4°C until stained and imaged.

Vibratome sectioning and wholemount preparation

Fixed, intact forelimbs were embedded in 3.5% low-melting-point agarose (Amresco) in 1× PBS and set at RT for 30 min to solidify. Agarose-embedded samples were then positioned in the appropriate plane and attached to the sample holder of a Leica VT-1000S vibratome with Loctite super glue, and the chamber was filled with PBS to maintain sample hydration. Forelimbs were sliced into 250 μm thick sections and placed in a tissue culture dish filled with PBS on ice. Extra agarose was dissected away from the tissue by forceps. Samples were stained via immunohistochemistry the same day.

For wholemount preparation, fixed, intact forelimbs were prepared by removing skin or embedded in 3% low-melting-point agarose and bisected in the dorsal-ventral plane using a scalpel.

Cryosectioning and immunohistochemistry

Forelimbs from Pax3Cre/Cre and WT embryos were fixed in 4% PFA for 1 h, washed with PBS 3 × 30 min before embedding in Optimal Cutting Temperature compound (OCT; Sakura Finetek), frozen with dry ice-cooled isopentane (Fisher Scientific), and stored at −80 °C. If ScxGFP+, the samples were then incubated overnight at 4°C in sucrose solution (30% wt/volume % solution of sucrose in PBS with 0.02% sodium azide). Subsequently, the samples were incubated in a 50% OCT: 50% sucrose solution for 30 min. Cryosections (10 µm thickness) were collected on His-bond glass slides (VWR) and processed following5. Sections were incubated with primary antibodies (Table S1) at 4 °C overnight, and washed with PBS for 3 × 5 min. Slides were then stained with secondary antibodies and probes (Table S1). Sections were imaged at 10× water using a Zeiss LSM 880 confocal.

Fluorescent labeling of ECM and imaging

After PFA fixation, decellularized forelimbs or vibratomed sections were incubated in blocking buffer [10% donkey serum diluted in 1× PBS with 1% Triton X-100 (PBST) and 0.02% sodium azide] for 16 h at 4°C to increase the ability of antibodies to permeate through the sample and to block non-specific binding. Samples were then incubated with primary antibodies (Table S1) diluted in blocking buffer, and gently rocked at 4 °C for 48 h. Samples were rinsed 3 × 30 min with 1% PBST at 25°C, and then incubated with secondary staining reagents diluted in blocking buffer, placed in a lightproof container, and rocked again at 4 °C for 48 h. WGA was used as a counter stain as it binds to sialic acid and N-acetylglucosamine and marks proteoglycans in the musculoskeletal architecture24. Finally, after rinsing 3 × 30 min with 1% PBST at 25 °C, samples were stored in 1× PBS at 4 °C until imaged. Samples were imaged using one of the following confocal microscopes: an inverted Zeiss LSM 880 confocal using either the 10× EC-Plan NeoFluar (NA = 0.3, working distance = 5.2 mm), or 25× multi-immersion LD LCI Plan-Apochromat (NA = 0.8, working distance = 0.57 mm), an upright Zeiss LSM 800 confocal using 20× water immersion Plan-Apochromat (NA = 1.0, working distance = 2.4 mm), or a Leica DM6 CFS STELLARIS upright confocal (Leica Microsystems) with LAS X software (V4.1.1.23273) using HC APO 10×/0.3 W U-V-I (NA = 0.3, working distance = 3.6 mm), HC FLUOTAR L 25×/0.95 VISIR (NA = 0.95, working distance = 2.4 mm), and HC APO L 63×/0.9 W U-V-I CS2 (NA = 0.9, working distance = 2.2 mm). Z-stack and tile functions were used to capture the entirety of the biceps and triceps muscles. Widefield images were acquired using a Leica M80 stereo microscope.

Forelimb optical clearing

To visualize the 3D morphology and spatial patterning of muscle cells embedded within the developing limb, and confirm that ECM structure was maintained after decellularization, we used SeeDB, a fructose-based clearing solution we previously utilized for the 3D visualization of musculoskeletal and kidney cells in the context of ECM organization7,76,77. Forelimbs were cleared following76,78. Fructose solutions of varying concentrations (20%, 40%, 60%, 80%, 100%, and 115% wt/vol) were generated by dissolving D-(-)-fructose (JT Baker) in Milli-Q water with 0.5% α-thioglycerol (Sigma-Aldrich) to prevent browning and 0.02% sodium azide to prevent fungal infection in the higher concentration solutions. Fixed and stained tissues were equilibrated to increasing concentrations of fructose by incubating in each formulation for 8-24 h under gentle rocking at RT.

Mechanical testing of limbs

Limbs were collected and decellularized as described in the forelimb decellularization section, but not fixed. After blocking, the limbs were stained with AF488-conjugated WGA and an antibody to COL22A1 as described above. Mechanical methods were modified from41,79. In brief, forelimbs were held using two bulldog clamps (Fine Science Tools) attached to the wrist and shoulder, respectively, taking care to avoid the long triceps tendon. Clamps were secured in a custom uniaxial loading rig controlled by a microrobotic system (FT-RS1002, FemtoTools) with samples submerged in PBS, and the humerus was cut from the bicep side to allow ECM fibers in the triceps to stretch freely during testing. The loading system was integrated under a confocal microscope for imaging at 25×, as shown in Fig. S5. To evaluate deformation in the ECM, rectangles were photobleached on either side of the MTJ to be used as fiducial markers. Nine laser lines centered at 488 nm were set to 100% power, and a z-stack was acquired to photobleach the markers. After photobleaching, a reference z-stack was taken of the tendon, MTJ, and muscle area. The distal end of the forelimb was then displaced 2.5 mm to stretch the limb, and a second z-stack was acquired for the same region of interest. Forelimbs were returned to the reference configuration to generate a third z-stack for the region of interest. Data shown representative of N = 3, with additional replicates shown in Figure S6.

Image processing

Confocal stacks were adjusted for brightness and contrast then rendered in 3D using FIJI (NIH). For vibratome sections, image stacks were processed into z-projections in FIJI. For mechanical testing, image stacks were processed into summed z-projections using FIJI. Average COL22A1+ MTJ fiber length was evaluated in FIJI by measuring the average length of 12 fibers from each MTJ region. To remove some of the optical distortion inherent to confocal microscopy, the Nearest Neighbor Deconvolution function within the ZenBlue software (Carl Zeiss Microscopy) or LAS X software (Leica Microsystems) was applied. Figures were assembled by taking snapshots of 3D-rendered image volumes or selecting 2D slices from stacks and were arranged using Adobe Photoshop and Illustrator. Data shown is representative of N ≥ 3.

Proteomic analysis

Forelimbs were microdissected from E13.5 Pax3Cre/Cre or WT embryos, the autopod region was removed, and the remaining tissue was snap frozen and stored at -80 °C. Dissected limbs from one embryo were pooled and processed for mass spectrometry as described in ref. 25, with slight modifications. Briefly, proteins were fractionated into five samples (C, N, M, CS, IN) using homemade buffers5, enzymatically digested into peptides, and peptides were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS). Data shown from N = 3 biological replicates.

Digested peptides were analyzed in an Dionex UltiMate 3000 RSLC nano System (Thermo Fisher Scientific) coupled on-line to Orbitrap Fusion Lumos Mass Spectrometer (Thermo Fisher Scientific) as previously described80. Briefly, reverse phase peptide separation was accomplished using a trap column (300 μm ID × 5 mm) packed with 5 μm 100 Å PepMap C18 medium coupled to a 50-cm long × 75 µm inner diameter analytical column packed with 2 µm 100 Å PepMap C18 silica (Thermo Fisher Scientific). The column temperature was maintained at 50 °C. One µg peptides (equivalent volume) was loaded to the trap column in a loading buffer (3% acetonitrile, 0.1% FA) at a flow rate of 5 µL/min for 5 min and eluted from the analytical column at a flow rate of 200 nL/min using a 130-min LC gradient. Column was washed and equilibrated by using three 30-min LC gradient before injecting next sample. All data were acquired in the Orbitrap mass analyzer and data were collected using an HCD fragmentation scheme. For MS scans, the scan range was from 350 to 1600 m/z at a resolution of 120,000, the automatic gain control target was set at 4 × 105, maximum injection time 50 ms for both MS1 and MS2, and dynamic exclusion was 30 s. MS data were acquired in Data Dependent mode with cycle time of 5 s/scan. MS/MS data were collected at a resolution of 7500 with isolation window of 1.2 m/z. The RF lens and collision energy were set at 30%.

Raw data files were processed by MaxQuant81, and raw or LFQ intensities were used for downstream data analysis82. Identified proteins were annotated by Cellular Compartment and Matrisome Classifications28. For both raw and LFQ values, intensities were summed across all five fractions following previously published methods25; using WT forelimbs of the same age (E14.5), we measured the relative amount of protein extracted in each buffer. Using these ratios, the intensity from each fraction was normalized to the relative amount of protein extracted in the corresponding buffer, and summed across all five fractions. After combining intensities, the distribution of these classifications were calculated as a percentage of total raw intensity25.

Statistics and Reproducibility

For volcano plot analysis of ECM proteins identified in the IN fraction, combined LFQ intensities were log2 transformed, the fold-change was calculated between Pax3Cre/Cre and WT samples, and Welsh’s t-test was used for statistical comparisons. LFQ intensities were also used for heatmap analyses, wherein proteins were first clustered by Matrisome Classification and then by abundance. Raw intensities were used for ECM protein identification in the other fractions. Data imputation was not performed83. Adjusted p-values for GO driver terms were calculated in g:Profiler84, by comparing the lists of the proteins that were exclusive or significantly more abundant in WT or Pax3Cre/Cre. Mechanical testing quantification was analyzed using GraphPad Prism using a repeated measures one-way ANOVA and Tukey’s post-hoc comparisons test on n = 3 biological replicates.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.