Organogenesis of the somatic musculature in Drosophila is directed by the precise adhesion between migrating myotubes and their corresponding ectodermally derived tendon cells. Whereas the PS integrins mediate the adhesion between these two cell types, their extracellular matrix (ECM) ligands have been only partially characterized. This study shows that the ECM protein Thrombospondin (Tsp), produced by tendon cells, is essential for the formation of the integrin-mediated myotendinous junction. Tsp expression is induced by the tendon-specific transcription factor Stripe, and accumulates at the myotendinous junction following the association between the muscle and the tendon cell. In tsp mutant embryos, migrating somatic muscles fail to attach to tendon cells and often form hemiadherens junctions with their neighboring muscle cells, resulting in nonfunctional somatic musculature. Talin accumulation at the cytoplasmic faces of the muscles and tendons is greatly reduced, implicating Tsp as a potential integrin ligand. Consistently, purified Tsp C-terminal domain polypeptide mediates spreading of PS2 integrin-expressing S2 cells in a KGD- and PS2-integrin-dependent manner. A model is proposed in which the myotendinous junction is formed by the specific association of Tsp with multiple muscle-specific PS2 integrin receptors and a subsequent consolidation of the junction by enhanced tendon-specific production of Tsp secreted into the junctional space (Subramanian, 2007).

The development of functional musculature depends on the correct encounter and adhesion of muscles with their corresponding tendon cells. In Drosophila the hemiadherens junctions, formed on both sides of the myotendinous junction, mediate the adhesion between muscles and their corresponding tendon cells. The muscle-specific integrin heterodimer αPS2ßPS accumulates at the muscle counterpart of this junction, and binds to its specific extracellular matrix (ECM) ligand Tiggrin. Correspondingly, the tendon-specific integrin heterodimer αPS1ßPS accumulates at the tendon counterpart of the myotendinous junction, and is thought to associate with the laminin ligand. Both hemiadherens junctions on each cell type exhibit a symmetrical distribution, raising the possibility that, although each cell utilizes a distinct integrin heterodimer, the formation of the myotendinous junction is coordinated between the two cell types. In the absence of the common ßPS subunit (Myospheroid), muscles initially interact with tendon cells; however, following muscle contraction the muscles detach from the tendon cells and round up (the myospheroid phenotype). Notably, lack of the muscle-specific αPS2 subunit similarly leads to muscle detachment; by contrast, however, lack of the tendon-specific αPS1 (e.g., in the mew mutant embryos) does not lead to muscle detachment. mew mutant embryos hatch, suggesting that occupation of the muscle-specific αPS2ßPS by its ligand may be sufficient for the formation of embryonic myotendinous junctions. The αPS1 belongs to the laminin-binding type α family of receptors and binds to laminin. Drosophila laminin may consist of ß1 and ß2 subunits and either of two laminin α subunits. The αPS1 is thought to associate with laminin containing the LanA subunit (also known as α3,5), which when deleted does not exhibit significant muscle-tendon attachment defects. By contrast, lack of the laminin α1,2 (wing blister), which associates with the αPS2ßPS (Graner, 1998), results in a mild muscle-detachment phenotype (e.g., wing blister mutants) (Martin, 1999), pointing to the crucial function of the muscle-specific PS2 in the formation of the myotendinous junction (Subramanian, 2007).

Tiggrin, a Drosophila-specific ECM component, has been shown to associate with the muscle-specific αPS2ßPS integrin. However, homozygous tiggrin mutant embryos do form muscle-tendon junctions and the adult flies are only semilethal (Subramanian, 2007).

In addition to its role in the establishment of myotendinous junctions, integrin-mediated adhesion is essential for several biological processes, including dorsal closure, visceral mesoderm development and the development of the adult fly wing. Wing epithelial cells from the dorsal and ventral aspects of the wing form specialized integrin-mediated adherens junctions required for the development of the adult fly wing. At morphogenesis dorsal wing epithelial cells expressing αPS1ßPS are brought together with ventral cells that express αPS2ßPS. Adhesion between these two epithelial sheets of cells is presumably mediated by specific ECM ligands. Although the involvement of the laminin α1,2 (wing blister) has been described, ligand specificity of each of the PS integrin receptors in this context has yet to be elucidated (Subramanian, 2007).

Tendon cells are specified in the Drosophila ectoderm as a result of the activity of the tendon-specific transcription factor Stripe. Embryos mutant for stripe do not develop normal tendon cells. Conversely, ectopic expression of Stripe leads to ectopic development of tendon cells. Thrombospondin was recovered in a search for genes that are regulated by the tendon-specific transcription factor Stripe (Subramanian, 2007).

Thrombospondins (Tsps) are a family of extracellular matrix proteins that mediate cell-cell and cell-matrix interactions by binding membrane receptors, extracellular matrix proteins and cytokines (Adams, 2001; Lawler, 2000). In vertebrates there are five tsp genes expressed in various tissues, including the brain (TSP1 and Tsp2), bones (Tsp5) and tendons (Tsp4). Tsp1 and Tsp2 are closely related trimeric proteins that share the same set of structural and functional domains. Tsp4 and Tsp5 are pentameric and differ from Tsp1 and Tsp2 in their domain arrangement. All Tsps share a typical C-terminal domain (CTD) that contains epidermal growth factor (EGF)-like repeats, and a Ca-binding domain. The N-terminal domain contains additional conserved regions including the laminin G-like domain (which is not present in Tsp5) (Adams, 2004). Drosophila tsp is encoded by a single gene that is spliced into four variants, among which only one (TspA) contains the conserved CTD, which in addition to the EGF repeats and Ca-binding domains also includes a putative integrin-binding KGD motif. The N-terminal domain contains a conserved heparin-binding domain and putative integrin-binding motifs RGD and KGD. Drosophila Tsp (Adams, 2003) is closest in structure to vertebrate Tsp-5/COMP, which is expressed mainly in cartilage and certain other connective tissues and has a role in chondrocyte attachment, differentiation and cartilage ECM assembly (Subramanian, 2007).

A wide range of functions has been attributed to the different Tsps, including a role in platelet aggregation, inflammatory response, regulation of angiogenesis during wound healing, and tumor growth (Adams, 2004). Tsp1 and Tsp2 have been described (Christopherson, 2005) as astrocyte-secreted components that promote synapse formation in the CNS (Subramanian, 2007).

The large isoform of Drosophila Tsp has been shown to form pentamers and exhibits heparin-binding activity. Its major sites of expression in the embryo are the muscle attachment sites, and also the precursors of the longitudinal visceral muscles. In larval stages (Adams, 2003) it is expressed in wing imaginal discs (Subramanian, 2007).

This study reports that Drosophila Tsp is a key ECM component that is required for muscle-specific adhesion to tendon cells. In tsp mutant embryos muscles fail to attach to tendon cells, and often aggregate and form ectopic integrin-mediated junctions with neighboring muscles. This leads to nonfunctional somatic musculature and embryonic lethality. In the embryo, Tsp is required for integrin-mediated adhesion as measured by Talin-specific accumulation. Furthermore, Tsp can functionally bind to αPS2ßPS-integrins; the purified CTD of Tsp mediates PS2 integrin-dependent cell spreading in a KGD- and PS2-dependent manner (Subramanian, 2007).

Taken together, these results suggest a model whereby Tsp produced by tendon cells is required for muscle-specific adhesion to tendons by binding the muscle-specific αPS2ßPS integrin receptors, and a subsequent consolidation of the junction by enhanced tendon-specific production of Tsp secreted into the junctional space (Subramanian, 2007).

It is suggested that the dynamics of myotendinous junction formation involve the following sequential steps. (1) When the myotube is very close to the tendon cell, Tsp secreted continuously from the tendon cell associates with the muscle leading edge and binds to the muscle-specific αPS2ßPS integrin receptors. Because Drosophila Tsp forms pentamers, each pentamer potentially associates with several PS2 receptors, leading to accumulation of αPS2ßPS receptors at the myotube leading edge. This association triggers integrin-mediated adhesion and Talin accumulation at the cytoplasmic tail of the PS2 integrin receptors. (2) Tsp may bind to the tendon surfaces through an unknown ligand. (3) Stripe levels in the tendon cell are elevated following the establishment of the muscle-tendon junction, because of Vein-EGF receptor (EGFR) signaling. Stripe induces the elevation of Tsp levels, creating a positive feedback loop that encourages further secretion and accumulation of Tsp at the junction site, strengthening the myotendinous junction (Subramanian, 2007).

The KGD site in the CTD of Tsp was shown to trigger PS2 integrin-dependent cell spreading. This sequence had been shown to bind certain types of vertebrate integrin receptors (Scarborough, 1993). The N-terminal domain of Tsp contains an additional KGD site, and an RGD site, both implicated in integrin-binding activity. These sites may also contribute to the binding of the PS2 muscle-specific integrins. Therefore, each Tsp pentamer contains multiple binding sites for PS2 integrin receptors, and thus may induce receptor aggregation at the muscle leading edge. It remains to be determined whether Tsp is capable of binding to PS1 integrins or other receptors expressed by the tendon cell (Subramanian, 2007).

Whether Tsp functions as an integrin ligand in other tissues (e.g., midgut, salivary gland, dorsal closure and the wing epithelium) is yet to be elucidated. Phenotypic analysis of the tsp^8R mutant embryos did not reveal a major phenotype in the gut, CNS or dorsal closure. Similarly, tsp^8R mutant clones induced at the larvae stage did not result with wing blisters as in integrin-induced clones. Although mutants for the tsp^8R allele did not show staining with the anti-Tsp antibody, it is still possible that residual Tsp activity is retained in the mutants because of the activity of the other TSP isoforms (which were not affected by the deletion of the EP excision at the CTD). In addition, maternal tsp transcripts were detected that may partially rescue the zygotic tsp phenotype in the early developmental stages (Subramanian, 2007).

An additional relevant ECM component at the myotendinous junction is laminin. Laminin α1,2 (encoded by wing blister) is required for the formation of the myotendinous junction (Martin, 1999). Laminin α1,2 contains an RGD sequence and also binds to the PS2 integrins (Graner, 1998), demonstrating the crucial role of these receptors in the formation of the myotendinous junctions. It is possible that laminin containing the laminin α1,2 subunit associates with Tsp in the myotendinous junctional space. Both laminin and Tsp carry a heparin-binding domain and it is possible that they interact indirectly through a putative heparin-containing proteoglycan. Because no changes in laminin distribution was detected following overexpression of Tsp (using anti-laminin antibody), it is not thought that there is any direct Tsp-laminin interaction. The heparan sulfate glycoprotein Syndecan is produced by the muscle cells. In syndecan mutant embryos the somatic muscle pattern is defective, a phenotype that is attributed to an effect of Syndecan on Slit distribution and function. However, Syndecan at the muscle cell membrane may contribute to a putative indirect interaction between Tsp and laminin through its heparin-containing domain. Such interaction may enhance the accumulation of ECM components such as Tsp and laminin at the myotendinous junction. In support of this hypothesis, vertebrate Tsp has been shown to bind Syndecan at its CTD (Adams, 2004). However, syndecan homozygous mutant embryos do not exhibit alterations in Tsp distribution, arguing against a central role for Syndecan in Tsp distribution. Nevertheless, it remains possible that another heparin domain-containing protein functions to promote Tsp and laminin deposition at the myotendinous junction (Subramanian, 2007).

It is considered that the Stripe-dependent positive feedback that upregulates tsp transcription contributes significantly to the establishment of the myotendinous junction. Previous studies have shown that muscle-tendon interactions form a signaling center, which is initiated by muscle-dependent Vein secretion and accumulation at the myotendinous junction. Vein activates the EGFR pathway in the tendon cell, leading to a significant elevation of the transcription factor Stripe. This study shows that Stripe induces upregulation of Tsp. Taking these results together, it is suggested that the initial formation of the hemiadherens junction creates a self-auto-regulatory nucleation center, which leads to additional deposition of Tsp and possibly other ECM components. These, in turn, gradually strengthen the hemiadherence junction formed between the muscle and the tendon cell (Subramanian, 2007).

Vertebrate Thrombospondins are essential for a variety of biological activities, including cell adhesion, migration, angiogenesis, etc. This work reveals an intriguing similarity between the role of Tsp in the formation of the myotendinous junction and the role of vertebrate Tsp1 and Tsp2 in the induction of synapses. It was shown that Tsp provided by oligodendrocytes is a potent inducer of synapse formation on the dendrites of cultured neurons (Christopherson, 2005). Although these synapses are not electrically active, the Tsp-induced synapses exhibit typical synaptic ultra-structures. The biogenesis of the myotendinous junction carries several similarities to the biogenesis of synapses, including the mutual crosstalk between the two cell types involved and the gradual formation of the junction at both cell membranes involved (Subramanian, 2007).

In summary, this analysis of Tsp function reveals the molecular dynamics and biogenesis of the myotendinous junction. A similar scenario may unfold during Tsp-dependent synapse formation in the development of vertebrate embryos (Subramanian, 2007).

GENE STRUCTURE

cDNA clone length - 4193 bases

Bases in 5' UTR - 904

Exons - 13

Bases in 3' UTR - 105

PROTEIN STRUCTURE

Amino Acids - 1060 (isoform A)

Structural Domains

Drosophila TSP gene is divided into 13 exons; The translation initiation site is in exon 2. The transcription start site was analyzed using a PCR procedure, and the longest transcripts were found to initiate about 300 bp 5' of the predicted start site. The open reading frame of the DTSP cDNA encodes a protein that has 1060 amino acid residues. The polypeptide is composed of domains or repeats characteristic of the TSP3/TSP4/COMP subfamily of thrombospondin proteins: Amino-terminal domain, four Type II repeats, seven Type III repeats, carboxyl-terminal domain. The protein is highly acidic, particularly in the region of Type III repeats, with an Asp + Glu content of 15.8%. A signal peptide was detected at the N-terminus, which indicates that DTSP, like other TSPs, functions as an extracellular protein. Ten Asn residues were identified as potential glycosylation sites. Alignment of the amino acid sequences of the Drosophila TSP with human TSP1-TSP4 and COMP demonstrated a high degree of homology between the four Type II repeats, seven Type III repeats, and C-terminal domain (Adolph, 2001).

Thrombospondins (TSPs) are multidomain oligomers that have complex roles in cell interactions and tissue organisation. The five vertebrate TSPs comprise two subgroups, A and B, that are assembled as trimers or pentamers, respectively. A new dataset containing the single TSP of Drosophila and four other newly identified invertebrate TSPs has been developed to examine the phylogenetic relationships of TSPs. These analyses clearly indicate pentamerisation as an early attribute of TSPs. Drosophila TSP is assembled as a pentamer, has heparin-binding activity and is a component of extracellular matrix (ECM). During embryogenesis, the TSP transcript is concentrated at muscle attachment sites and is expressed by a subset of myoblasts and in imaginal discs. These novel results establish TSPs as highly conserved ECM components in both invertebrates and vertebrates and open fresh perspectives on the conservation of structure and biological function within this family (Adams, 2003).

The vertebrate TSP family includes five members that form two subgroups, A and B, according to their oligomerisation status and molecular architecture. TSP-1 and TSP-2, in subgroup A, form homotrimers. TSP-3, TSP-4 and TSP-5 (also known as cartilage oligomeric matrix protein, COMP), in subgroup B, form homopentamers (Adams, 2003 and references therein).

Each subunit of the TSP-1 and TSP-2 trimers contains an amino-terminal domain, a coiled-coil oligomerisation domain, a procollagen homology domain (also known as von Willebrand factor type-C domain), three repeated TSP type 1 or properdin domains, three repeated TSP type 2 or EGF-like domains and seven TSP type 3, calcium-binding repeats and a globular C-terminal region that is, to date, unique to TSPs. The subunits of the pentameric TSPs differ in having four of the TSP type 2 domains and do not contain the procollagen or the TSP type 1 domains. TSP-5/COMP also lacks a distinct amino-terminal domain. The hallmark of a thrombospondin subunit is thus the contiguous structural grouping of repeated EGF domains, TSP type 3 repeats and the unique C-terminal region. Physical characterisations have indicated that the TSP type 3 repeats and C terminus undergo close structural interactions and may together constitute a protein structure domain (Adams, 2003 and references therein).

TSPs are widely dispersed in vertebrates and phylogenetic analysis of these TSPs has given rise to discussion of the possible existence of TSPs in modern invertebrates. However, although the coiled-coil, procollagen-homology, type 1, or EGF-like domains are each found separately in invertebrate proteins, the Caenorhabditis elegans genome does not appear to encode a TSP. Recently, however, a single TSP-like coding sequence was found within the Drosophila melanogaster genome by rigorous database homology searching and was confirmed by cDNA sequencing (Adams, 2001). In parallel, the same open reading frame was characterised at cDNA level by PCR based on the genome project-predicted gene sequence (Adams, 2003 and references therein).

The D. melanogaster TSP-like protein (D-TSP) contains a unique amino-terminal domain, a heptad-repeat oligomerisation sequence, six repeated TSP type 2/EGF-like domains, seven TSP type 3 repeats and a carboxy-terminal domain. To identify further invertebrate TSPs, the TSP-1, TSP-5 and D-TSP amino acid sequences were used in TBLASTX searches of the expressed sequence tag (EST) database, with exclusion of human and mouse ESTs. All searches identified the same coding sequences in three other invertebrate species, that included a second dipteran insect, Anopheles gambiae; the lepidopteran insect Bombyx mori, and the urochordate, Ciona intestinalis. These sequences were identified as TSPs with high confidence, because they were the only matches that had high, statistically significant, overall sequence identity with all other TSPs and because the predicted protein sequences contained contiguous type 2 domains and type 3 repeats. In the case of A. gambiae, an EST with 52% sequence identity to the amino-terminal domain of D-TSP was also found. Both of the cDNA sequences were encoded in the same region of the A. gambiae genome and other TSP-like sequences were not detected elsewhere in the genome. These studies established that the A. gambiae genome encodes a single TSP with a domain structure similar to that of D-TSP. As an ascidian, C. intestinalis is representative of the smallest chordate genome. Strikingly, the searches of Ciona ESTs and genome revealed two forms of TSP that corresponded to one example each of the A and B subgroups. The domain organisation of Ciona TSP-B corresponded to that of TSP-3 and TSP-4. Ciona TSP-A was very similar to TSP-1 and TSP-2, except that the cysteine residues potentially involved in oligomerisation were not located within a heptad repeat sequence. Both Ciona TSPs contained amino-terminal domains related to the laminin-G fold predicted for TSPs (Adams, 2003)

TSP-1 is the vertebrate TSP that has been studied most intensively, and several short, linear peptide motifs have been identified that function in matrix-binding and cell attachment. Some of these are present in the other vertebrate TSPs. It is of interest that several of these peptide motifs are present in insect TSPs. Of particular interest, the amino-terminal domains of both insect TSPs contained two BBXB (where B stands for basic and X for any residue) heparin-binding motifs. The insect N-terminal domain was unique in relation to other TSP amino-terminal domains in containing a potential integrin-binding RGD motif. Certain species orthologues of TSP-4 and TSP-5/COMP contain an RGD motif in the third type 3 repeat. None of the invertebrate TSPs had an RGD motif at this site; however, D-TSP had a KGD motif in the last type 3 repeat, at the position equivalent to that of the RGD motif of TSP-1 that is known to function in integrin-binding. The KGD motif, identified in the snake venom barbourin is a binding motif for certain integrins. In contrast, A. gambiae TSP had an RGD motif in the sixth type 3 repeat. The C terminus of TSP-1 contains two motifs that bind CD47, IRVVM and RFYVVM. The RFYVVM motif is found in all vertebrate TSPs. The IRVVM motif was not present in the insect TSPs and the sequence at the RFYVVM site was KYYVVQ (fly) or KFYAVM (mosquito). Experiments with peptides have indicated that the VV dipeptide is needed for activity and that substitution of any one of the I, R or M residues in IRVVM abrogates activity. CD47-binding thus appears to be a feature of vertebrate TSPs. This is in accord with the absence of a CD47-like molecule encoded in the Drosophila and Anopheles genomes (Adams, 2003).

The phylogenetic relationships between the modern invertebrate and vertebrate TSPs was studied. Initially, sequence identities in D-TSP were compared on a domain-by-domain basis. The type 3 repeats had 40% to 44% identity with the type 3 repeats of all vertebrate TSPs and the globular C terminus had 53% to 57% identity with all the vertebrate TSP C termini. This is unlike the situation for the vertebrate and Ciona TSPs, where, despite high overall conservation, the sequence identity of these regions clearly differs between the A and B subgroups. For example, the C terminus of human TSP-1 has 82% identity with that of TSP-2 and only 61% identity with the C terminus of TSP-4. The three most C-terminal type 2, EGF-like domains of D-TSP had 31% to 33% identity with the type 2 domains of TSP-1 and TSP-2, and 37% to 40% identity with the type 2 domains of TSP-3, TSP-4 and TSP-5. The amino-terminal region did not correspond to the laminin-G-like fold structure predicted for the amino-terminal domains of TSPs 1-4. In contrast, D-TSP amino-terminal domain had 47% sequence identity with the Anopheles TSP amino-terminal domain. This is in line with the documented average identity of 56% between clearly orthologous genes in these two species (Adams, 2003).

These conclusions were substantiated by a detailed phylogenetic analysis, which was performed independently on three regions of TSPs. The selected regions were the coiled-coil oligomerisation domain (39 residues), the three EGF domains preceding the TSP type 3 repeats (EGF domains 4-6; 228 residues), and the long C-terminal region that contains the seven TSP type 3 repeats and globular C terminus (474 residues) (Adams, 2003).

The phylogenetic diagrams for the three regions were rather similar in their basic features. The overall quality of the diagrams is reflected by the clear grouping of orthologs for TSPs 1-5, with the TSPs from species of more ancient origin generally found closer to the center of the trees. The tree for the EGF domains is probably the most accurate, because it includes the EGF domains from an additional insect, B. mori. In all cases, other insect sequences were closely related to D-TSP and formed a group placed centrally between the vertebrate A and B groups. For all three domains, Ciona TSP-B was placed as a separate branch between the insect TSPs and the vertebrate B group. Because Ciona TSP-A lacked a clear coiled-coil sequence, it was omitted from this analysis. For the other domains, Ciona TSP-A consistently grouped most closely to TSP-1 and TSP-2. Thus, although the length of sequence for comparison of the coiled-coil domains was of necessity short, the data obtained matched well with the results obtained with the longer domains (Adams, 2003).

In light of the new information on invertebrate TSPs and their relationship to the vertebrate TSPs, it is hypothesised that modern insect TSPs should be pentameric. This question, and other characterisations, was examined with regard to D-TSP. Multiple sequence alignment of the coiled-coil domains of D. melanogaster, A. gambiae TSPs and Ciona TSP-B with the coiled-coil oligomerisation domains of vertebrate TSPs showed that they aligned most closely with TSPs 3, 4 and 5. Furthermore, two cysteine residues were present at positions corresponding to the cysteine residues that form inter-subunit disulphide bonds in TSP-3, TSP-4 and TSP-5. These features suggested that insect TSPs might indeed oligomerise as pentamers. Intriguingly, the alignment also highlighted the fact that Ciona TSP-A lacked a clear coiled-coil sequence. It is possible that this molecule exists as a monomer or dimer (Adams, 2003).

The following two reviews deal with Thrombospondins:

Bornstein, P. (2001). Thrombospondins as matricellular modulators of cell function. J. Clin. Invest. 107(8): 929-34. Medline abstract: 11306593

Murphy-Ullrich, J. E. (2001). The de-adhesive activity of matricellular proteins: is intermediate cell adhesion an adaptive state? J. Clin. Invest. 107(7): 785-790. Medline abstract: 11285293

date revised: 10 August 2007

Home page: The Interactive Fly © 2006 Thomas Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.