Kon-tiki enhances PS2 integrin adhesion and localizes its ligand, Thrombospondin, in the myotendinous junction

Cell adhesion is essential for the development and maintenance of animal tissues. It is controlled by the regulation of the binding properties of cell surface receptors and their ligands. Moreover, altered cell adhesion is a hallmark for cancer cell progression (Sahai, 2007). Complex tissue development requires tight control of cell adhesion between different cell layers. Cells can adhere to each other directly or indirectly through the adhesion to the extracellular matrix (ECM) surrounding them (Rozario and DeSimone, 2010). This is mainly achieved by the transmembrane integrin receptors (Brown, 2000; Maartens and Brown, 2015) although it can also be mediated by other receptors such as Dystroglycan (Bozzi et al., 2009) and other proteoglycan receptors (Couchman, 2010).

Integrins are the main adhesion receptors that mediate the link between the ECM and the cytoskeleton inside the cell. They are widely expressed in different tissues and conserved in evolution, and have been shown to be essential for tissue morphogenesis and homeostasis. Integrins are heterodimeric receptors composed of an α and a β subunit that bind the ECM with their extracellular domain. They recruit many intracellular adaptors (integrin-associated proteins) via their cytoplasmic domain, and these mediate interaction with the actin cytoskeleton (Hynes, 2002). Integrin activity and regulation must be tightly controlled both in embryogenesis and in adult tissues since impaired or increased activation leads to different disorders (Legate et al., 2009; Pouwels et al., 2012). Thus, investigating the mechanisms of integrin activation and regulation are crucial to better understand integrin function. Integrins are unusual in their ability to respond both to extracellular and intracellular stimuli and thus enable bidirectional signaling. Extracellular stimuli can induce intracellular signaling (outside-in), and intracellular signaling can cause extracellular changes (inside-out). There is a lot of information about the intracellular molecules that participate in the inside-out activation of integrins; however, little is known about the molecules or mechanism by which integrins can be activated outside-in. It has been shown that growth factor receptors cooperate with integrins to mediate intracellular signaling (Ivaska and Heino, 2011). In addition, cell surface proteoglycans are cell adhesion receptors that can function as co-receptors alongside high-affinity growth factor receptors or adhesion receptors such as integrins (Couchman, 2010). For example, the heparan sulfate Syndecan can promote integrin-mediated adhesion in different cells (Beauvais et al., 2009; Couchman, 2010). In addition, the vertebrate chondroitin proteoglycan receptor CSPG4 (also known as NG2) has also been shown to be engaged in a protein complex with integrins and to promote integrin activation in different cellular systems. However, the molecular mechanisms by which cell surface proteoglycans modulate integrin activation remain unknown (Chekenya et al., 2008; Fukushi et al., 2004; Iida et al., 1998; You et al., 2014). Here, we have used the Drosophila embryonic muscle–tendon adhesion sites as a model system to study the function of the Drosophila ortholog of CSPG4, Kon-tiki (Kon, also known as Perdido) (Estrada et al., 2007; Schnorrer et al., 2007), in the context of the αPS2βPS integrin-mediated adhesion.

Muscle–tendon adhesion leads to the formation of the myotendinous junction (MTJ), which is required for the translation of the muscle contractile force into movement via tendon attachment to the cuticle. The construction of the MTJ involves the coordinated development and crosstalk between tendons and muscles (Maartens and Brown, 2015; Schweitzer et al., 2010). During MTJ formation, different cellular processes take place in a timely manner, such as cell specification, muscle migration toward tendons, and muscle–tendon recognition and attachment. Tendon cells not only serve as attachment sites but also provide guiding cues for the migrating myotube. Furthermore, tendon-specific gene expression and terminal differentiation depends on muscle attachment (Bate, 1990; Gilsohn and Volk, 2010; Martin-Bermudo, 2000; Volk, 1999). Muscle–tendon attachment is a sequential process, in which the muscle first transiently comes in close contact with the epidermal cell, thereby initiating assembly of a hemiadherens-type junction. Once they are in contact, it is thought that integrins are activated, and they trigger a strong attachment to the ECM, thus stabilizing the attachment prior to muscle contraction (Brown et al., 2000; Prokop et al., 1998; Tepass and Hartenstein, 1994; Yuan et al., 2010). In Drosophila, integrin mutant embryos contain spherical myotubes due to muscle detachment. The Drosophila integrin subunits present in the muscle attachment are encoded by the genes multiple edematous wing (mew, αPS1), inflated (if, αPS2) and myospheroid (mys, βPS) (Bökel and Brown, 2002; Brown et al., 2000). These subunits form two different heterodimers expressed at the MTJ, αPS1βPS and αPS2βPS. αPS1βPS integrin is expressed in the tendon cells and might be involved in early events of the formation of the MTJ (Estrada et al., 2007; Roote and Zusman, 1995). The αPS2βPS, which is expressed in the muscle cells, is not required for initial formation of the attachment but to form a strong muscle attachment (Wright et al., 1960). Different integrins interact with distinct types of ECM proteins: αPS1βPS interacts with laminin (Gotwals et al., 1994), whereas αPS2βPS interacts with Thrombospondin (Tsp) and Tiggrin (Bunch et al., 1998; Chanana et al., 2007; Fogerty et al., 1994; Subramanian et al., 2007). In fact, controlling the right levels of these ECM molecules is key to the formation of the MTJ (Gilsohn and Volk, 2010; Maartens and Brown, 2015; Yatsenko and Shcherbata, 2014). In particular, the role of Tsp in the MTJ has been studied in detail. Tsp is secreted from tendon cells and progressively accumulates at the junction, being essential for the biogenesis of the MTJ. In its absence, some muscles round up as a result of muscle detachment (Chanana et al., 2007; Subramanian et al., 2007). This phenotype, as well as its localization at the MTJ, is similar to that seen upon mutation of Kon.

Kon is essential for the targeting and adhesion between embryonic muscles and tendons (Estrada et al., 2007; Schnorrer et al., 2007), both in the embryo and in the adult (Perez-Moreno et al., 2014; Weitkunat et al., 2014). The molecular nature of Kon and its vertebrate ortholog CSPG4, together with some experimental data, indicates that they function as ECM receptors (Couchman, 2010; Staub et al., 2002). In addition, CSPG4 (Stallcup, 2002; Staub et al., 2002) is expressed in the sarcolemma of human postnatal skeletal muscle, as well as in regenerating myofibers (Petrini et al., 2003). In spite of this, it remains unclear what is its function in myogenesis. In Drosophila, loss-of-function of this gene results in rounded, detached muscles. Kon is expressed in muscles and localizes to muscle tips and to the muscle attachment site. Kon contains laminin globular extracellular domains and a small intracellular domain with a C-terminal PDZ-binding consensus sequence. We hypothesized that Kon primes the formation of a protein complex at the myotendinous junction that would activate a signaling pathway within the muscle that is essential for myotube guidance, recognition and attachment (Estrada et al., 2007).

Here, we find that kon and if genetically interact and can form part of the same protein complex at the muscle membrane in the embryo. Together they specifically promote cell adhesion both in cell culture and in the embryonic MTJ. In addition, Kon is required for integrin-dependent signaling at the muscle attachment, and together with the αPS2βPS integrin, Kon recruits the tendon-secreted αPS2βPS integrin ligand, Tsp, at the MTJ. We propose that Kon mediates muscle–tendon adhesion by enhancing integrin signaling and adhesion by helping to localize its ligand to the muscle membrane.

In order to understand how different cell adhesion receptors may cooperate in the development of the MTJ, we studied the genetic interactions between two receptors expressed in the muscles. These are encoded by kon and if, and are required to form a strong muscle attachment (Brown, 1994; Estrada et al., 2007; Schnorrer et al., 2007; Wright, 1960) (Fig. 1A–D). Loss of αPS2βPS integrin causes muscle detachment, although this detachment takes place after the muscles have started contracting, suggesting that αPS2βPS integrin is not involved in muscle tendon targeting (Brown, 1994; Estrada et al., 2007). Indeed, stage 16 if mutant embryos show spindle-shaped muscles but no muscle detachment (Fig. 1D). To analyze the relationship between kon and if, we performed a genetic interaction experiment. We compared the extent of muscle detachment in embryos hemizygous mutant for if (Fig. 1D), and embryos hemizygous mutant for if and heterozygous for kon (if/Y; kon/+) (Fig. 1E). We observed that if hemizygous mutants (Fig. 1D) do not present muscle detachment as expected (Brown, 1994), while if/Y; kon/+ embryos present a severe muscle detachment phenotype with many myospheres (Fig. 1E,G). In fact, these embryos present 37% more myospheres per hemisegment than kon homozygous mutant embryos (Fig. 1C,G). Moreover, we analyzed double homozygous mutant embryos for if and kon (if/Y; kon) (Fig. 1F) and observed the presence of 1.8 and 2.8 times more myospheres per segment than found in if; kon/+ or kon homozygous mutant embryos, respectively (Fig. 1G). This strong genetic interaction indicates that Kon and αPS2βPS integrin could act together in mediating the attachment of the muscle to the tendon cell, even though they may have additional independent functions.

In our previous work (Estrada et al., 2007), we assessed the genetic interaction between kon and integrins by depleting their levels through a low dose co-injection of their double-stranded RNAs (dsRNAs) – these experiments did not show a significant interaction between kon and if, as opposed to upon the co-injection of dsRNAs against kon and mew, where the interaction was statistically significant. These data, together with the in vivo visualization of embryos injected with single dsRNA for these genes, suggested that αPS1βPS integrin is required earlier during the muscle guidance process, for the formation of proper projections and muscle attachment, and that the αPS2βPS integrin may participate in muscle attachment in a different manner from the inferred αPS1βPS–Kon complex, possibly by stabilizing MTJs after they have formed (Estrada et al., 2007). In this work, we have further analyzed the relationship between Kon and αPS2βPS integrin by studying different genetic combinations of kon and if null alleles, which cause a complete loss of function of the genes, and where we found a clear genetic interaction between them (Fig. 1).

Having found that kon interacts genetically with if, we wondered whether the proteins encoded by these genes interact molecularly. To test this, we immunoprecipitated Kon from embryo extracts and tested whether the βPS subunit co-immunoprecipitated with Kon. Indeed, we observed that the βPS subunit co-immunoprecipitated with Kon in embryos (Fig. 2A). This result suggests that integrins and Kon form part of the same protein complex in the embryo.

To further study the interaction between Kon and integrins in mediating cell adhesion, we used an S2 cell aggregation assay. The Drosophila S2 cell line lacks intrinsic self-adhesive properties and cells grow individually as round, non-adherent and non-aggregating cells (Bunch and Brower, 1992; Cherbas et al., 2011) (Fig. 2C). In fact, S2 cells express very little endogenous integrin and kon transcripts. However, they can form cell aggregates if they are transfected with cell adhesion molecules (Hortsch and Bieber, 1991). We transfected S2 cells with Kon, αPS1βPS or αPS2βPS alone and then, Kon in combination with either of the two integrins, in order to quantify the formation of cell aggregates. We observed a small number of cell aggregates when cells were transfected with αPS1βPS, αPS2βPS or Kon alone, but the frequency of aggregates was significantly increased when cells were co-transfected with Kon and either αPS1βPS or αPS2βPS, being significantly higher when we co-transfected with αPS2βPS (Fig. 2B,C). These aggregates were similar to the ones found in cells transfected with the known cell adhesion protein Dumbfounded (Galletta et al., 2004; data not shown), which we used as a positive control. In addition, these aggregates mostly contain transfected cells (Fig. 2C), suggesting that the aggregates are formed by cells which co-express αPS2βPS and Kon. These results suggest that the receptors Kon and αPS2βPS specifically cooperate in mediating cell adhesion, possibly by recruiting an ECM protein that could mediate cell adhesion. This protein(s) could either be expressed by the S2 cells or be present in the culture medium.

In the embryo, Kon and αPS2βPS integrin are both present on the muscle side of the MTJ (Estrada et al., 2007; Maartens and Brown, 2015). To gain insight into the mechanism underlying the interaction between Kon and integrin, we analyzed whether integrin function was affected in the absence of Kon. To do this, we first analyzed αPS2βPS integrin localization in kon mutant embryos. We found that αPS2βPS is still localized in kon mutants, even in detached muscles (Fig. 3A–B″; Fig. S1). Next, we studied whether signaling downstream of integrins was affected in the absence of Kon. Focal adhesion kinase (Fak) is an evolutionary conserved non-receptor protein kinase involved in a myriad of cellular responses. In vertebrates, the Src–Fak complex is referred to as the major hub for integrin signaling (Martin et al., 2002). Furthermore, in the Drosophila embryo, the phosphorylation of Fak takes place at the muscle attachment site in an αPS2βPS integrin-dependent manner (Grabbe et al., 2004) (Fig. S1B–C′). Interestingly, the expression of phosphorylated Fak (P-Fak) is also increased in the vertebrate MTJ, highlighting the conservation of the pathway (Snow and Henry, 2009). We have studied the localization of P-Fak in kon mutant embryos and found that the levels of P-Fak are reduced. Since the muscle attachment is reduced in kon mutants, we quantified P-Fak levels normalized to the junctional area, and found that P-Fak levels are reduced by 39% at the muscle attachment (Fig. 3C–D″; see also Fig. 5A,D), suggesting that the αPS2βPS integrin signaling is compromised in kon embryos.

Finally, we tested a molecular interaction between Kon and P-Fak. Co-immunoprecipitation experiments performed in embryos suggest that P-Fak forms part of a protein complex together with Kon in vivo (Fig. 2A). This result further supports that Kon and integrins form part of the same protein complex in Drosophila embryos.

As we mentioned before, integrin signaling can take place in an inside-out and/or outside-in manner, and is regulated by a complex network of signals. In order to understand how Kon is mediating integrin signaling at the MTJ, we studied the function of the Kon extracellular and intracellular domains. Kon contains a small intracellular domain with a C-terminus PDZ-binding domain. PDZ-binding domains serve as a linkage to PDZ protein networks. This domain is required to bind and localize the PDZ-containing protein Grip to the muscle membrane at the muscle attachment site. This intracellular signaling through Grip is conserved and is required for muscle targeting (Estrada et al., 2007; Schnorrer et al., 2007; Stegmuller et al., 2003). In order to study the function of this domain in the muscle attachment, we constructed a Kon protein where the cytoplasmic domain of Kon was deleted (KonΔcyt), containing only the transmembrane and extracellular domains (Fig. 4A). KonΔcyt expressed under the control of the twist promoter in kon mutant embryos properly localized at the MTJ (Fig. 4C–D″). However, we found that, as expected (owing to the function of the intracellular domain in muscle targeting), it only partially rescued the muscle detachment phenotype in kon mutant embryos (Fig. 4C–E). In contrast, we observed the complete absence of muscle detachment in kon mutants expressing the full version of Kon protein (Fig. 4B,E). These results suggest that the intracellular domain is not essential for the localization of Kon to the MTJ, even though, in accordance with previous results (Estrada et al., 2007; Schnorrer et al., 2007; Stegmuller et al., 2003), it is essential for the muscle–tendon attachment.

In order to test whether the cytoplasmic domain of Kon is required to mediate αPS2βPS integrin signaling in the muscle, we analyzed whether KonΔcyt is able to rescue P-Fak expression at the MTJ in kon mutant embryos. We found that expression of KonΔcyt in the muscles of kon embryos was able to rescue P-Fak levels in a similar manner to the full version of Kon protein (Fig. 5A–D).

Taken together, these results suggest that even though the cytoplasmic domain is required to fulfill the complete Kon function, as already shown to be required for muscle–tendon targeting and interaction with the essential protein Grip (Estrada et al., 2007; Schnorrer et al., 2007), it is not essential to localize Kon to the MTJ. In addition, we show that the cytoplasmic domain is not necessary to mediate αPS2βPS integrin downstream signaling. Thus, the extracellular domain of Kon is sufficient to mediate its function in modulating the αPS2βPS integrin signaling at the MTJ.

We have observed that Kon is required for αPS2βPS integrin signaling and that Kon intracellular domain is not essential for this signaling. Thus, we asked how the extracellular domain of Kon could mediate integrin signaling and adhesion. The molecular nature of Kon and its vertebrate orthologs, together with some experimental data, indicates that they function as ECM receptors (Couchman, 2010; Staub et al., 2002). For this reason, we explored whether Kon could be regulating integrin signaling by localizing the αPS2βPS integrin ligands. Tsp, produced by tendons and essential for the biogenesis of the MTJ, has been suggested to be a ligand for αPS2βPS integrin, as well as being required for integrin-mediated MTJ formation (Chanana et al., 2007; Subramanian et al., 2007). In fact, Tsp levels are reduced in mys mutants (Subramanian et al., 2007). In addition, we have observed that Tsp mutants show a similar muscle detachment phenotype to that in kon embryos (Chanana et al., 2007; Subramanian et al., 2007) (Fig. S2B), suggesting that they could be involved in the same process. Moreover, we analyzed the levels of P-Fak in Tsp mutants compared to the controls and found that they are reduced by 27% (Fig. S2A–C). Thus, we wondered if Kon could mediate its adhesion through Tsp at the MTJ. We first studied the localization of Tsp in kon mutants and found that it is reduced compared to the controls (Fig. 6A–B″,E). Since Tsp has been proposed to be a ligand for the αPS2βPS integrin, it is not completely absent in kon mutants, and its localization is affected in mys mutant embryos, we wondered whether the αPS2βPS integrin was also responsible for its localized expression at the MTJ. For this, we quantified the levels of Tsp in if mutant embryos. Indeed, we found that Tsp levels were also reduced in if mutants compared to the controls (Fig. 6C–C″,E). The quantification of Tsp levels showed that they were significantly more reduced in kon mutants (65%) than in if mutants (58%) (Fig. 6E). Moreover, Tsp levels were reduced by 76% in if; kon double mutants, compared to the controls (Fig. 6D,E). This significant reduction of Tsp in if; kon double mutants suggests that kon and if are both required to localize Tsp at the MTJ, and thus cooperate in recruiting Tsp.

We also studied whether the overexpression of Kon in wild-type muscles enhanced the accumulation of Tsp at the MTJ, and found that there are no differences in the levels of Tsp between embryos where Kon is overexpressed compared to control embryos (Fig. S3), also suggesting that Kon is not enough to recruit Tsp to the MTJ and that it needs the αPS2βPS receptor.

We then studied whether the Kon extracellular domain had a role in the localization of Tsp at the MTJ. To do this, we quantified the levels of Tsp in control embryos, kon mutant embryos and kon embryos where we expressed KonΔcyt or full-length Kon protein in the muscles. We found that the expression of the Kon extracellular domain partially rescued the levels of Tsp in kon mutants, similar what occurs upon expression of the full-length Kon protein (Fig. 7A–E). These results suggest that Kon is required to recruit Tsp at the MTJ through its extracellular domain.

Finally, in order to test whether Kon was sufficient to localize Tsp at the MTJ, we expressed Kon ectopically in the tendons of kon mutant embryos. We observed that Kon partially rescues Tsp localization at the junction in these embryos (Fig. 8B–B″; compare with Fig. 6A–A″ and Fig. S4A), suggesting that Kon is also able to recruit Tsp at the MTJ from the tendon cell. As we have proposed that Kon could be regulating αPS2βPS integrin signaling by localizing its ligand, we studied whether Kon expression in the tendons of kon mutants could also rescue the levels of P-Fak. We observed that P-Fak levels are restored in this genotype (Fig. 8C–D″; Fig. S4B). Thus, the expression of Kon in the tendon is sufficient to localize the αPS2βPS ligand Tsp and restore the αPS2βPS integrin signaling.

Moreover, expressing Kon in the tendons of kon mutant embryos also helped us to study the cis and trans requirements of Kon function. In fact, the expression of Kon in the tendons of kon mutant embryos rescues around half of the muscle detachment observed in kon embryos (Fig. 8E), suggesting that the localization of Tsp and P-Fak contributes to the function of Kon in the formation of the MTJ. In addition, we observed that the expression of KonΔcyt in the tendons of kon mutant embryos rescues the muscle detachment in kon embryos in a similar way to expression of full-length Kon (Fig. 8E), suggesting that the extracellular domain of Kon is key for Kon function.

Transmembrane proteoglycans can associate with other receptors, and they function as co-receptors for growth factor and cell adhesion receptors, such as integrins, thereby, affecting their function (Couchman, 2010). Understanding how integrins and transmembrane proteoglycans are regulated and coordinated is largely unknown and is key to understanding cell adhesion in development and disease. Here, we find that the transmembrane receptor Kon cooperates with the αPS2βPS integrin to mediate cell adhesion both in cultured cells and in the Drosophila embryo MTJ. We find that embryos lacking both proteins present a stronger muscle detachment phenotype than embryos lacking either Kon or the αPS2βPS integrin alone, indicating that Kon and αPS2βPS in embryonic muscles cooperate to mediate adhesion to tendon cells. Moreover, we propose a mechanism by which this transmembrane proteoglycan can enhance integrin signaling and adhesion in the development of the MTJ, by helping to localize the integrin ligand.

The vertebrate ortholog of Kon, CSPG4, has been shown to be engaged in a protein complex with integrins and promote integrin activation in different cellular systems. For example, CSPG4 can bind directly to α4β1 integrin and enhance integrin-mediated adhesion in melanoma cells (Iida et al., 1998; Chekenya et al., 2008; Fukushi et al., 2004; You et al., 2014). How does Kon modulate αPS2βPS integrin signaling? One possibility is that Kon helps to localize integrins at the MTJ. In fact, it has been shown that Kon helps to localize integrins at the MTJ in adult flight muscles (Weitkunat et al., 2014), although this is not the case in adult abdominal muscles (Perez-Moreno et al., 2014). Here, we have found that Kon does not regulate αPS2βPS integrin localization at the MTJ in Drosophila embryos. However, we find that co-immunoprecipitation experiments suggest that Kon forms part of the same protein complex with the βPS subunit in the embryo. Although we have only been able to detect co-immunoprecipitation of Kon with the βPS subunit and not αPS2, we would like to suggest that Kon forms a complex with the αPS2βPS integrin, and not αPS1βPS, because they are both expressed in the muscle membrane. In addition, we show that Kon and P-Fak, which is downstream of αPS2βPS integrin, form part of the same protein complex. This interaction is probably indirect through integrins, since the expression of a KonΔcyt rescues P-Fak localization in kon mutant embryos. These results suggest that the extracellular domain of Kon is key in modulating integrin-mediated signaling. In addition, our results showing that the expression of KonΔcyt in the muscles of kon mutant embryos only partially rescues the muscle detachment caused by the lack of Kon, suggest that the intracellular domain of Kon, while being dispensable for integrin signaling, is required for Kon-mediated MTJ formation. As this domain has been shown to interact with the PDZ protein Grip, which is required to mediate muscle–tendon recognition in the embryo (Estrada et al., 2007; Schnorrer et al., 2007; Swan et al., 2004), we would like to propose that the intracellular domain of Kon might mediate the earlier events of muscle guidance and targeting to the tendon cells, but not the later requirement of Kon function, enhancement of integrin signaling. In fact, Kon lacks intrinsic enzymatic activity. Thus, we suggest that it modulates integrin signaling indirectly through the extracellular domain by regulating the localization or the activity of other proteins, which in turn may modulate integrin function.

Kon is an ECM receptor and its orthologs in vertebrates are involved in the recruitment of ECM components (Burg et al., 1996; Staub et al., 2002). Among the αPS2βPS ligands, Tsp has been shown to bind directly to αPS2βPS integrin in Drosophila S2 cells (Subramanian et al., 2007). Furthermore, here we show that Tsp levels are also reduced in if mutants, and that Kon is necessary to properly localize Tsp at the Drosophila MTJ. Moreover, the ectopic expression of Kon in tendon cells in kon mutants restores the localization of Tsp at the MTJ, further suggesting the essential role of Kon in localizing Tsp. But how could Kon be recruiting Tsp to the MTJ? Tsps are secreted multimeric, multidomain glycoproteins that function at cell surfaces and in the ECM, acting as regulators of cell interactions and attachments (Adams, 2001). The different conserved domains within Tsps are involved in interactions with other ECM molecules, and transmembrane proteoglycans, such as integrins, and heparan sulfate proteoglycans (Bentley and Adams, 2010). Even though, in vertebrates, Tsp has been shown to bind the heparan sulfate syndecan (Adams and Lawler, 2004), in Drosophila, the heparan sulfate proteoglycan expressed in the muscles, Syndecan, does not mediate this interaction (Subramanian et al., 2007). Our studies suggest that the chondroitin sulfate proteoglycan Kon may be the adhesion receptor that localizes, together with αPS2βPS integrin, the secreted ECM protein Tsp to the MTJ. Kon could interact directly with Tsp through a Tsp GAG-binding domain, or indirectly through another ECM molecule. In the context of Kon interaction with the ECM, we also show that the presence of Kon and the αPS2βPS integrin in S2 cells significantly enhances cell adhesion in a cell aggregation assay. In this case, Kon and the αPS2βPS integrin could recruit some ECM protein expressed by S2 cells or a protein from the culture medium to mediate adhesion. This recruitment would be similar to the function of Kon and αPS2βPS integrin in the embryonic MTJ. Although the nature of these proteins is unknown, Collagen IV, Perlecan, Laminin and Sparc are some candidates expressed by S2 cells (Cherbas et al., 2011). Moreover, it is unlikely that Tsp mediates this adhesion in the S2 cell aggregates since it is not expressed in S2 cells (Cherbas et al., 2011 and FlyBase).

Thus, we propose that the extracellular domain of Kon modulates integrin signaling by helping to localize Tsp to the MTJ. Kon could localize Tsp and thus increase its availability to the αPS2βPS integrin, which would enhance downstream integrin signaling and adhesion. This is supported by the fact that some ECM proteins, including Tsp (Adams and Watt, 1993; Asch et al., 1991; Sun et al., 1992), present both proteoglycan- and integrin-binding sites thereby promoting the formation of adhesion receptor clusters. These interactions work co-operatively to support stable cell attachments (Iida et al., 1998; Adams, 2001). Alternatively, Kon could help to localize Tsp by interacting with αPS2βPS integrin leading to integrin clustering and/or an increment of integrin ligand affinity (but not localizing the αPS2βPS integrin), which would help to localize Tsp and then enhance integrin signaling and adhesion. Our co-immunoprecipitation data together with experiments with the vertebrate orthologs (Iida et al., 1998) also support this possibility (Fig. 8F).

Finally, the fact that the expression of Kon in tendon cells in kon mutants is able to localize Tsp and P-Fak at the MTJ, and can partially rescue the muscle detachment, suggest that the recruitment of Tsp and the restoration of integrin signaling are necessary for Kon function. This role of Kon in localizing Tsp and P-Fak to the MTJ would lead to increased adhesion and consolidation of the MTJ. This functional interaction between Kon and αPS2βPS integrin builds a specialized ECM microenvironment, which is essential for the development of the MTJ. Ultimately, these studies could be useful to design new drugs that prevent CSPG4-promoted integrin-dependent chemoresistance in tumor cell survival (Chekenya et al., 2008).

The following stocks were used (all from Bloomington Stock Center unless stated otherwise): strain y¹w118 as wild type; strain perd^F1-3 as kon mutants, perd^F1-3/CTG (Estrada et al., 2007); strain perd^F1-3/CyO, ftz-lacZ (Estrada et al., 2007); strain if^B2/FTG; strain twist-GAL4; strain UAS-HA-kon-tiki (UAS-HA-kon) (Schnorrer et al., 2007); strain UAS-HA-KonΔcyt (this work); strain Tsp^8R/CyoYFP (Subramanian et al., 2007). MD710 (stripe-GAL4) (Calleja et al., 1996). The FTG, CTG and TTG balancer chromosomes, carrying twist-Gal4 UAS-2EGFP, were used to identify homozygous mutants (Halfon et al., 2002).

Drosophila S2 cells (from the Drosophila Genomics Resource Center) were grown and transfected as published previously (Bunch and Brower, 1992). Transfections were performed with FuGene (Invitrogen) following the manufacturer's instructions. Transfection efficiencies were ∼30%. Aggregation assays were performed as previously described (Hortsch and Bieber, 1991). Cells were transfected with plasmids encoding Actin-Gal4 and UAS-HA-Kon-tiki (Schnorrer et al., 2007), and plasmids pHSPSβ (mys cDNA under an Hsp70 promoter) (Bunch and Brower, 1992), pHSPS1 (mew cDNA under an Hsp70 promoter) (Bunch and Brower, 1992), pHSPS2 (if cDNA under an Hsp70 promoter) (Bunch and Brower, 1992) or UAS Duf-HA (Galletta et al., 2004). Positively transfected cells were identified by specific antibodies against HA (1:500, cat. no. 11867423001; Roche), PS2 integrin-specific antibody CF.2C7 (1:100; Hybridoma Bank; Brower et al., 1984), PS1 integrin-specific antibody DK1A4 (1:100; Hybridoma Bank; Brower et al., 1984), βPS1 integrin-specific antibody CF.6G11-c (1:100; Hybridoma Bank; Brower et al., 1984). Each experiment consisted of the simultaneous transfection of S2 cells with all the different proteins to be compared in a six-well plate. Similar numbers of S2 cells were placed in each transfection plate. Experiments were run in triplicate, with two independent scorings per experiment, for a total of six observations per transfection. Large cell aggregates (composed of 20 or more cells) were quantified from the differently transfected S2 cells (n=6). We tested whether cell cultures with different transfections differed in the number of cell aggregates they contained by means of generalized linear models with an underlying Poisson error distribution and a log link function using SAS 9.2 (SAS Institute, Cary, NC). Post hoc tests were carried out testing for differences among the least square means and applying a Sidak correction of the observed P-values.

For in vivo co-immunoprecipitations, embryonic lysates were prepared from eight collections of stage 16 w¹¹¹⁸ embryos. Co-immunoprecipitations were carried out as described previously (Slovakova and Carmena, 2011) except that the lysis buffer was: 50 mM Tris-HCl pH 8, 150 mM NaCl, 1% NP40, 0.3% DOC, 50 mg/ml PMSF and Complete protease inhibitors (Roche). Two immunoprecipitations were performed in parallel, one with rabbit anti-Kon (1:2000; Schnorrer et al., 2007) and one with rabbit anti-β-galactosidase (cat. no. 55976; Cappel) as a control. The same amount of protein extract was loaded on the beads in each immunoprecipitation. Immunoblotting was performed with rabbit anti-βPS 185A-E (1:1000; Gotwals et al., 1994) and rabbit anti-P-Fak (1:200, cat. no. 44-642; Biosource).

Embryo antibody stainings were carried out as described previously (Carmena et al., 1998). All embryos were at stage 16. The following primary antibodies were used: rabbit and mouse anti-myosin heavy chain (MHC) (1:10, Dan Kiehart, Dept. of Biology, Duke University, USA), rat anti-MHC (1:400, cat. no. BT-GB-147S; Babraham Bioscience Technologies), rabbit and mouse anti-β-galactosidase (rabbit anti-β-galactosidase, 1:10000, cat. no. 55976; Cappel; mouse anti-β-galactosidase, 1:10000, cat. no. Z378A; Promega), rabbit and mouse anti-GFP (1:5000, cat. no. A-6455 and A-11120, respectively; Invitrogen), chicken anti-GFP (1:1500, cat. no. 13970; AbCam), rabbit anti-β-tubulin (1:3000; Leiss et al., 1988), rat anti-HA (1:400, cat. no. 11867423001; Roche), rat anti-Tsp (1:200; Subramanian et al., 2007), rabbit anti-P-Fak (1:1000, cat. no. 44-642; Biosource), rat anti-Tropomyosin (Babraham Bioscience Technologies), rat anti-PS2 (1:100; Bogaert et al., 1987). We tested for differences among genotypes in the number of myospheres observed per hemisegment by fitting a linear model since the dependent variable met all parametric assumptions. Confocal images were obtained using a Leica SP2 microscope and processed with Adobe Illustrator and ImageJ.

The construction of KonΔcyt was performed as follows. DNA from plasmid pFS135 (Schnorrer et al., 2007) containing the complete Kon cDNA was PCR amplified using primers Kon-SgrA (5′-GTCACGCCGGCGTAATCTAGAGGATCTTTGTGAAG-3′) and UAS-StuI (5′-GTCAAGGCCTCCCGGGTCTAGTGGATCCAG-3′). Primer Kon-SgrA introduces a stop codon before a SgrA1 site and eliminates 735 bp from the cytoplasmic domain of Kon. The amplified DNA fragment and the pFS135 plasmid were digested with SgrA1 and StuI for oriented cloning. Ligation mixtures were used to transform E. coli DH5α, with selection of Apr transformants on LB medium plates containing ampicillin. This plasmid was used to obtain the transgenic flies.

Quantification of muscle detachment was performed by counting the number of rounded muscles or myospheres in stage 16 embryonic ventral muscles. Quantification of fluorescence intensity was performed from maximal projections confocal stacks using ImageJ software. The quantification was performed by measuring the mean gray value on manually selected regions normalized to the area of the junction. Mean gray value is the sum of the gray values of all the pixels in the selection divided by the number of pixels. The regions were selected with ImageJ software and were muscle attachment sites in the ventral longitudinal muscles. The background value, taken from intersegmental signal-free regions, was subtracted from the mean gray value in each embryo. The quantifications data was represented as in each case in a box plot, where center lines show the medians, box limits indicate the 25th and 75th percentiles as determined by R software, whiskers extend to 1.5 times the interquartile range from the 25th and 75th percentiles and outliers are represented by dots (Spitzer et al., 2014). All statistical analyses were conducted in R (R Core Development Team). We statistically compared differences in the variables of interest (number of myospheres or fluorescence intensity) among the different genotypes by fitting general linear models using the lm function. Residuals were systematically checked for goodness of fit to a Gaussian error distribution using visual assessments of normal quantile plots and Kolmogorov–Smirnov tests (function lillie.test from package nortest). We also tested for heteroskedasticity of the data using the Breusch–Pagan test with function bptest from package lmtest. In most cases, variables met parametric assumptions, and when they did not, we fit the linear model on the ranked variable. Multiple comparisons were conducted by means of post-hoc Tukey tests using the TukeyHSD function. Asterisks over the graphs indicate significant differences, while n.s. indicates results did not show significant differences.

We thank the Bloomington Stock Center and the Vienna Stock Center for fly stocks, the Developmental Hybridoma Bank for antibodies. T. Bunch, R. Hynes, S. Abmayr, T. Volk, F. Schnorrer, D. Kiehart, and Gerd Vorbrüggen for flies and antibodies, Ivan Gómez-Mestre for help with the statistical analysis, M. D. Martín-Bermudo, A. E. Rosales-Nieves, C. S. Lopes, and A. González-Reyes for helpful comments on the manuscript, Laura Tomás from the CABD proteomics facility, and the Drosophila Genomics Resource Center, supported by The National Institutes of Health (NIH) (grant 2P40OD010949-10A1). We also thank M. D. Martín-Bermudo in whose laboratory most of the experiments were performed, and C. J. O'Kane, in whose laboratory some final experiments were performed.