DE-Cadherin regulates unconventional Myosin ID and Myosin IC in Drosophila left-right asymmetry establishment

L/R patterning is required for the functional organization of the body plan and the L/R asymmetric development of visceral organs, such as the heart, liver or intestine in human. To date, two distinct molecular mechanisms responsible for symmetry breaking have been identified in vertebrates (for a review, see Levin, 2006; Levin and Palmer, 2007; Speder et al., 2007). The earliest known event is the left-sided expression of ion pumps in the early Xenopus embryo generating an asymmetric gap junction-mediated ion flow, leading to a directional transport of putative L/R determinants through the induced polarized electric field (Levin et al., 2002). The second mechanism operates during gastrulation and relies on tilted motile cilia localized at the embryonic node, which create a leftward fluid flow over the midline of the body leading to unidirectional transport and polarized accumulation of L/R determinants (for a review, see Raya and Belmonte, 2006). Propagation and maintenance of the initial asymmetric L/R signal is ensured by the expression of an asymmetric gene cascade, including the left-sided expression of Nodal and its downstream target genes Pitx2 and Lefty (for reviews, see Raya and Belmonte, 2006; Tabin, 2006).

Most of the defects associated with aberrant L/R patterning lead to a randomization of L/R asymmetric traits (situs ambiguus), causing severe congenital disorders and embryonic lethality (Aylsworth, 2001). By contrast, rare but highly instrumental situs inversus mutations lead to a complete inversion of the L/R axis. To date, only two situs inversus genes have been molecularly identified in animal models: the inversin (Inv; Invs – Mouse Genome Informatics) gene in mouse (Yokoyama et al., 1993) and, more recently, myosin ID (myoID; Myo31DF – FlyBase) in Drosophila melanogaster (Hozumi et al., 2006; Speder et al., 2006). MyoID has been shown to act as an L/R determinant responsible for the clockwise (dextral) rotation of genitalia in male flies as well as the stereotyped looping of tubular organs, such as the embryonic gut, spermiduct and testis (Fig. 1A) (Ádám et al., 2003; Coutelis et al., 2008; Hozumi et al., 2006; Speder et al., 2006; Speder and Noselli, 2007; Speder et al., 2007; Suzanne, 2010). Loss of MyoID function causes a situs inversus phenotype characterized by counter-clockwise (sinistral) genitalia rotation and inverted organ looping. MyoID is required functionally in a critical time-window of 3 hours (Fig. 1A), one day before genitalia rotation occurs (Speder et al., 2006). The myoID gene encodes an unconventional type ID myosin: a one-headed, monomeric actin-based motor, bearing an N-terminal motor head domain, a neck domain with two IQ motifs and a short basic tail domain (Mooseker and Cheney, 1995).

A first hint towards the cellular function of MyoID came from the discovery of a physical interaction between this protein and the adherens junction component β-Catenin (Speder et al., 2006). The MyoID tail domain binds to β-Catenin in vitro and both proteins colocalize in vivo at the adherens junctions in the A8 segment of the male genital disc, the L/R organizer for genitalia (Fig. 1A). Adherens junctions are adhesive cell-cell contacts and signalling platforms, localizing apically in epithelial cells (Miyoshi and Takai, 2008). Their core component is the dimeric Ca²⁺-dependent transmembrane protein E-Cadherin, establishing cell adhesion through extracellular domain binding of homodimers at the apical surfaces of adjacent cells (Niessen and Gottardi, 2008).

Here, we investigate the role of adherens junctions in L/R asymmetry establishment in Drosophila. Using targeted genetic invalidation and biochemical approaches, we show that DE-Cadherin (Shotgun – FlyBase) serves as a signalling platform for L/R asymmetry and promotes its implementation through MyoID direct binding and the additional repression of the anti-dextral activity of MyoIC (Myo61F – FlyBase), a closely related unconventional type I myosin. We show that MyoIC antagonizes MyoID and that these myosins have mutually exclusive localization domains at the cell cortex. Altogether, we describe a novel regulatory network of L/R asymmetry establishment through the adherens junction-dependent regulation of two type I myosins, one activating (MyoID) and the other inhibitory (MyoIC).

The following mutant strains were used: myoID^K2 is a myoID null allele (Speder et al., 2006); Df(3L)920 is a deficiency covering the myoIC locus (Hegan et al., 2007). Expression of transgenes was carried out using the bipartite Gal4-UAS system (Brand and Perrimon, 1993). The following transgenic lines were used: UAS::DE-cadherin^RNAi (#103962 from Vienna Drosophila RNAi Stock Center); UAS::DE-Cadherin, UAS::DE-Cadherin^DN (a gift from M. Mlodzik, Mount Sinai School of Medicine, NY, USA). The UAS::myoID^RNAi, UAS::myoIC^RNAi, UAS::MyoIC, UAS::MyoID, UAS::MyoID-GFP and UAS::MyoIC-RFP constructs were generated following standard procedures (Speder et al., 2006) (this study). The following Gal4 driver lines were used: Abd-B::Gal4^LDN (de Navas et al., 2006), myoID::Gal4 (NP1548, National Institute of Genetics Stock Center), hh::Gal4, tsh::Gal4, arm::Gal4 (Bloomington Stock Center). All crosses were performed at 30°C unless stated otherwise.

To improve readability, the Gal4 driver lines were named after the gene used for Gal4 expression specificity and the ‘::Gal4’ was substituted for ‘>’. Similarly, UAS constructs were simplified to the sequence expressed and placed after the ‘>’ sign. For example, UAS::MyoID-GFP; Abd-B::Gal4^LDN becomes Abd-B>MyoID-GFP.

We raised new polyclonal antibodies against MyoIC and MyoID. For the generation of anti-MyoIC antibodies, we used for immunization a GST-fusion protein containing the MyoIC-tail residues 765-1026 (PB isoform), and followed the procedure described in Speder et al. (Speder et al., 2006). An anti-MyoID rat polyclonal antibody was generated by Eurogentec against the two following peptides: residues 987-1001 and 824-838. We also used a rabbit polyclonal anti-MyoID, described in Speder et al. (Speder et al., 2006). These antibodies were all used at a 1/50 dilution. Other primary antibodies were polyclonal rat anti-DE-Cadherin (1/50), polyclonal mouse anti-β-catenin (1/50), polyclonal mouse anti-Dlg (1/500) and anti-En (1/100) and were from Developmental Studies Hybridoma Bank. Fluorescent secondary antibodies were Alexa 488, Alexa 546 or Cy5 633 goat anti-rat, anti-rabbit or anti-mouse from Invitrogen (1/400), and phalloidin-TRITC or phalloidin-FITC FluoProbes from Sigma (0.2 μg/ml). We carried out immunostaining according to standard procedures (Speder et al., 2006). For each genotype, 15-20 discs were analysed. Images were captured on a Zeiss LSM 510 META confocal microscope using 40× and 63× objectives. Picture processing was carried out in Adobe Photoshop 7.0 and CS4.

The experiments were carried out as described in Speder et al. (Speder et al., 2006) by crossing the tub::Gal80^ts; Abd-B::Gal4^LDN and UAS::MyoIC lines or the myoID::Gal4; tub::Gal80^ts and UAS::DE-cadherin^RNAi lines. Temperature shifts were performed from 25°C to 29°C and from 29°C to 25°C in a temperature-controlled water bath.

myoID^RNAi flip-out clones were generated by crossing the hs-flp; +; UAS::myoID^RNAi / SM5-TM6B to act[FRT]CD8[FRT]Gal4, UAS::GFPnls lines. Heat shock was applied for 10 minutes at 34°C at 72-96 hours after egg-laying.

One hundred genital discs of each genotype were dissected and lysed in RIPA buffer (1% Triton X-100, 150 mM NaCl, 50 mM Tris-HCl, pH 8) by vortexing. After addition of Laemmli buffer (1×) followed by 5 minutes at 96°C, the whole protein extract was charged on a 10% SDS-PAGE gel. Separated proteins were transferred onto Optitran BA-S 85 reinforced nitrocellulose membranes (Whatman) and detected using HRP Substrate Peroxide Solution (Millipore). Proteins were detected using the following primary antibodies: rat anti-MyoID (1:50), rabbit anti-MyoIC (1:500), rat anti-DE-Cadherin (1:500), polyclonal mouse anti-β-catenin (1:200). Secondary antibodies were: rat, rabbit or mouse anti-HRP from GE Healthcare (1:10,000). For membrane stripping, blots were incubated for 1 hour at 37°C with constant shaking in stripping buffer (0.2 M glycine, 0.1% SDS, 1% Tween, pH 2.2).

Mid-third instar larvae expressing MyoIC-RFP or MyoID-GFP under the control of arm::Gal4 driver were lysed in lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 1 mM MgCl₂). Larval extracts (1 mg of protein) were incubated overnight at 4°C with 6 μg of rabbit anti-RFP (Rockland) or rabbit anti-GFP (Sigma-Aldrich) antibodies. The antigen–antibody complex was precipitated with 80 μl of a 50% slurry of proteinA/proteinG Sepharose beads and incubated for 2 hours at room temperature. The beads were recovered by centrifugation for 1 minute at 400 g and washed four times with lysis buffer. Samples were denatured for 5 minutes at 75°C and loaded onto NuPAGE Novex gel (4%-12% Tris-Glycine, Invitrogen). Enhanced chemiluminescence was used for antibody detection of DE-Cadherin and β-Catenin after blotting on Immobilon-P^SQ membrane (Millipore).

MyoID plays a major role in L/R determination in Drosophila (Hozumi et al., 2006; Speder et al., 2006). To test whether other myosins are involved in L/R asymmetry establishment, we systematically expressed RNAi and available overexpression constructs against all known Drosophila myosins in the A8 segment of the male genital disc, the genitalia L/R organizer (Table 1), using the UAS-/Gal4 targeting system (Brand and Perrimon, 1993). We found that overexpression of MyoIC leads to a strong situs inversus phenotype, with flies showing sinistral genitalia rotation (Table 1; Table 2, row 4), instead of the wild-type dextral rotation. Interestingly, MyoIC is also a type I unconventional myosin and the closest homologue of MyoID. It was shown previously that upon MyoIC overexpression the direction of embryonic gut looping, an early marker of L/R asymmetry, is inverted (Hozumi et al., 2008; Hozumi et al., 2006). However, depletion of MyoIC activity in the L/R organizer did not affect the rotation of the genitalia which remained wild type (i.e. dextral) (Table 1; Table 2, row 3). We could not observe L/R axis inversion for any of the other Drosophila myosins. In some cases, however, their RNAi-mediated silencing led to mild rotation defects without affecting directionality, thus probably affecting the mechanical aspect of the rotation process (for details, see Table 1). These results indicate that only type I myosins have a specific role in L/R determination in Drosophila.

Two models can explain the sinistral phenotype caused by MyoIC overexpression: (1) MyoIC might act as a sinistral determinant, overriding the dextral activity of MyoID, or (2) MyoIC might act as an anti-dextral factor, inhibiting MyoID activity. If MyoIC harbours a sinistral activity, then the simultaneous loss of dextral (MyoID) and potentially sinistral (MyoIC) functions should lead to a ‘no-rotation’ phenotype. To test whether MyoIC is a sinistral determinant, we generated double mutant flies lacking both MyoID and MyoIC activities. In these flies, we observe a clear sinistral phenotype (Table 2, rows 5, 6), which is identical to the loss of MyoID alone. This result indicates that MyoIC does not bear an independent sinistral activity, but rather acts as a negative regulator of MyoID activity and, therefore, exerts an anti-dextral function.

To gain better insights into the MyoIC-MyoID interaction, we generated antibodies against each protein (see Materials and methods) and analysed their expression pattern in wild-type discs. MyoIC was expressed in the entire male genital disc and was enriched in the A8 segment (Fig. 1C), overlapping with the A8-specific expression of MyoID (Fig. 1B). We performed co-immunostaining analysis to determine the intracellular localization of both MyoID and MyoIC. Although both myosins were detected in the same cells, careful analysis revealed that at the cellular level MyoID and MyoIC have mutually exclusive cortical localization patterns (Fig. 1D).

In an effort to characterize the anti-dextral activity of MyoIC, we analysed the intracellular localization of both myosins in a MyoIC gain-of-function condition and compared it with the wild-type situation (Fig. 2A). We observed a fully penetrant reduction of MyoID signal (Fig. 2B), which is consistent with the sinistral phenotype of these flies and indicates a loss of MyoID activity. Reciprocally, MyoID overexpression induced a decrease of MyoIC signal (Fig. 2C). The wild-type dextral phenotype of these MyoID overexpressing flies is consistent with the wild-type dextral phenotype of MyoIC loss-of-function flies (see above; Table 2, row 3). From these data, we conclude that MyoIC and MyoID are mutually inhibiting each other’s localization to specific cortical domains.

To test whether MyoID expression levels could counteract the dominant effect of MyoIC overexpression, MyoID and MyoIC were co-overexpressed in the A8 segment. In this context, we observed a wild-type dextral rotation phenotype. A dilution effect of the Gal4 activator was ruled out using a UAS::lacZ control overexpressed with either myosin (Table 2, rows 8-11). These results thus indicate that MyoID overexpression is able to rescue MyoIC overexpression (Table 2, row 7). Accordingly, we detected a colocalization of MyoID and MyoIC at the membrane and in the cytoplasm (Fig. 2D), indicating that MyoID is no longer excluded by MyoIC and reoccupies its wild-type cortical domain. These results show that the observed MyoIC-induced situs inversus phenotype is linked to the loss of MyoID localization and activity in the A8 segment cells. Furthermore, it shows the importance of the MyoID-MyoIC protein balance for proper L/R asymmetry establishment.

To gain a further understanding of the molecular nature of the mutual inhibition between MyoIC and MyoID, we investigated whether the protein levels of either myosin were changed upon overexpression of the other myosin. To address this question, we performed western blot analysis on dissected genital discs in which the myosin protein or RNAi overexpression is limited to the A8 segment. The control lanes show that MyoID or MyoIC protein levels were increased upon overexpression, thus validating the approach (Fig. 3A, lane 3, upper panel and lane 5, middle panel). The differences of endogenous expression domains of MyoID (restricted to the A8 segment) and MyoIC (expressed in all genitalia segments A8, A9 and A10) probably account for the 45-fold change of MyoID and only fourfold change of MyoIC observed upon overexpression. Surprisingly, we found that MyoID protein concentration remains unchanged upon MyoIC overexpression (Fig. 3A, lane 5, upper panel). Reciprocally, upon MyoID overexpression MyoIC protein level was only slightly altered (Fig. 3A, lane 3, middle panel).

Therefore, neither myosin is degraded upon overexpression of the other myosin, although they became undetectable by immunohistochemistry (Fig. 2B,C). The loss or reduction of antigen detection could be due to, first, a conformational change of the protein or concealment of the epitopes in a protein complex, or, second, the displacement of the protein from the membrane into the cytoplasm, leading to its dilution. Therefore, MyoID function assessment should not be limited to its detectability by immunofluorescence but rather on the genitalia rotation phenotype.

To characterize further the interaction between MyoIC and MyoID, we analysed the impact of the loss of one of the myosins on the expression and localization of the other. In MyoIC loss of function, we observed a decreased MyoID signal (Fig. 2E) and a concomitant twofold reduction of MyoID protein by western blot (Fig. 3A, lane 4). However, the wild-type dextral phenotype seen in this condition (Table 2, row 3) suggests that half a dose of MyoID protein is sufficient for its function (MyoID/MyoIC ratio>1). Nevertheless, MyoID intracellular localization and/or organization are changed, as suggested by the reduced antibody detection (Fig. 2E).

By contrast, MyoID loss of function had no strong effect on MyoIC expression, as neither MyoIC localization nor protein level was altered (Fig. 2F and Fig. 3A, lane 2, middle panel). This result was confirmed cell-autonomously with myoID^RNAi-expressing clones (Fig. 2G).

In summary, MyoID remains active upon MyoIC loss of function and becomes inactive when MyoIC protein levels are increased or, more generally, when MyoIC levels are much above the levels of MyoID (MyoID/MyoIC ration <<1). In wild-type flies, MyoID is able to act and to promote dextral determination in spite of MyoIC presence. MyoID is not equally able to affect MyoIC localization and expression. These results confirm the anti-dextral function of MyoIC as an antagonist of MyoID.

MyoID has been shown to interact physically and to colocalize with the adherens junction component β-Catenin (Speder et al., 2006). We thus investigated whether adherens junctions could act as a regulatory signalling platform for MyoID function during L/R asymmetry establishment. To test this hypothesis, we first silenced the adherens junction component DE-Cadherin specifically in the A8 segment. This depletion resulted in a highly penetrant no-rotation (0°) and close-to-no-rotation (1-90°) phenotype (Table 2, row 13). Interestingly, immunofluorescence analysis revealed a concomitant loss of MyoID signal in this context (Fig. 4B), consistent with the no-rotation phenotype and a loss of MyoID activity. In stronger conditions of DE-Cadherin depletion, a close to fully penetrant no-rotation phenotype was observed, suggesting that no-rotation could represent the end phenotype of a full DE-cadherin loss of function.

Furthermore, we also observed an increased MyoIC signal upon DE-cadherin silencing (Fig. 4B), suggesting that DE-Cadherin negatively regulates MyoIC. Similar results were obtained upon overexpression of a dominant-negative form of DE-Cadherin (DE-Cad.^DN) in the posterior compartment of all the genital disc segments (A8, A9, A10), indicating that the regulation of MyoIC by DE-Cadherin is not specific to the A8 segment (Fig. 4C). These results show that a reduction in DE-Cadherin leads to an increase in MyoIC expression and a concomitant reduction in MyoID immunodetection.

Strikingly, the effect of DE-Cadherin reduction led to a decrease in MyoID signal in the A8 segment (Fig. 4B) and a loss of MyoID activity leading to rotation defects (Table 2, rows 13, 14). Furthermore, upon DE-Cadherin reduction we observed an increased MyoIC signal by immunohistochemistry and an increase in MyoIC protein level by western blot (Fig. 3A, lane 6, middle panel). In this context, MyoID protein level is also increased, suggesting that DE-cadherin controls the protein level of both myosins.

To test whether the increase in MyoIC level impacts on the rotation defects of DE-Cadherin depletion, we examined the effect of the absence of both DE-Cadherin and MyoIC on MyoID and on genitalia rotation. In the double DE-cadherin, myoIC mutant context, the rotation defects are shifted towards a more wild-type phenotype, not due to the mere Gal4 dilution (Fig. 4E; Table 2, rows 14-17). The level of MyoID protein was also slightly reduced, consistent with MyoIC reduction as previously observed (Fig. 3, lane 4). However, MyoID remained higher than normal upon DE-cadherin silencing, suggesting that DE-Cadherin regulates MyoID protein level through an unknown and MyoIC-independent mechanism. Additionally, MyoID was not detected by immunohistochemistry (Fig. 4E) as in the myoIC mutant context (Fig. 2E). These results suggest that both the depletion of DE-Cadherin and the increase in MyoIC contribute to the final genitalia rotation defects in the DE-Cadherin depletion context. Thus, DE-Cadherin appears to have two effects on L/R asymmetry: first, promoting MyoID activity and second, repressing MyoIC inhibition (Fig. 7A).

We tested this model further by increasing DE-cadherin dosage. Overexpression of DE-Cadherin in the A8 segment did not affect direction of genitalia rotation, which is dextral (Table 2, row 19), indicating that MyoID functions normally. However, MyoIC signal appeared to be strongly decreased (Fig. 4D, compare with wild type in 4A), suggesting a repression of MyoIC by DE-Cadherin. Consistent with the increase of both myosins in DE-cadherin loss-of-function conditions (Fig. 3A,B, lane 6), both myosin protein levels were reduced upon DE-Cadherin overexpression (Fig. 3A,B, lane 7, middle panel). It is important to note that in these conditions the silencing is restricted to the small A8 segment, but the contribution of DE-Cadherin and MyoIC in the other segments is not affected, therefore minimizing the effect observed on the western blot. Thus, DE-Cadherin indirectly increases MyoID activity further by repressing MyoIC specifically in the A8 segment.

MyoID was not detectable by immunohistochemistry in the DE-Cadherin overexpression context (Fig. 4D), and its protein level was reduced (Fig. 3A, lane 7, upper panel). Nevertheless, in this context MyoID level is sufficient for proper L/R establishment, owing to the strong decrease in MyoIC (and a favourable MyoID/MyoIC ratio of ∼0.5, equivalent to a heterozygous situation for myoID), thus consistent with our previous findings in the direct MyoIC silencing conditions (Table 2, row 3; Fig. 2E).

MyoID activity in directing dextral L/R asymmetry is temporally restricted and only required for 3 hours before pupariation (Speder et al., 2006). To test whether DE-Cadherin, MyoIC and MyoID have synchronous functions in L/R asymmetry, we made use of the temperature-dependent TARGET gene expression system (McGuire et al., 2004) to switch on or off the overexpression of MyoIC or the depletion of DE-Cadherin in the A8 segment at different developmental stages.

The overexpression of MyoIC shows that MyoIC anti-dextral activity takes place during days 6-7 of development and overlaps with the dextral determinant function of MyoID (Fig. 5A) (Speder et al., 2006). Interestingly, the temporal requirement for DE-Cadherin appeared to be broader between days 4-9 of development (Fig. 5B). Thus, the time window of DE-Cadherin function in L/R asymmetry overlaps with the activity peaks of MyoID and MyoIC (Fig. 5A) (Speder et al., 2006). The wider time window for DE-Cadherin can be explained by the applied method, which depends on the half-life of the mRNA and protein. Additionally, reduced DE-Cadherin might not only affect L/R determination at day 6 but also affect cell adhesion and tissue mechanics during the actual rotation of the genitalia at a later stage of morphogenesis. To distinguish between these temporally separate functions, we performed double temperature-shift experiments to silence DE-cadherin for short periods of 8 hours during development to narrow down DE-Cadherin temporal requirement. Using this approach, we show two peaks of DE-Cadherin requirement, the first overlapping with L/R determination and the second with the asymmetric rotation of the genitalia (Fig. 5C). These results suggest an uncoupling of DE-Cadherin functions in L/R determination at day 6, together with MyoID, and in the later genitalia morphogenesis and rotation independently of MyoID. The overlap at day 6 between the temporal functional windows of DE-Cadherin, MyoIC and MyoID is consistent with a synergistic effect of these proteins for L/R determination.

To understand the interaction between DE-Cadherin and type I myosins further, we performed immunofluorescence analysis of wild-type discs and examined the localization of the three proteins. We found that MyoID colocalizes with DE-Cadherin (Fig. 4F), which is consistent with the earlier reported colocalization of MyoID with β-Catenin (Speder et al., 2006). By contrast, DE-Cadherin did not colocalize with MyoIC (Fig. 4G), which is consistent with an inhibition of MyoIC by DE-Cadherin and our finding of mutually exclusive MyoIC and MyoID cortical localization domains (Fig. 1D).

To characterize the interactions between the adherens junction and both myosins, we performed co-immunoprecipitation experiments from larvae expressing either MyoID-GFP or MyoIC-RFP. As suggested by immunofluorescence analyses and previous results (Speder et al., 2006), β-Catenin could be co-immunoprecipitated with MyoID-GFP from larval extracts (Fig. 6A, lane 1). By contrast, β-Catenin and MyoIC-RFP did not show any interaction (Fig. 6A, lane 2). Furthermore, we showed a novel interaction between MyoID and DE-Cadherin. Indeed, DE-Cadherin co-immunoprecipitated with MyoID but not MyoIC (Fig. 6A′, lanes 1 and 2). Interestingly, both the interactions between MyoID and β-Catenin and MyoID and DE-Cadherin were abolished upon addition of MyoIC-RFP, leading to a strong reduction in, or even absence of, DE-Cadherin, and β-Catenin co-immunoprecipitated with MyoID-GFP (Fig. 6A,A′, lane 3). Therefore, the presence of MyoIC in the protein extracts is sufficient to disrupt the physical interaction between MyoID and the adherens junction components DE-Cadherin and β-Catenin, both in vivo and in vitro.

Taken together, these results suggest that the interactions between MyoID and the adherens junction components, which are inhibited by the antagonist MyoIC, are essential for MyoID function (Fig. 6D).

To support our in vitro data showing an interaction between MyoID and both DE-cadherin and β-Catenin (Fig. 6A) and a model in which the adherens junction acts as a myosin function regulating platform, we silenced other adherens junction components and looked for rotation defects and myosin’s localization. Depletion of either α- or β-Catenin in the L/R organizer led to dextral partial phenotypes (Table 3, rows 1-5), whereas stronger mutant conditions using stronger drivers or a combination of RNAi constructs to deplete either catenin together with DE-Cadherin led to lethality (Table 3, rows 6-10).

Interestingly, staining of α- or β-Catenin-depleted genital discs for MyoIC and MyoID showed an increase in MyoIC and a loss of MyoID signal, identical to our observation for DE-cadherin silencing (Fig. 6B,C).

These results further support the hypothesis that the adherens junction acts as a platform for myosin function required for L/R asymmetry establishment.

The mouse inversin (Inv) and Drosophila myoID are to date the only two known situs inversus genes. Inv was found to act as a switch between canonical and non-canonical Wnt signalling. Inv inhibits the canonical Wnt signalling through targeting Dishevelled (Dsh) to degradation (Simons, 2005). Diego, the Drosophila homologue of inversin, is a component of the non-canonical Wnt signalling pathway responsible for planar polarization. Therefore, we investigated whether disruption of the planar cell polarity (PCP) pathway, and especially Diego, affects L/R asymmetry in genitalia rotation. To this end, we expressed RNAi against most described PCP genes involved in the PCP pathway into the A8 segment of the genital disc and scored rotation phenotypes (supplementary material Table S1). Several PCP genes, including diego, showed dextral partial rotation defects, although none caused an inversion of directionality or a no-rotation phenotype. The partial rotation defects observed are likely to be due to polarization defects that affect the rotation process per se. Altogether, these results indicate that diego and other PCP genes do not play a major role in LR directionality in Drosophila.

Although several L/R molecules and signalling pathways have been identified in vertebrates, the molecular mechanisms underlying stereotyped L/R asymmetry establishment in invertebrates are largely unknown (Speder et al., 2007). Recently, an actin-based molecular motor, the unconventional type ID myosin, was identified as a L/R determinant in Drosophila, sufficient to orient the L/R axis (Hozumi et al., 2006; Speder et al., 2006). In this study, we describe a novel regulatory network controlling MyoID activity that is adherens junction dependent and can be antagonized by MyoIC, the closest homologue of MyoID.

We show that DE-Cadherin has a dual role in promoting MyoID activity. DE-Cadherin directly stabilizes MyoID at the adherens junctions and inhibits the antagonistic activity of MyoIC, thus eliciting MyoID function. We show further that MyoID activity and organization are dependent on the level of MyoIC protein. In MyoIC gain of function, MyoID intracellular pattern is modified with a reduced overall signal and the protein level remains unchanged (Fig. 2B; Fig. 3, lane 5). In MyoIC loss of function, MyoID is not detected by immunohistochemistry and the protein level measured by western blot is reduced (Fig. 2E; Fig. 3, lane 4). Nevertheless, in MyoIC loss of function, sufficient MyoID activity remains as confirmed by the wild-type rotation phenotype (Table 2, row 3). Thus, MyoID activity is only impaired when MyoIC is in large excess (MyoID/MyoIC ration <<1; Fig. 2B, Table 2, row 4).

It is important to note that these results were obtained through two independent approaches, either by direct modulation of MyoIC expression levels (Fig. 2B,E) or indirectly by affecting DE-cadherin (Fig. 4B,D), silencing of which in the A8 segment leads to an increase of MyoIC (and of MyoID) levels (Fig. 3). These results indicate that DE-Cadherin negatively regulates both MyoID and MyoIC levels. Taken together, these data indicate that DE-Cadherin controls both the L/R determinant MyoID and its repressor MyoIC activity and protein levels (Fig. 3, lanes 6, 7; Fig. 4D).

Specific depletion of DE-Cadherin in the A8 segment, the L/R organizer (Speder et al., 2006), leads to a no-rotation phenotype. Could stronger depletion of the adherens junction components leads to a sinistral phenotype? In other words, does DE-Cadherin depletion also affect sinistral development or pathway, which is taking over in absence of MyoID? To tackle this question, we depleted both DE-Cadherin and MyoID. We find that, unlike the sole depletion of MyoID that leads to a majority of sinistral flies, the double depletion leads to a majority of non-rotated flies (Table 2, rows 20-22). These data therefore indicate that DE-Cadherin acts both on the dextral (MyoID-dependent) and on the sinistral pathways, making it a general regulator of L/R asymmetry in flies.

Adherens junctions have previously been connected to MyoID-mediated L/R asymmetry establishment through biochemical analysis, which identified a physical interaction between the MyoID-tail domain and β-Catenin in vitro (Speder et al., 2006). In this study, we show a role of the adherens junction in the Drosophila L/R pathway and place the adherens junction component DE-Cadherin as a regulator of both unconventional myosins MyoIC and MyoID. DE-Cadherin dually promotes MyoID and β-Catenin association at the adherens junctions and inhibits the antagonistic function of MyoIC (Fig. 7A). Indeed, we observe a partial colocalization of MyoID and DE-Cadherin and the concomitant exclusion of MyoIC from the adherens junction (Fig. 4F,G). Further supporting this model is the additional biochemical evidence that MyoID, but not MyoIC, can directly interact with DE-Cadherin and β-Catenin in vitro (Fig. 6). Excess of MyoIC is able to disrupt this interaction both in vivo and in vitro, leading to MyoID loss-of-function phenotypes (Fig. 7B). This disruption can be rescued by MyoID overexpression (Fig. 7C). We propose that MyoID L/R function depends on its physical interaction with adherens junction through both β-Catenin and DE-Cadherin (Fig. 7A). Upon adherens junction reduction, MyoID does not associate with β-Catenin, MyoIC is no longer repressed and in turn inhibits MyoID activity, leading to L/R defects (Fig. 7D). Finally, the temporal synchrony of MyoID, MyoIC and DE-Cadherin requirement in genitalia rotation implies a concomitant activity of these proteins in L/R establishment.

Interestingly, MyoIC has previously been reported to affect L/R looping of the embryonic gut, another marker of L/R asymmetry in Drosophila (Hozumi et al., 2006). It was shown that MyoIC overexpression also causes fully penetrant embryonic gut inversion (Hozumi et al., 2006). Thus, the antagonistic function of MyoIC appears to be conserved in Drosophila L/R tissues. Furthermore, it was recently shown that DE-Cadherin is required for L/R asymmetry establishment of the Drosophila hindgut (Taniguchi et al., 2011). The dextral curving of the hindgut was lost in a myoID and DE-cadherin mutant background accompanied by a loss of L/R biased asymmetric cell shape and a differential DE-Cadherin localization. The authors suggest that both factors are implicated in the creation of asymmetric cortical tension prior to asymmetric curving of the hindgut. Taken together, these data indicate that the Drosophila L/R tissues, hindgut and genitalia, show a similar dependency on adherens junction and type I unconventional myosins for L/R determination.

Interestingly, despite the apparent differences between vertebrates and invertebrates mechanisms of L/R asymmetry establishment (Speder et al., 2007), a common theme can be found in the only two molecularly described situs inversus genes to date. In addition to myoID, the mouse inversin gene also causes constant L/R axis inversion in homozygous mutants (Yokoyama et al., 1993). Interestingly, the inversin protein was shown to co-precipitate with β-catenin and N-cadherin and to localize to the adherens junction of polarized epithelial cells (Nurnberger et al., 2002). Therefore, both situs inversus proteins interact with β-catenin and associate with adherens junction molecules, such as N- or E-cadherin. Additionally, N-cadherin was reported to play a role in L/R asymmetry establishment in chicken, as N-cadherin absence at the Hensen’s node leads the randomization of L/R asymmetry (Garcia-Castro et al., 2000), similar to our results showing a role of DE-Cadherin in L/R determination in Drosophila. The Drosophila inversin homologue diego belongs to the planar cell polarity gene family. Depletion of diego, or of any other of PCP gene, in the L/R organizer does not affect the directionality of genitalia rotation, suggesting that Diego does not play a critical role in Drosophila L/R asymmetry (supplementary material Table S1). In conclusion, the situs inversus proteins MyoID and Inversin, which play a central and upstream role in the establishment of L/R asymmetry, both require an interaction with the adherens junctions for their function.

It is important to note that the L/R axis is established after and, most importantly, relatively to the other two main axes, the anterior-posterior and dorsal-ventral axes (Hayashi and Murakami, 2001). Indeed, L/R determinants have to be oriented along existing spatial coordinates so that the L/R axis is positioned perpendicular to existing axes. We propose that the association of MyoID (and more generally of situs inversus proteins) with the adherens junctions is an essential mechanism to orient MyoID activity along the apical-basal axis, which in epithelia is perpendicular to the anterior-posterior axis and is thus equivalent to a dorsal-ventral axis. Therefore, association of MyoID with the adherens junctions would represent a way to orient and polarize its activity. Following polarization of MyoID in epithelia, the intrinsic chirality of myosins, through their directional activity towards one end of the actin filaments, could create de novo an asymmetric axis similar to the F-molecule model proposed by Brown and Wolpert (Brown and Wolpert, 1990).

In conclusion, our findings reveal an important link between the adherens junction and L/R asymmetry determinant in Drosophila and suggest an evolutionary conservation in vertebrates and invertebrates of the interactions between the adherens junction and the situs inversus proteins linking cell architecture and polarity to the patterning of the L/R axis. Future work on the molecular targets of MyoID and Inversin will help understand how the L/R axis becomes oriented to lead to stereotyped asymmetric organogenesis.

We are grateful to the anonymous reviewers for their constructive comments on the paper. We thank M. Mlodzik and E. Sanchez-Herrero for fly strains; and M. Clark and V. Van De Bor for critical reading of the manuscript. We thank the PRISM platform for its imaging facilities.