Src42A is required for E-cadherin dynamics at cell junctions during Drosophila axis elongation

Gastrulation represents a crucial morphogenetic process in early embryogenesis by which the blastomeres acquire cell fates in three different germ layers and the principal body axes become established (Williams and Solnica-Krezel, 2017). Much of the cellular basis of gastrulation movements is well known from studies in model organisms, including Drosophila melanogaster (Gheisari et al., 2020; Paré and Zallen, 2020). In Drosophila the anterior-posterior (AP) pattern originates during oogenesis to set up molecular gradients in the fertilized egg, which inform the elongation of the AP axis in a process called germband extension. The translation of the AP patterning into directional movement of the germband critically involves the generation of planar cell polarity through Toll-like receptor proteins (Paré et al., 2014).

A band of ventral-lateral epidermal cells forms the germband, which, during gastrulation, more than doubles its length along the anterior-posterior axis of the embryo, while narrowing in the dorsal-ventral direction (Kong et al., 2017). The cell behaviours associated with the extension of the germband include rearrangements of the apical cell contacts and basolateral cell protrusions (Sun et al., 2017). Three different modes of cell intercalations involving the apical cell junctions have been described: T1 transition, multiple rosette formation and vertex sliding (Bertet et al., 2004; Blankenship et al., 2006; Vanderleest et al., 2018). All three behaviours require the adherens junctions (AJs) to undergo precisely controlled remodelling (Levayer and Lecuit, 2013; Levayer et al., 2011; Rauzi et al., 2010). According to their position within the blastoderm epithelium, AJs can be classified into bicellular junctions (bAJs), referring to a two-cell interface, and tricellular junctions (tAJs), referring to a cell vertex where three cells attach to each other; both bAJs and tAJs are modulated during germband elongation.

The T1 transition describes the intercalation between four cells that exchange their neighbours by shrinking the bAJs in vertical direction and extending a new junction in horizontal direction (Bertet et al., 2004). The localized contraction and extension of the bAJ is facilitated through a planar polarized distribution of non-muscle myosin II (MyoII), the levels of which are enriched at vertical (anterior/posterior, AP) cell interfaces, while the scaffolding protein Bazooka (Baz) along with the E-cadherin/catenin complex is enriched at the horizontal (dorsoventral, DV) interfaces (Kong et al., 2017; Paré and Zallen, 2020; Zallen and Wieschaus, 2004). Interference with the planar polarization of either of these proteins negatively affects cell intercalation. For example, disruption of AP patterning via Krüppel RNAi knockdown affects junctional remodelling and MyoII AP planar polarity (Bertet et al., 2004). Junctional remodelling is also dependent on tyrosine phosphorylation. For example, the DV planar polarity of β-catenin is compromised after Abelson (Abl) tyrosine kinase knockdown (ablⁱ), resulting delayed cell intercalations (Tamada et al., 2012).

More recent studies have shown that tAJs also play a role in germband extension. The transmembrane protein Sidekick (Sdk) exclusively localizes at the tAJs and is vital for maintaining the length of the AP and DV interfaces. sdk mutants show defects in apical vertex adhesion and less strain in the tissue, resulting in a reduced rate of T1 transition (Finegan et al., 2019). In addition, Sdk localization changes with increasing mechanical tension. Expression of Myosin light chain phosphatase resulted in reduced Sidekick accumulation at the tAJs, suggesting that Sidekick protein levels are controlled by mechanical tension (Letizia et al., 2019). More recently, it has been shown that Canoe and its vertex localization are vital for cell intercalation; the translocation of Canoe from the tAJs to bAJs is crucial for germband extension, and this mobilization is mediated by phosphotyrosine signalling through Abl kinase (Yu and Zallen, 2020).

In addition to Abl, Src kinases have also been shown to be involved in germband elongation. A first report provided evidence that Src42A, in conjunction with the small GTPase Rac1, is required for the formation of actin-rich protrusions on the basal side of multicellular rosettes (Sun et al., 2017). Src42A was shown to be a component of a signalling pathway mediating the cell surface information from Toll-2 receptors towards PI3-kinase in order to promote the formation of MyoII planar polarity (Tamada et al., 2021). However, despite the well-known functions of Src family kinases in cell adhesion, the role of Src42A in regulating the distribution and dynamics of the E-cadherin adhesion complex has not yet been studied.

In this study we have explored the function of one of the two Src kinases encoded in the Drosophila genome, Src42A, in fine-tuning tAJs and bAJs during T1 transition. Consistent with previous reports, we find that Src42A has a substantial impact on T1 transitions. After Src42A single knockdown (Src42Aⁱ), the contraction of the AP border and the planar polarity of the bAJs are impaired. The delayed T1 transitions in Src42Aⁱ embryos led to slower cell intercalation, and correlated with reduced tension at the AP cell border and a delay in the rate at which the AP borders shrink during T1 transition. Src42Aⁱ embryos exhibit increased E-cadherin levels at the AP cell border, suggesting that the turnover of E-cadherin is affected. Consequently, E-cadherin intensity at the tAJs is increased in intercalating cells. E-cadherin dynamics at tAJs show an accumulation and dispersion cycle, which is affected by reducing the level of Src42A. Our data suggest that Src42A is involved in two not mutually exclusive processes; in addition to its function in Toll-2-like receptor signalling instructing the planar polarity of MyoII, we propose that it also acts as a signalling factor in controlling the amount of time that E-cadherin is present at tAJs.

The Drosophila genome encodes two Src homologs: Src42A and Src64B. To detect endogenous Src42A protein, we generated an antiserum against a full-length GST-Src42A fusion protein. The antibody was used to examine protein levels in embryos zygotically mutant for the loss-of-function allele Src42A^26-1 (Takahashi et al., 2005). Src42A levels were significantly reduced in late-stage embryos homozygous for Src42A^26-1 compared with heterozygous controls. At gastrulation stages, Src42A protein levels were unimpaired between homozygous and heterozygous Src42A^26-1 embryos, and were markedly reduced only at late stages of embryogenesis (Fig. 1A,B; Fig. S1A-E). Immunostaining of embryos expressing HA-tagged Src42A or Src64B revealed that the antibody was specific for Src42A, and did not crossreact with Src64B (Fig. S2). We conclude that our Src42A antibody detects maternally provided Src42A protein that decreases during embryogenesis, but persists at low levels into late stages of embryonic development.

Consistent with earlier studies, we found that Src42A exhibits a differential and dynamic association with plasma membrane domains in the early embryo from cellularization onwards (Takahashi et al., 2005). In early to mid-cellularization, Src42A was localized at the apical aspect of the growing plasma membrane, where it partially colocalized with Baz at the plasma membrane cortex (Fig. 1C,C′). A second accumulation of Src42A protein was present at the basal tip of the furrow canal, where the localization overlapped with Myo II (Fig. 1C,C′). During late cellularization, when spot AJs move apically, Src42A accumulated at the emerging apical AJs, whereas levels of Src42A at the furrow canals decreased (Fig. 1C,C′). During gastrulation, Src42A was observed at bAJs and tAJs on the apical side of the lateral epidermis during germband extension. (Fig. 1D,D′; Fig. S3A-C). Src42A was slightly enriched at the AP interface, where Myo II was localized during the generation of planar cell polarity (Fig. 1D′; Fig. S3D,E). We conclude that, by using a specific antibody, we detected a differential localization of Src42A at distinct plasma membrane domains in early embryos. The observation that Src42A was especially enriched at tension-generating sites prompted us to examine the function of Src42A during gastrulation.

Gastrulation stage embryos that are homozygous mutant for Src42A^26-1 harbour considerable amounts of maternal Src42A protein (Fig. S1E). The elimination of maternally supplied Src42A is not possible using the conventional autosomal DFS-FRT technique, as the Src42A locus is present between the centromere and the most proximal available FRT site (Chou and Perrimon, 1996). Therefore, to deplete maternal Src42A, RNAi knockdown experiments were conducted by employing the UAS/GAL4 system to express a transgenic short hairpin RNA using the P{TRiP.HMC04138} line (Brand and Perrimon, 1993; Staller et al., 2013) (Fig. 2A).

Embryos expressing TRiP04138 showed reduced hatching rates with variable penetrance and expressivity, depending on the maternal Gal4 driver used (Fig. 2A′). A combination of P{matα4-GAL4-VP16}67; P{matα4-GAL4-VP16}15 [called mat67 (mat15 hereafter)] was most efficient as embryos showed about 80% reduction in hatching rates and were therefore used for all the experiments (called Src42Aⁱ in this study). At late syncytial and early cellular blastoderm stages, 40% of Src42Aⁱ embryos exhibited a defect in cytoplasmic clearing indicated by an irregular interface between the central yolk and the clear cortical cytoplasm (Fig. 2B,D; Movie 1). By immunoblotting, we found that embryos with clearing defects exhibited the most severe reduction in Src42A protein levels (Fig. 2C). Similar clearing defects have also been reported for embryos injected with dsRNA against Src42A mRNA (Sun et al., 2017). Src42Aⁱ embryos displayed variable cuticle phenotypes, which were stronger compared with zygotic Src42A^26-1 homozygous mutants (Fig. 2E).

Although the Src42Aⁱ embryos did not exhibit major defects at the onset of gastrulation movements, defects in cytoplasmic clearing suggest a requirement for Src42A in cellularization, which may lead to an abnormal blastoderm epithelium in the knockdown embryos. We therefore examined the localization of markers for membrane domains during cellularization and found no major differences of Src42Aⁱ embryos compared with controls (Fig. S4). Despite the formation of a rather normal blastoderm epithelium, we found that occasionally nuclei dropped from the cell cortex into the centre of the embryo during early cellularization (Fig. S4B; Movie 1). We conclude that Src42Aⁱ embryos exhibit mild cellularization defects, including cytoplasmic clearing and nuclei dropping from the cortex, but that the overall polarized structure of the blastoderm epithelium remains largely unimpaired.

Our results showed that maternal Src42A is required for embryogenesis, but that the depletion of maternal Src42A only causes minor defects in the formation of the blastoderm. We found by immunofluorescence that during germband extension Src42A was localized at bAJs and tAJs (Fig. 1B; Fig. S3). bAJ- and tAJ-resident proteins have been reported to undergo dynamic redistribution during germband extension (Finegan et al., 2019; Uechi and Kuranaga, 2019). Src42Aⁱ embryos exhibit defects in germband elongation, when analysed by video time-lapse recordings (Fig. 3A; Movie 2). Germband elongation occurs in two phases: an initial fast phase, in which the germband doubles its length within 25 min at 25°C, is followed by a slow phase with an additional 0.5-fold elongation within 70 min (da Silva and Vincent, 2007; Irvine and Wieschaus, 1994) (Fig. 3B,C). Plotting of cumulative displacement length of the germband over time revealed a substantial delay in its AP extension (Fig. 3B). Germband cells showed a 0.25- and 0.16-fold delay in Src42Aⁱ embryos for the fast and for the slow phase, respectively (Fig. 3C-E). We further examined germband cells in Src42Aⁱ embryos at a subcellular level. Src42A has several known substrates, which are present at bAJs and tAJs (Brunet et al., 2013; Takahashi et al., 2005). Indeed, at the onset of germband extension (stage 7), Src42Aⁱ embryos exhibited significantly reduced phosphotyrosine levels at bAJs and tAJs (Fig. 3F-H). This indicates that Src42A is required for normal levels of protein phosphorylation at tyrosine residues at bAJs and tAJs. Consistent with earlier findings that Src42A controls germ band elongation by changing the properties of signalling factors at the subcellular level (Sun et al., 2017; Tamada et al., 2021), these results demonstrate that Src42A-dependent tyrosine phosphorylation is required for normal germband elongation.

T1 transitions are a key cell behaviour during the fast phase of germband extension. It involves junction contraction at the AP cell interfaces followed by neighbour exchange along the DV axis, resulting in tissue elongation (Bertet et al., 2004). Recently, it was shown that Src kinases play a role downstream of Toll-2 in planar polarization of Myosin II and Baz (Tamada et al., 2021); when both Src42A and Src64B were knocked down, the rate of intercalation determined by altered arbitrary cell edge contraction rates was reduced. Based on the finding that Src42Aⁱ embryos display drastic reduction in the fast phase of germband elongation (Fig. 3D), we sought to investigate T1 transitions in Src42Aⁱ embryos. Using Utrophin-GFP as a marker for the plasma membrane-associated actin cortex, T1 transitions were recorded in control and Src42Aⁱ embryos (Movie 3). The shrinkage of the junction was measured as the time it takes for the AP interface junction between two adjacent vertices on the DV axis to shrink into a single four-cell vertex. In four-cell vertices undergoing T1 transitions, the shrinking junction along the AP axis reduces its length with an average of 1.25 min (Fig. 4A). When compromising Src42A function, this aspect of the T1-process was delayed (Fig. 4B). In 60% of the cases in the Src42A knockdown, the shrinking of the AP junction interface was not completed after 2 min (Fig. 4B′,D). This suggests that the rate at which T1 transitions occur is delayed due to less contractility at the AP junctions. We noted that the number of germband cells in Src42Aⁱ embryos was slightly decreased and the average apical cell area was larger than in wild-type controls (Fig. S5). These defects may be related to the nuclear dropping observed during cellularization in Src42Aⁱ embryos. However, it is unlikely that these effects cause the delay in T1 transitions, as cell size and cell number were shown to have little if any effect on germband extension (Edgar and O'Farrell, 1989; Irvine and Wieschaus, 1994). We conclude that Src42A is required for contraction of AP junctions during T1 transitions in germband elongation.

In T1 transitions, the AP cell border is enriched with MyoII, which mediates tension through its contractile properties. Moreover, a supracellular enrichment of MyoII was also reported to occur due to tensile forces along the AP cell interfaces (Bertet et al., 2004; Fernandez-Gonzalez et al., 2009). In Src42A, Src64B double-knockdown embryos, MyoII planar polarity was reduced (Tamada et al., 2021). We found that, in single-knockdown Src42Aⁱ embryos T1 transitions are delayed and we asked whether this correlates with a reduction in MyoII-dependent tension at the AP border of the cells undergoing T1 transition. To measure the tension at the AP cell boundaries, we performed laser ablation experiments in intercalating cells and measured the recoil velocity. When the AP border was cut, the detached tAJs moved slower in Src42Aⁱ embryos compared with control (Fig. 5A). The displacement of tAJs was normalized to initial length and plotted over time (Fig. 5B,C). A non-linear regression analysis on the displacement data revealed that the initial recoil velocity was reduced in T1 transitions in Src42Aⁱ embryos (Fig. 5D). From these experiments, we conclude that Src42A is involved in generating or maintaining tension in intercalating cells at T1 transitions.

In Drosophila, Src42A genetically interacts with components of the E-cadherin/β-catenin complex and Src42A can be detected in a ternary complex with E-cadherin and β-catenin (Takahashi et al., 2005). Furthermore, the adhesive properties of the mammalian E-cadherin-catenin complex are regulated by phosphorylation of Y654 of β-catenin through c-src (Röper et al., 2018). β-Catenin exhibits a planar-polarized distribution and is preferentially localized at the DV cell border of germband cells; the phosphorylation of β-catenin Y667 by Abelson kinase (Abl) is vital to achieve this polarization (Tamada et al., 2012). It has also been shown in the Drosophila wing epithelium that Src42A can potentially activate Abl kinase (Singh et al., 2010). These data raised the hypothesis that Src42A affects the E-cadherin/β-catenin complex indirectly via Abl. In Src42Ai embryos, there was no significant change in protein levels at the DV cell border, but the AP borders showed enhanced levels of β-catenin and E-cadherin (Fig. 6). These data were also supported by quantification of live imaging of E-cadherin-GFP in wild-type and Src42Aⁱ embryos (Fig. S6). We conclude that the planar polarity of the E-cadherin-catenin complex is compromised in Src42Aⁱ embryos (Fig. 6A-E).

To examine the relationship of Src42A and Abl in axis extension, we conducted a double knockdown experiment targeting both kinases. Germband elongation was examined by time-lapse video microscopy in controls compared with single knockdowns and Src42Aⁱ+Ablⁱ double knockdowns (Fig. 7). The cumulative displacement values were plotted over time and the speed of elongation was calculated from the displacement data (Fig. 7A-C). We found that, in the initial fast phase, Src42Aⁱ+Ablⁱ exhibits a slightly stronger delay in germband extension than Src42Aⁱ or Ablⁱ alone; Ablⁱ alone did not affect the speed of germband elongation in the fast phase (Fig. 7B,C). During the slow phase of extension, Src42Aⁱ+Ablⁱ showed stronger delay compared with Src42Aⁱ or Ablⁱ alone (Fig. 7B,C). As the overall effect on germband extension in Ablⁱ single knockdowns was rather mild (Fig. 7A-C), we examined Abl protein levels in Ablⁱ embryos and found that Abl protein was still detectable at bAJs, indicating perdurance of maternal Abl protein (Fig. 7D,D′). Although Src42A levels were unimpaired in Ablⁱ knockdown embryos, Abl protein was slightly reduced in Src42Aⁱ embryos (Fig. 7E-F′). These data indicate that Src42A and Abl are both required for germband extension. As Src42A is required for normal bAJ localization of Abl, we propose that, similar to the cell invasion process in the wing imaginal disc epithelium (Singh et al., 2010), Src42A may act upstream of Abl in a common pathway.

We found that Src42A showed a distinct accumulation at the tAJs (Fig. 1B). The tAJs is one of the tension hotspots during germband elongation, and Sidekick (Sdk) protein is a core component of this subcellular domain (Finegan et al., 2019; Salomon et al., 2017; Uechi and Kuranaga, 2019). Sdk also regulates E-cadherin endocytosis during cell intercalation when new junction formation occurs at the horizontal interface (Letizia et al., 2019). Despite the requirement of Sdk, the mechanisms controlling E-cadherin dynamics at the tAJs are not well understood. Therefore, we recorded E-cadherin-GFP localization in the background of Src42Aⁱ during T1 transition. The localization of E-cadherin-GFP exhibits an accumulation and dispersion cycle during T1 transition at tAJs (Fig. 8A) (Vanderleest et al., 2018). In Src42Aⁱ embryos, E-cadherin levels were continuously increasing at the tAJs, suggesting that the turnover of E-cadherin is affected during T1 transition (Fig. 8A; Movie 4). From the vertex intensity ratio calculation, E-cadherin levels were not fluctuating in embryos deficient for Src42A (Fig. 8B-D). The prolonged residence times and accumulation of E-cadherin at the junctions could be a result of alterations in the turnover rates of E-cadherin. We therefore performed fluorescence recovery after photobleaching (FRAP) experiments on control and Src42Aⁱ embryos expressing E-cadherin-GFP (Fig. S7). The data show that the E-cadherin-GFP fluorescence recovery rates at tAJs in Src42Aⁱ are similar to controls (Fig. 8E,F). In the case of bAJs, E-cadherin-GFP recovery is initially unimpaired, but ∼110 s after bleaching, fluorescence intensity in Src42Aⁱ overshoots the levels of the controls, which run into equilibrium (Fig. 8E,F). The observed increase in Src42Aⁱ embryos is consistent with the accumulation of E-cadherin at bAJs: when E-cadherin recovery approached equilibrium in controls, E-cadherin-GFP levels further increased in Src42Aⁱ embryos. Furthermore, stage 7 Sdkⁱ embryos were subjected to immunostaining experiments to examine E-cadherin and Src42A localization (Fig. 8G). The results show that there were no changes in E-cadherin levels, however, Src42A levels at the tAJs were increased (Fig. 8H,I). Together, these data suggest that E-cadherin turnover at adherens junctions is compromised after Src42A RNAi knockdown.

The use of transgenic RNAi to knock down maternal Src42A has caveats, including off-target effects and incomplete depletion due to perdurance of maternal Src42A gene product. Attempts to confirm the specificity of the RNAi knockdown by rescuing Src42Aⁱ embryos using Src42A transgenes were inconclusive, because Src42A overexpression, in particular of tagged versions, can induce lethality (Fig. S8) (Irby and Yeatman, 2000). Therefore, to confirm the specificity of the effects caused by RNAi knockdown experiments, we generated germline clones by inducing double-strand breaks at a chromosomal site (stlk; 41A3) proximal to the Src42A (42A6-7) locus using CRISPR-Cas9 (Allen et al., 2021). In germline clones homozygous for the complete loss-of-function allele Src42A^26-1 (named Src42A^GLC in this study), the cytoplasmic clearing was consistently defective during the transition from cell cycle 13 to 14. The number of nuclei in blastoderm stage embryos during cellularization was reduced in Src42A^GLC embryos compared with controls, suggesting that nuclei dropping is more severe under these conditions (Fig. S9A,B, Movie 5). These defects suggest that Src42A may have a yet unknown function in blastoderm development. The shape and size of eggs derived from Src42A germline clones were more irregular and smaller, respectively, compared with Src42Aⁱ or wild type (Fig. 9A,C). In ovaries from Src42A^GLC-carrying females, Src42A is absent from nurse cell interfaces. The absence of Src42A from the germline also affected the localization of E-cadherin at the nurse cell boundaries (Fig. 9B). We conclude that depletion of Src42A in the germline affects the structure of the egg chamber in oogenesis and blastoderm formation in embryogenesis.

We further analysed the speed of germband elongation (Fig. 9C,D) and the planar polarized distribution of β-catenin and E-cadherin (Fig. 9E-I). In all aspects examined, the phenotype of Src42A^GLC was stronger compared with Src42Aⁱ, confirming the notion that some Src42A maternal gene products escaped the RNAi knockdown. This is also consistent with an increased lethality of Src42A^GLC embryos compared with Src42Aⁱ (Fig. S9C). This experiment provides evidence that the knockdown effects in Src42Aⁱ are specific to reduction of Src42A and indicates that Src42A is also involved in normal germline and blastoderm development.

In this study, we aimed to understand the function of Src42A in regulating the dynamics of E-cadherin at adherens junctions during germband extension. Depletion of maternal and zygotic Src42A by RNAi knockdown resulted in defects in early embryogenesis. We also generated, for the first time, germline clones for a mutant Src42A allele using a CRISPR/Cas9 approach that can be used to generate germline or somatic mosaic clones for genes that are present proximal to the FRT site in the centromeric region. The embryos derived from Src42A germline clones recapitulate the phenotypes of Src42A knockdown experiments and extend the spectrum of requirements of maternal Src42A to oogenesis and blastoderm formation.

The most penetrant and robust phenotype of Src42A knockdown and maternal mutant embryos was a strongly compromised germband extension, which was associated with impaired planar polarity of E-cadherin and β-catenin. In particular, the E-cadherin intensity at AP cell borders was abnormally high, suggesting that E-cadherin turnover at the AP border is affected. During this stage, E-cadherin polarity was shown to control the flow of actomyosin to the AP cell border (Levayer and Lecuit, 2013). In the absence of E-cadherin endocytosis, the MyoII flow required for vertical junction constriction may be compromised. This conclusion is consistent with reduced Myo II polarity (Tamada et al., 2021) and with the reduced recoil velocity that we observed in Src42Aⁱ embryos when the AP border was ablated. Less tension at the AP border implies that the constriction rate of vertical AJs during the T1 transition is compromised. Our findings support the view that Src kinases (both Src42A and Src64B) act downstream of Toll-2 and Toll-6 receptors, and thereby alter the formation of Myo II cables at the AP cell border and Bazooka localization at the DV cell border (Tamada et al., 2021). Therefore, knocking down Src42A alters the dynamics of molecular components in T1 transition cells and thereby compromises the speed of germband extension.

Previous studies on Abelson kinase (Abl) show that it affects the planar polarity of β-catenin during rosette formation (Tamada et al., 2012). During cell invasion, Src42A can act upstream of Abl and there is a potential feedback from Abl, with a positive impact on Src42A (Singh et al., 2010). Despite the known interplay between Src and Abl kinases, the relationship between these two kinases was not studied during germband elongation. Here, we show that RNAi knockdown of both Src42A and Abl kinases resulted in an enhanced phenotype during the slow phase of germband extension. Although the molecular details of this genetic interaction remain to be determined, our data support the model that Src42A acts upstream of Abl in a common genetic pathway in particular during the slow phase, where the rosettes resolve by extending new horizontal junctions.

Abl is required for the mobility of Canoe (the Drosophila homolog of afadin) from tAJs to bAJs, and this mobility depends on tyrosine phosphorylation of Canoe by Abl, as a Canoe phospho-mutant displayed delayed T1 transitions (Yu and Zallen, 2020). Canoe maternal mutants show similar defects to Src42Aⁱ embryos: both display less phosphotyrosine signal at the tAJs. Src42A can also phosphorylate the LIM domain-only protein Smallish, which is required for the planar polarity of Baz and Canoe at the bAJs, and Smallish is enriched at tAJs during germband extension stages (Beati et al., 2018). It is known that Canoe acts as linker between the junctional cadherin-catenin complex and the actin cytoskeleton. We hypothesize that Src42A might activate Abl and therefore lead to an enhancement of Canoe mobility. When Abl alone was knocked down, the mobility of Canoe at the AJs was impaired (Yu and Zallen, 2020). If Src42A acts directly on Canoe or indirectly through Abl to mobilize Canoe between bAJs and tAJs, this would result in a change in the proportions of Canoe at tAJs and bAJs in Abl/Src42A double knockdowns. In mammals, Afadin regulates E-cadherin turnover at cell junctions by trans-interacting with Nectin. If there is no trans-interaction, E-cadherin is internalized by Clathrin-mediated endocytosis (de Beco et al., 2009; Hoshino et al., 2005; Takeichi, 2014). In our analysis, we find that E-cadherin intensity at tAJs remains high in Src42Aⁱ embryos compared with wild-type embryos. This implies that E-cadherin is accumulating at the tAJs, which could possibly occur due to reduced levels of Canoe causing increasing E-cadherin clusters at the tAJs during T1 transition.

At the mid-stage of germband extension, E-cadherin protein levels at the tAJs exhibit an accumulation and dispersion cycle that correlates with cell intercalation (Vanderleest et al., 2018). The resident tricellular junction protein Sdk regulates E-cadherin endocytosis during genitalia rotation, where cell intercalation occurs continuously (Uechi and Kuranaga, 2019). Sdk connects to the actin cytoskeleton through Canoe and Polychaetoid, which transmit tension at the tAJs (Letizia et al., 2019). In Src42Aⁱ embryos, E-cadherin intensity is continuously increased at tAJs during cell intercalation, suggesting a specific role for Src42A in maintaining E-cadherin turnover in tAJs. What acts upstream of Src42A at tAJs in mediating E-cadherin turnover remains elusive. We aimed to understand the relationship between Sdk and Src42A in the context of E-cadherin turnover. Under Sdkⁱ conditions, we were not able to observe any difference in E-cadherin intensity in fixed embryos but the level of Src42A was increased at tAJs. These data suggest that Sdk directly or indirectly controls the recruitment of Src42A to the tAJs, but the exact relationship between Sdk, Src42A and E-cadherin turnover remains to be addressed.

In this study, we provide evidence of a requirement for Src42A in regulating molecular components at the bAJs and tAJs in axis extension in Drosophila. During T1 transitions, Src42A regulates the residence time of E-cadherin, suggesting that Src42A may function at the tAJs. Such a role of Src42A had not been reported so far. For example, Src42A might be involved in transducing a signal in response to mechanical tension at the tAJs into the cell. Given the reciprocal interactions between actomyosin planar polarization and E-cadherin stabilization, we propose that, in addition to translating the planar polarized information of Toll receptors to MyoII, Src42A may affect E-cadherin dynamics at bAJs and tAJs through phosphorylation of additional, yet unknown components, which may or may not include the E-cadherin/catenin complex itself. It will therefore be interesting to identify tAJ-resident proteins that act upstream and downstream of Src42A.

Full-length Src42A cDNA was cloned into pGGWA vector using the Gateway cloning method (Katzen, 2007) for antibody generation. The final pGGWA+Src42A vector contains a GST (Glutathione S-transferase) tag at the N-terminal end of Src42A; furthermore, the recombinant protein GST-Src42A was expressed in BL21(DE3) E. coli cells at 18°C overnight and purified using affinity chromatography (using Glutathione Sepharose beads). Guinea pigs were used for antibody generation (Eurogentec). After immunization, the serum from the final bleed was used directly as a Src42A primary antibody with a dilution of 1:500 for all experiments.

All fly stocks were raised at 25°C. To analyse the dynamic localization of Src42A during cellularization and germband extension, embryos from Sqh::KI-GFP flies were fixed and stained. Additionally, Src42A^26-1 flies (zygotic mutant) (Takahashi et al., 2005) were used to study the specificity of the generated antibody. To check the cross-reactivity of Src42A antibody with Src64B, UASp>Src42A-HA and UASp>Src64B-HA female flies (gifts from Andreas Wodarz, University of Cologne, Germany) were crossed with engrailed>Gal4 male flies.

Maternal and zygotic knockdown of Src42A was performed using the UAS/GAL4 system (Brand and Perrimon, 1993; Staller et al., 2013). Female flies carrying shRNA for Src42A under the control of UAS promoter (y¹ sc* v¹ sev²¹; P{TRiP.HMC04138}attP2/TM3, Sb¹) were crossed to male driver flies (y¹ w*; P{matα4-GAL4-VP16}67; P{matα4-GAL4-VP16}15). In the next generation P{matα4-GAL4-VP16}67/ +; P{matα4-GAL4-VP16}15/ P{TRiP.HMC04138}attP2 flies were collected and both males and females from these genotypes were crossed to obtain embryos in which maternal and zygotic Src42A (named Src42Aⁱ in this work) was knocked down in the F2 generation. As controls in germband elongation assays, male driver lines (y¹ w*; P{matα4-GAL4-VP16}67; P{matα4-GAL4-VP16}15) were crossed to w¹¹¹⁸ females, the F1 generation was backcrossed and the F2 embryos were analysed. For all immunostaining quantification experiments, His-GFP embryos were used as a control and stained together in a same tube along with Src42Aⁱ embryos. For dynamic analysis of T1 transitions, y¹ w*; P{matα4-GAL4-VP16}67 sqh::Utrophin-GFP; P{matα4-GAL4-VP16}15/TM6 flies were used as a driver to obtain Src42Aⁱ embryos; Sqh::Utrophin GFP flies were used as a control.

For E-cadherin vertex intensity analysis and laser ablation experiments, y¹ w*; P{matα4-GAL4-VP16}67 Shg::DECadGFP/ Shg::DECadGFP; P{matα4-GAL4-VP16}15/ P{TRiP.HMC04138}attP2 flies were crossed to obtain Src42ⁱ embryos along with an E-cadherin marker; Shg::DECadGFP flies were used as a control. For Src42A and Abelson double knockdown (Src42Aⁱ_Ablⁱ) P{TRiP.HMC05140}attP40; P{TRiP.HMC04138}attP2/TM3, Sb¹ females were crossed with same maternal driver line as mentioned above. RNAi experiments were also performed for sidekick (sdk) using y¹ sc* v¹ sev²¹; P{TRiP.HMS00292}attP2 fly line (mentioned as Sdkⁱ⁾. All the TRiP fly lines were generated at Harvard Medical School for the Transgenic RNAi Project (Ni et al., 2011).

Germ-line clones lacking Src42A were generated using a CRISPR-Cas9 approach based on gRNA-induced double-strand (ds) breaks at a site (stlk locus; 41A3) proximal to the Src42A (42A6-7) locus. A transgene constitutively expressing stlk gRNA (P{TKO.GS00956}attP40(y⁺); Bloomington 76505) was recombined with Src42A^26-1. To induce ds breaks in the non-essential stlk gene, y w; stlk-gRNA-attP40(y⁺) Src42A^26-1 females were crossed with y w act5c-Cas9 lig4/Y; FRT42D ovoD(w⁺)/+ males. All eggs produced by F1 females of the genotype y w act5c-Cas9 lig4 / y w; stlk-gRNA-attP40(y⁺) Src42A^26-1 / FRT42D ovoD(w⁺) are derived from germline cells lacking the dominant female-sterile ovoD transgene (Chou and Perrimon, 1996) and are therefore homozygous for Src42A^26-1. Control females carrying the FRT42D ovoD(w⁺) chromosome in the absence of the act5c-Cas9 source or of the stlk-gRNA source did not lay any eggs. Females carrying Src42A^26-1 germline clones were crossed with w*/Y; Src42A^26-1/CyO[twi::GFP] males to obtain embryos lacking maternal and zygotic Src42A. The genotypes of the germline clone embryos were identified by the absence of twi::GFP expression.

Embryos were collected from 0 to 7 h and fixed using 4% formaldehyde made in phosphate-buffered saline as described previously (Müller, 2008). After fixation, the embryos were blocked in 5% BSA and stained with the following primary antibodies: guinea pig anti-Src42A (1:500), rabbit anti-Bazooka (1:2000) (Wodarz et al., 1999), mouse anti-phosphotyrosine (PY20) (1:1000) (BD Biosciences, 610000), rabbit anti-β galactosidase (1:1000) (Cappel/ICN/MP 55976), rat anti-E-cadherin DCAD2 (1:20) (Developmental Studies Hybridoma Bank), mouse anti-Arm N2 (1:250) (Developmental Studies Hybridoma Bank), rat anti-haemagglutinin 3F10 (1:1000) (Roche) and rabbit anti Abl (1:50) (a gift from Mark Peifer, University of North Carolina, Chapel Hill, USA). The following secondary antibodies were used, goat anti-guinea pig Alexa 647 (Invitrogen A-21450), goat anti-rabbit Cy3 (Stratech BS-0432R-CY3-BSS), donkey anti-mouse Cy3 (Jackson ImmunoResearch 715-165-150), donkey anti-rat Cy2 (Jackson ImmunoResearch 715-225-153) and donkey anti-rat Cy3 (Jackson ImmunoResearch 715-165-153). All secondary antibodies were used at 1:250 dilution.

For immunoblotting experiments, staged embryos were collected and lysates were prepared under denaturing conditions using 2×SDS sample buffer [0.125 M Tris (pH 6.8), 20% glycerin, 4% SDS, 0.004% bromophenol blue and 10% β-mercaptoethanol]. To check the knockdown efficiency, embryos were collected from 0 to 7 h, protein lysates were prepared using RIPA buffer (50 mM Tris, 150 mM NaCl, 1% NP40, 0.1% SDS, 0.5% Na-deoxycholate and 1% Triton X-100) under non-denaturing conditions and separated on a 15% SDS-PAGE gel (Wodarz, 2008), transferred onto nitrocellulose membrane (Amersham Protron 0.2 µm NC), and the membrane blocked and stained using the primary antibodies guineapig anti-Src42A (1:500, this study), rabbit anti-haemagglutinin (1:1000) (Sigma-Aldrich H6908) and mouse anti-alpha tubulin 12G10 (1:2000) (Developmental Studies Hybridoma Bank). The secondary antibodies donkey anti-guinea pig IR dye 800 (1:10,000) (LI-COR 925-32411), donkey anti-mouse IR dye 680 (1:10,000) (LI-COR 926-68072) and donkey anti-mouse HRP (1:2000) (Jackson ImmunoResearch 715-035-150) were used to detect the protein levels in a LI-COR Odyssey Fc imaging system. The immunoblots were simultaneously stained with antibodies against the antigen to be detected and with anti-alpha tubulin as a loading control. The immunoblot was imaged in two separate channels and is depicted in two panels.

Images were acquired using confocal laser scanning microscope (Zeiss LSM880) for 1.2 µm depth from the surface in z-plane with 0.2 µm interval. All the images were processed and analysed using Fiji. In total, six images were acquired, and maximum intensity projection (MIP) images were generated from these images. Using the MIP images, quantifications were performed. Average pixel intensity at the two and three cell contacts were measured and subtracted with the background value. The pixel intensities were normalized to the maximum value of the control image. Both control and Src42Ai embryos were stained in the same tube for quantification purposes. The planar polarized distribution of immunolabeled proteins was analysed by measuring pixel intensities of bicellular junctions at AP and DV borders; these pixel intensities were normalized to the maximum value of AP- and DV-border measurement in control embryos. The ratios of AP- to DV-border were calculated by dividing the raw data of DV-border measurement to AP-border measurements.

Bright-field images were taken from Zeiss Axiophot and Olympus BX61 microscope, and processed using Fiji. The length of germband elongation was tracked using ‘manual tracking’ plug-in, and the cumulative displacement of germband were plotted over time. To access the speed of germband elongation, a linear curve was drawn for the slow and fast phase of germband elongation, and the linear curve slope was calculated for respective phases.

Live-imaging experiments were performed using sqh::Utrophin-GFP as a marker for cell membrane. Imaging was performed using a confocal laser scanning microscope (LSM880) in airy scan mode. The transition between stage 6 and 7 were recorded with a 15 s time interval. All the acquired images were analysed using ‘Tissue Analyzer’ plug-in in Fiji. Images from each timepoints were segmented and the length of AP axis and DV axis was measured from segmented images. Initial AP axis length (L0) is normalized to the length over time (Lt) and the normalized value was plotted over time.

Images were taken using a confocal laser scanning microscope in airy scan mode. Germband cells undergoing T1 transitions were imaged for a depth of 1 µm in z-plane with 0.2 µm interval. MIP images were generated from the acquired images for all time points. Cells undergoing T1 transition containing 2.5 µm AP axis length were selected and used for analysis. From the final images, E-cadherin vertex intensity ratio is calculated using the following formula (ItAJs-IB)/(IbAJs-IB) (Vanderleest et al., 2018). The vertex intensity ratio was then plotted over time.

Various types of statistical tests were applied and indicated in the figure legends. When a paired Student's t-test was performed, the normal distribution of the data was calculated using D'Agostino and Pearson test for datasets with more than eight elements (D'Agostino and Pearson, 1973). Alternatively, the Kolmogorov–Smirnov Test was used (Smirnov, 1948). The sample sizes (n) for each experiment are indicated in the figure legends.

Stage 7 embryos expressing ECadGFP were prepared for live imaging and recorded in the GFP channel with 1 s time interval on a confocal laser scanning microscopy (Zeiss, LSM980 with 100× magnification using oil immersion, 1.4 NA). A 355 nm pulsed laser (DPSL355/14, 355 nm, 70 µJ/pulse, Rapp OptoElectronic) was employed for ablation and manipulated on the ‘REO-SysCon-Zen’ platform (Rapp OptoElectronic). The 355 nm pulsed laser was mounted on epiport of the confocal laser scanning microscopy. Laser ablation was performed with 5% of laser power, with 200 ms (around 40 pules) exposure time at the AP axis of the cells undergoing T1 transition. For analysis, the displacement length of ablated tAJs [L(t)] were measured manually in Fiji. The displacement value were normalized to the initial length [L(0)] between tAJs and plotted over time. The initial recoil velocity were calculated using Kelvin-Voigt fibre model to the following equations using Prism8 (Fernandez-Gonzalez et al., 2009; Liang et al., 2016).

We thank Andreas Wodarz and Mark Peifer for providing fly stocks and antibodies. We thank Tanja Wilhelm, Christine Otto and Nicole Schleinschok for fly maintenance and expert technical assistance, and Monika Winneknecht and Birgit Simon for preparing fly media. We thank the Bloomington Drosophila stock centre for Drosophila stocks, the Developmental Studies Hybridoma Bank for antibodies and the Transgenic RNAi Project for providing RNAi lines. Work in H.-A.J.M.'s laboratory was supported by core funding from the Universität Kassel and the PhosMOrg Consortium.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201119.reviewer-comments.pdf