The Abl-interactor Abi suppresses the function of the BRAG2 GEF family member Schizo

BRAG proteins are a subgroup of the Arf-guanine nucleotide exchange factor (GEF) family that are mandatory for developmental and physiological processes, e.g. myoblast fusion, neuronal pathfinding and synaptic transmission. However, they also play an important role during disease progression, e.g. cancer metastasis and X-chromosome-linked intellectual disability (D'Souza and Casanova, 2016). The BRAG family is characterized by an N-terminal located calmodulin-binding IQ motif, a catalytic Sec7 domain of ∼200 amino acids and a pleckstrin homology (PH) domain that is immediately located downstream of the Sec7 domain. The Sec7 domain stimulates the release of GDP to allow binding of GTP on ADP-ribosylation factor (Arf) family members. These Arf GTPases serve as master regulators of intracellular membrane traffic and actin dynamics. A primary challenge in understanding the activation of small GTPases in development, tissue homeostasis and disease is to study the molecular mechanism underlying GEF activation.

The human GEFs for Arf family proteins are grouped into six evolutionarily conserved families known as BIG, BRAG/IQSec, Cytohesins, EFA6/PSD, FBX8 and GBFs. Arf GEFs of the Cytohesin and BIG family are regulated by auto-inhibition (DiNitto et al., 2007; Richardson et al., 2012; Stalder et al., 2011). The GEF activity of Cytohesin is suppressed through the Sec7-PH linker and the C-terminal helix/polybasic region that mask the active side in the Sec7 domain (DiNitto et al., 2007). The auto-inhibition of the yeast BIG family member Sec7 depends on an intramolecular interaction that involves the HSD domains (Richardson et al., 2012). Auto-inhibition is released by the binding of membrane-bound Arf-GTP to the PH domain or the HSD1 domain. A positive feedback loop arises through the generation of more Arf-GTP by the Sec7 GEF, which leads to the recruitment of more Arf-GEF. In contrast, high resolution crystal structure of the unbound Sec7-PH domain of the Sec7 GEF BRAG2 revealed that the lipid-binding side in the PH domain and the Arf-binding site in the Sec7 domain are both constitutively active (Karandur et al., 2017). The finding that the Sec7 and PH domains are constantly accessible for protein­–protein and protein–membrane interactions raises the question of how the spatial–temporal activation of BRAG2 is achieved.

The Drosophila BRAG family member Schizo is required for the guidance of neuronal axons (Önel et al., 2014) and muscle development (Chen et al., 2003; Dottermusch-Heidel et al., 2012). Muscles are multinucleated cells that arise by the fusion of mononucleated myoblasts. Besides muscle formation, myoblast fusion is crucial for the maintenance, growth and repair of muscles in mammals and Drosophila (Abmayr and Pavlath, 2012; Chaturvedi et al., 2017). However, the precise function of Schizo during myoblast fusion is still unknown. Rescue experiments with Drosophila Arf1, Arf2 and Arf6 that represent the three classes of mammalian Arf GTPases (Donaldson and Jackson, 2011), suggest that Schizo acts through the Arf1-GTPase (Dottermusch-Heidel et al., 2012). In a global yeast two-hybrid screen the cell adhesion molecule N-cadherin was identified as Schizo interaction partner (Dottermusch-Heidel et al., 2012). Genetic interaction studies revealed that the schizo myoblast fusion phenotype is suppressed by the loss of N-cadherin. Based on these findings we proposed that the removal of N-cadherin brings the apposing myoblast membranes into close proximity to allow membranes to fuse.

In this study, we have analyzed the function of the N-terminal domain of Schizo. We found that the expression of Schizo lacking this N-terminal domain in a wild-type background leads to a nuclear localization of Schizo. However, rescue experiments suggest that this truncation is still able to rescue the schizo myoblast fusion phenotype. Furthermore, we demonstrate that the schizo mutant phenotype can be rescued by the expression of the Sec7-PH domain during muscle development. These findings prove that the Sec7-PH domain serves as a constitutive-active form of Schizo during development. Accordingly, the expression of the Sec7-PH domain in a wild-type background reduces amounts of N-cadherin. To understand by which mechanism amounts of N-cadherin are regulated, we have investigated whether Graf-1 dependent CLIC/GEEC endocytosis is involved in this process. However, neither a graf-1 mutant generated by CRISPR/Cas9 nor the expression of truncated Graf-1 showed schizo-like phenotypes or increased amounts of N-cadherin. Instead, we found that the Scar/WAVE complex member Abi binds to the N-terminal region of Schizo. Abi, but not Scar/WAVE, antagonizes Schizo function in a dosage-dependent manner. These findings provide a new conceptual framework for the regulation of Schizo activity.

schizo encodes for two Arf GEFs (Schizo P1 and Schizo P2) that only differ in the first 12 amino acids at their N-terminal region and share all the conserved domains (Önel et al., 2004). These Schizo proteins correspond to the Loner isoforms Iso1 and Iso2 that have been described by Chen et al. (2003). In Dottermusch-Heidel et al. (2012) we screened a Drosophila yeast two-hybrid cDNA library with the first 753 amino acids of Schizo P2 (Fig. 1A) and identified N-cadherin as interaction partner. To determine whether the N-terminal region of Schizo is important for Schizo P2 localization, we placed the wild-type schizo cDNA LP01489 lacking the first 2,3 kb of the schizo ORF under the control of UAS activating sequences. The full-length Schizo protein P2 Siz_1-1313, the Schizo protein Siz_1-753 from the yeast-two hybrid screen and Siz_753-1313 lacking the first 2,3 kb of the LP01489 schizo cDNA are shown in Fig. 1A. GFP-tagged Siz_1-1313 and Siz_753-1313 were expressed in the mesoderm and in somatic muscle cells with the Mef2-GAL4 driver. In the mesoderm of stage 10 embryos, Siz_1-1313 and N-cadherin are both present at the plasma membrane (Fig. 1B–B″ arrow), but Siz_1-1313 is also detectable in the cytoplasm (Fig. 1B–B″ asterisks and D–D″). Since the N-cadherin antibody detects the extracellular region of N-cadherin, Siz_1-1313 and Ncadherin do not co-localize in Fig. 1B–B″. In contrast, Siz_753-1313 lacking the N-terminal region is only detectable in the cytoplasm and not at the plasma membrane like observed for Siz_1-1313 (Fig. 1C–C″). In somatic muscles Siz_753-1313 is expressed in a very specific pattern that is reminiscent of the nuclear expression pattern of the transcription factor Mef2 (Fig. 1E′ and E″ arrows).

To confirm the different localization of Siz_1-1313 full-length and truncated Siz_753-1313, both proteins were expressed in Drosophila S2R+ Schneider cells (Fig. 1F,F′,G,G′ and Movie 1). Siz_1-1313 is distributed in the cytoplasm of S2R+ cells in a punctuated manner (Fig. 1F and F′). The act that Siz_1-1313 is not found at the plasma membrane is probably due to the overexpression situation. As suspected from the localization in Fig. 1E′ and E″, we observed a co-localization of Siz_753-1313 with DAPi in Drosophila S2R+ cells confirming the nuclear localization of Siz_753-1313 (Fig. 1G and G′). The import of proteins into the nucleus depends on a short peptide sequences called nuclear localization signal (NLS) sequences. To analyze whether Schizo P2 contains such NLS sequences, we have performed a data-bank search as described by Lin and Hu (2013) and identified two predicted NLS sequences in Siz_1-1313 (Fig. 1A). One of these predicted NLS sequences is located in the Sec7 domain of Schizo P2 and is still present in Siz_753-1313. However, the expression of Siz_753-1313 in the mesoderm of wild-type embryos and its nucleocytoplasmic shuttling does not disturb myoblast fusion. Taken together, these results show that the N-terminal region of Schizo P2 is essential for the localization of the Schizo protein to the plasma membrane. In the absence of this region Schizo translocates to the nucleus although the Sec7 and PH domain that mediate lipid binding are still present.

Schizo function is required in the two types of Drosophila myoblasts: founder cells (FCs) and fusion-competent myoblasts (FCMs) (Dottermusch-Heidel et al., 2012), where it interacts with N-cadherin. The loss of Schizo function leads to severe defects in myoblast fusion (Fig. 2A-A″). In the CNS, schizo mutants lack commissural axons (Fig. 2B-B″, arrows) due to increased levels of the axon guidance molecule Slit (Önel et al., 2004). To assess the importance of the Sec7 and PH domain for Schizo P2 function, we generated Siz-ΔSec7 and Siz-ΔPH deletion mutants (Fig. 2C). Expression of Siz_1-1313 in the mesoderm with twist-GAL4 rescues the myoblast fusion phenotype (Fig. 2D and E), whereas homozygous mutants expressing Siz-ΔSec7 or Siz-ΔPH in their mesoderm still showed a myoblast fusion phenotype (Fig. 2F and G). These data confirm that the Sec7 and PH domains are both pivotal for Schizo function.

Studies on the Schizo homologue BRAG2 have demonstrated that the Sec7-PH module possesses a 10-fold higher activity towards Arf1 and a 15–20-fold higher activity in the presence of PIP2 (Jian et al., 2012; Aizel et al., 2013). To investigate whether the Schizo Sec7-PH module serves as an activated form of Schizo, we amplified the coding region of the Sec7-PH module Siz_753-1081 and cloned it under the control of the UAS activating sequences (Fig. 3A). We observed that the expression of Siz_753-1081 in FCs and FCMs with twist-GAL4 as well as the exclusive expression of Siz_753-1081 in FCMs with sns-GAL4 is able to rescue the schizo myoblast fusion phenotype (Fig. 3D and E). Furthermore, we noticed that the overexpression of Siz_753-1081 with Mef-GAL4 in wild-type embryos also disturbs myoblast fusion (Fig. 3H). To determine whether the altered localization of Siz_753-1313 affects Schizo function, we also performed a rescue experiment with Siz_753-1313. We found that the expression of Siz_753-1313 in both myoblast-types with twist-GAL4 partially rescues the schizo mutant phenotype (Fig. 3G). However, the comparison of the dorsal muscle group in wild-type (Fig. 3B, asterisks), schizo^C1-028 (Fig. 3C, asterisk) and rescued embryos with Siz_753-1081 (Fig. 3D and E, asterisk) or Siz_753-1313 (Fig. 3G, asterisk) let suggest that the rescue capacity of Siz_753-1313 differs from the rescue capacity of Siz_753-1081.

Still the question remained whether the Sec7-PH module is crucial for regulating amounts of N-cadherin. To address this question, we first elevated the amount of N-cadherin in homozygous schizo mutants and in embryos expressing dominant-negative Arf1 (Arf1T31N) by using two different approaches. Consistent with our previous hypothesis that was based on genetic interaction studies, we observed increased amounts of N-cadherin in whole-mount embryos (Fig. 4A′,A″,B′,B″,C′ and C″). To analyze the amount of N-cadherin, we used cytofluorograms of single embryos (Fig. 4A″,B″ and C″) and measured the fluorescence intensity from 6 to 15 embryos (Fig. 4D). Additionally, we examined the N-cadherin fluorescence intensity of embryos expressing Sec7-PH_753-1081 in the mesoderm with twist-GAL4 or with Mef-GAL4 in muscle cells. With both GAL4 drivers, we observed decreased amount of N-cadherin (Fig. 4E). Additionally, we validated decreased amounts of N-cadherin by Western blot analysis (Fig. 4F). Taken together, these data show that the Schizo full-length protein is required for the removal of N-cadherin and that Schizo-Sec7-PH_753-1081 represents an activated form of Schizo.

The finding that amounts of N-cadherin are decreased in Schizo-Sec7-PH_753-1081 expressing embryos combined with our previous findings that N-cadherin is increased in the absence of schizo, prompted us to investigate whether Schizo controls N-cadherin amounts by endocytosis. The observation that the schizo mutant phenotype can be rescued by GTP-bound Arf1 suggests that Schizo acts via Arf1 (Dottermusch-Heidel et al., 2012). Consistently, we detect a transient colocalization between Siz_1-1313 and Arf1 in Drosophila S2R+ cells (Fig. S1). The small Arf1-GTPase is involved in the endocytosis of lipid-anchored proteins such as GPI-APs and does not involve Dynamin. The endocytotic structures are termed GEECs (GPI-AP enriched early endosomal compartments). The molecular mechanism of this pathway is initiated by the recruitment of the Sec7 GEF GBF1, which activates Arf1 (Gupta et al., 2009). Arf1 recruits the RhoGAP protein Graf-2 (alias ARHGAP10, alias ARHGAP21) to the cell surface (Kumari and Mayor, 2008). The activity of this protein complex promotes GTP-hydrolysis on Cdc42, which is necessary for the endocytosis process. Additionally, Graf-1 (alias ARHGAP26) has been described to colocalize with activated Cdc42 and controls like Graf-2 Cdc42 activity (Lundmark et al., 2008). During muscle development, the downregulation of Graf-1 in murine C2C12 cells or the loss of Graf-1 or Graf-2 in primary myoblasts significantly reduces the capability of myoblasts to fuse (Doherty et al., 2011; Lenhart et al., 2014). Due to the role of Graf in GEEC endocytosis and its role in mammalian myoblast fusion, we examined Graf function in Drosophila.

Like the mammalian Graf proteins, Drosophila Graf-1 contains a BAR, PH, RhoGAP and SH3 domain (Kim et al., 2017) (Fig. 5A). To determine whether Drosophila Graf-1 is involved in the removal of N-cadherin, we first performed protein interaction studies between Graf-1 and Schizo and the intracellular domain of N-cadherin. Although we did not detect any interaction between Graf-1 full-length and N-cadherin, we found that Graf-1 lacking the BAR domain interacted with the intracellular domain of N-cadherin (Fig. 5A, Table 1, Fig. S2). The BAR domain of mammalian Graf-1 has been shown to directly interact with the GAP domain to inhibit its activity (Eberth et al., 2009). We assume that the failed interaction between Graf-1 full-length and the intracellular domain of N-cadherin is caused by the autoinhibited conformation of Graf-1. Domain mapping experiments revealed that the SH3 domain of Graf-1 is responsible for the interaction of GrafΔBAR with the intracellular domain of N-cadherin (Fig. 5A, Fig. S3). However, we found no protein interactions between Graf-1 and Schizo in the yeast-two hybrid assay. Subcellular localization assays revealed that Graf-1-eGFP is distributed in a punctuated manner in Drosophila S2R+ cells (Fig. 5B and C). Furthermore, we found a partial colocalization between Graf-1, Siz and GTP-bound Arf1 (Fig. 5D and E). To determine whether Graf is involved in the removal of N-cadherin, we used the CRISPR/Cas9 strategy to generate graf mutants. Fig. 5F shows the localization of the guide target (gRNA) within the open reading frame of the graf-1 gene which spans a genomic region of 8.8 kb. The gRNA is located in the third exon of graf-1 which encodes for the BAR domain. The largest deletion we observed comprises 20 base pairs and was termed grafΔ20. The deletion in grafΔ20 leads to a seven amino acid truncation in the BAR domain of Graf-1, which should result in a premature stop codon after 64 amino acid. To investigate whether the Graf-1 protein is still detectable in grafΔ20 mutants we employed the Graf-1 antibody from Kim et al. (2017) and performed cytofluorograms of single embryos. The Graf-1 antibody detects a peptide sequence located between the RhoGAP and the SH3 domain. In wild-type embryos Graf-1 is expressed in muscles (Fig. 5G′, arrowhead) and the central nervous system (Fig. 5G′, arrow). In the grafΔ20 deletion we observed a reduced expression of Graf-1 in muscles (Fig. 5I′). When we analyzed Graf-1 protein levels in comparison to β3-Tubulin, we found that the Graf protein is reduced in grafΔ20 mutants (Fig. 5H and J). Sequence analyses revealed an alternative open reading frame for Graf-1, which might explain the reduced protein levels. Homozygous grafΔ20 mutants were viable and showed no defects in myoblast fusion (Fig. 5G and I). In Kim et al. (2017) a graf-null mutant was generated by imprecise P-element excision. Homozygous graf-null mutants are viable and fertile. In addition, we quantified amounts of N-cadherin in wild-type embryos expressing UAS-grafΔBAR and UAS-grafΔBARΔSH3 with Mef-GAL4 (Fig. 5K). However, we failed to observe increased amounts of N-cadherin. Based on these findings, we propose that Graf-1-dependent endocytosis does not play a major role in the removal of N-cadherin from the plasma membrane.

The actin polymerization machinery has been implicated to aid in clathrin-independent endocytosis (Chadda et al., 2007; Römer et al., 2010; Sathe et al., 2018). Indeed, actin polymerization seems to be a key step to power local membrane deformation and carrier budding in clathrin-independent endocytosis (Hinze and Boucrot, 2018). Since Arp2/3- and Formin-dependent F-actin polymerization is essential for the fusion of myoblasts (Kim et al., 2007; Massarwa et al., 2007; Richardson et al., 2007; Schäfer et al., 2007; Berger et al., 2008; Deng et al., 2015), we determined whether members of the Arp2/3 activation machinery or the formin Diaphanous interact with Schizo (Table 1). Surprisingly, we identified that Abi, a component of the Scar/WAVE complex, interacts with Schizo and Graf-1 (Table 1, Fig. 6 and Fig. S4).

We next mapped the interaction domain on Abi by testing deletion mutants of Abi for binding to full-length Siz_1-1313. Full-length Abi and the deletion mutant proteins examined are depicted schematically in Fig. 6A. The deletion of the carboxy-terminal SH3 domain eliminated the interaction with Schizo Siz_1-1313 (Fig. 6A). Subsequently, we performed the converse experiment and only used the SH3 domain of Abi for determining the interaction with Schizo (Fig. 6A and C). These data indicate that the SH3 domain is mandatory for the observed interaction. To identify the interaction domain on Schizo, we used the deletion mutant proteins illustrated in Fig. 6B. Abi full-length interacts with Siz_1-323, Siz_753-972, Siz_972-1313 and Siz_1-753. As demonstrated in the controls, Siz_972-1313 shows a false positive interaction with the empty pBGKT7 vector. By using the SH3 domain of Abi as bait protein, we could confirm the interaction of Siz_1-323, Siz_972-1313 and Siz_1-753, but not with Siz_753-972. From these data, we concluded that the N-terminal region of Schizo from 1–753 amino acid is responsible for interacting with Abi. Because of the observed protein interaction, we next transfected Drosophila S2R+ cells with UAS-siz-mcherry and UAS-abi-eGFP to determine the distribution of both proteins in Drosophila S2R+ cells. We found that UAS-abi-eGFP colocalizes with full-length UAS-schizo-mcherry and that both proteins are distributed in a punctuated manner (Fig. S5 and Movie 2).

The observation that Abi and Schizo colocalizes in a punctuated manner raises the question whether Abi and Schizo act in concert to regulate amounts of N-cadherin by endocytosis. To determine whether both proteins contribute to the same process, we performed epistasis experiments and used meitotic recombination to generate schizo abi-double mutants. In a first attempt we analyzed the muscle phenotype of the hypomorphic schizo allele siz^C1-028, the abi-null allele abi^Δ20 and siz^C1-028 abi^Δ20-double mutants (Fig. 7Aa–e). We observed that homozygous siz^C1-028 abi^Δ20 double mutants show a strong myoblast fusion phenotype like siz^C1-028 (Fig. 7Ad and Ab). The muscle phenotype of abi^Δ20-null mutants is comparable to the wild-type muscle pattern (Fig. 7Ac and Aa), but some muscles are missing. If schizo and abi both contribute to regulation of N-cadherin, we expected to enhance the abi muscle phenotype by reducing the schizo gene dose. However, the musculature of transheterozygous abi^Δ20/Df(abi) mutants lacking one copy of schizo looks like the musculature of homozygous abi^Δ20 mutant embryos (Fig. 7Ae and Ac). From the phenotypes, we cannot conclude whether schizo and abi contribute together to regulating N-cadherin. Therefore, we have determined amounts of N-cadherin in homozygous siz^C1-028, abi^Δ20/Df(abi) single and siz^C1-028 abi^Δ20 double mutants. Fig. 7Af shows the measurement of the N-cadherin fluorescence intensity from 5 to 12 embryos. Interestingly, we found that amounts of N-cadherin are reduced to wild-type levels in homozygous siz^C1-028 abi^Δ20 double-mutant embryos. The reduced amounts of N-cadherin in the double mutants indicate that schizo and abi might act antagonistically. Further support for this notion comes from the analyses of the central nervous phenotype with anti-N-cadherin (Fig. 7Ba–f). N-cadherin is not only expressed in the mesoderm and during myoblast fusion, but also shows a strong expression in the ventral nerve cord (Iwai et al., 1997). When we imaged siz^C1-028 abi^Δ20 double mutants to determine the fluorescence intensity of N-cadherin, we noticed that the commissural phenotype of homozygous schizo mutants (Fig. 7Bb arrowheads) is suppressed in some hemisegments of siz^C1-028 abi^Δ20 double mutants (Fig. 7Bd arrowheads).

Abi is a member of the Scar/WAVE complex, which is required for the activation of the Arp2/3 complex that nucleates branched F-actin filaments (Rotty et al., 2013). The nucleation ability of the Arp2/3 complex depends on nucleation promoting factors of the WASp family to which Scar/WAVE belongs to (Tyler et al., 2016). Besides its function as member of the Scar/WAVE complex, Abi is also known to interact physically with WASp (Bogdan and Klämbt, 2003). The finding that abi antagonizes schizo during myoblast fusion and commissural formation suggests that also other members of the Scar/WAVE complex, e.g. scar/wave counteract schizo. To test this assumption, we generated scar siz double mutants by using the hypomorphic scar allele scar^k13811. Commissures are reduced in the siz alleles siz^C1-028 (Figs 2B″ and 7Bb arrowheads). In homozygous scar^k13811 mutants, commissures are not reduced, but both commissures are sometimes observed in close proximity (Fig. 7Be arrowheads). In contrast, we found that the commissural phenotype of homozygous siz^C1-028 mutant embryos is clearly enhanced in homozygous scar^k13811 siz^C1-028 double-mutant embryos (Fig. 7Bf, arrowheads). Taken together, these data suggest that only abi counteracts schizo, but not scar/wave.

In Drosophila about 60% of the genome is maternally contributed as mRNA to ensure embryonic development (De Renzis et al., 2007; Lécuyer et al., 2007). The mRNA of abi and scar are maternally transcribed (Zallen et al., 2002; Lin et al., 2009). We found that the schizo mRNA is still detectable in the schizo deficiency Df(3L)ME178 at embryonic stage 11, when myoblasts start to fuse and at stage 14 (Fig. S6). This suggests that during embryogenesis, reduced protein levels of Abi are required to antagonize Schizo function, but not Scar/WAVE. To support this notion, we overexpressed Schizo in photoreceptor cells in adult flies that have no maternal mRNA using the eye-specific GMR-GAL4 driver line. Expression of UAS-siz leads to a rough-eye phenotype (Fig. 7Cb). This rough-eye phenotype is suppressed in flies heterozygous for the abi^Δ20 mutation (Fig. 7Cc). Consistently with our previous results, the rough-eye phenotype is not suppressed in flies heterozygous for the hypomorphic scar allele scar^k13811 (Fig. 7Cd). In addition, we observed no suppression of the rough-eye phenotype in flies heterozygous for the rac null allele rac1^J11, which is another member of the Scar/WAVE complex (Fig. S7B). Moreover, we examined whether the Abi interacting partner WASp is able to antagonize Schizo function. However, we found that the reduction of the wasp dosage using the dominant-negative EMS allele wasp^3D3-035 enhances the GMR-GAL4>>UAS-siz induced rough-eye phenotype (Fig. S7C). In summary, these experiments confirm our hypothesis that only reduced levels of Abi antagonize Schizo function.

Understanding the multiple regulatory layers of GEF activation that allows coordinating the GDP/GTP exchange on small GTPases is an important issue. To date, the multitude of molecular interactions leading to GEF activation are most advanced for the Ras activator Son of Sevenless (Bandaru et al., 2019). Studies on the subfamily of Arf GEFs that carry a PH domain associated with a catalytic Sec7 domain have concentrated on the interaction of the GEF with Arf GTPases and phospholipids. Membrane recruitment of Arf GTPases is mediated by a myristoylated N-terminal amphipathic helix (Franco et al., 1996; Goldberg, 1998; Liu et al., 2009) and is crucial for the activation by the GEF (Pasqualato et al., 2001; Randazzo et al., 1995). The PH domain binds phophatidyl inositol 3,4,5-triphosphate (PIP3) and phosphatidyl insositol 4,5-bisphosphate (PIP2) (Chardin et al., 1996; Kavran et al., 1998; Klarlund et al., 2000). In structural studies with myrArf/BRAG2 bound to a PIP2-containing bilayer, the myristoylated N-terminal helix of Arf is close to the Sec7 domain and it has been proposed that the Sec7 domain might recognize conformational information from the amphipathic helix (Karandur et al., 2017). However, so far, the influence of receptor binding for Arf GEF activation has not been taken into account.

In this study, we found that N-cadherin levels were elevated in mutants of the Arf1 GEF schizo and reduced in embryos expressing only the Sec7-PH domain of Schizo. These findings are in line with studies on mammalian GEP100/BRAG2. The siRNA-mediated depletion of GEP100/BRAG2 in HepG2 cells resulted in increased E-cadherin content (Hiroi et al., 2006). Furthermore, increased amounts of β1-integrin were observed in the ‘knock-down’ of BRAG2 in HeLa cells and it was proposed that BRAG2 served specifically for β1-integrin internalization (Dunphy et al., 2006). Rescue experiments with the Sec7-PH domain and overexpression studies further support the notion that it acts as a constitutively active form of Schizo during myoblast fusion.

The finding that the member of the BRAG2 subfamily Schizo interacts with N-cadherin via its N-terminal region let us investigate the importance of this region for Schizo function. Our data imply that the N-terminal region provides additional layers of GEF regulation. First, we identified that Schizo undergoes nucleocytoplasmic shuttling in the absence of the N-terminal region. A nuclear localization for mammalian BRAG2a and BRAG2b was also detected in HeLa and MDCK cells after overexpression (Dunphy et al., 2006). BRAG2 possesses like Schizo a nuclear localization sequence in the Sec7 domain. However, the finding that the expression of Siz_753-1081 partially rescues the schizo mutant phenotype let suggest that nuclear localization does not disturb Schizo function. Furthermore, the overexpression of Siz_753-1081 does not result in myoblast fusion defects as observed for Siz_753-1313 suggesting that it does not act as a constitutive-active form. Second, we found that Abi binds to the N-terminal region of Schizo. Abi has been reported to regulate actin polymerization by formation of complexes with Scar/WAVE and WASp (Innocenti et al., 2005). Furthermore, it has been described to modulate EGFR endocytosis (Tanos and Pendergast, 2007). Besides, Abi-1 has been identified to interact with the Ras activator Son of Sevenless (Scita et al., 1999; Fan and Goff, 2000).

Vesicle scission in Clathrin-independent endocytosis depends on specialized actin-based platforms (Doherty and McMahon, 2009; Mayor et al., 2014). However, there is no unifying theme and multiple mechanisms may co-exist. The molecular machinery of CLIC/GEEC-dependent endocytosis involves the activation of Arf1 by the Arf GEF GBF1 and the RhoGAP protein Graf that removes Cdc42 from the plasma membrane (Kumari and Mayor, 2008; Gupta et al., 2009). A recent study has addressed the spatiotemporal localization of known molecules affecting CLIC/GEEC endocytosis by using real-time TIRF microscopy (Sathe et al., 2018). In this study it was reported that Arp3 recruitment occurred earlier to endocytic vesicles than Cdc42. Furthermore, N-WASp failed to recruit to form CLIC/GEEC endocytotic sites. These data imply alternative pathways for Arp2/3 activation in CLIC/GEEC endocytosis. In Drosophila, WASp lacking the Cdc42-binding domain (WaspΔCRIB) is still able to rescue the adult phenotype of wasp mutants and it was proposed that other elements than Cdc42 contribute to Drosophila WASp activation (Tal et al., 2002). Such an alternative element might be Abi. However, our genetic interaction studies suggested that Abi counteracts Schizo function.

Dissecting the function of Arp2/3-dependent endocytosis during myoblast fusion is challenging since the fusion of myoblasts depends on Scar/WAVE- and WASp-dependent Arp2/3 activation (Kim et al., 2007; Massarwa et al., 2007; Richardson et al., 2007; Schäfer et al., 2007; Berger et al., 2008). This might explain why the myoblast fusion phenotype in abi siz double mutants is not suppressed although amounts of N-cadherins are reduced. Therefore, we overexpressed Schizo in photoreceptor cells and performed gene dose experiments and found that only the reduction of abi suppressed the Schizo-induced overexpression phenotype.

In Salmonella host cell invasion, the Arf6 GTPase was shown to affect indirectly actin polymerization by activating the GEF ARNO (Humphreys et al., 2013). Arf6 is activated by EFA6 or BRAG and recruits the autoinhibited GEF ARNO to the plasma membrane. The autoinhibition of ARNO is released by the binding of activated Arf6 to its PH domain. As a consequence, ARNO activates Arf1, which induces together with Rac1 the activation of the WAVE Regulatory Complex (Koronakis et al., 2011; Humphreys et al., 2013). Since we found no evidence that scar/wave or wasp exert a regulatory influence on Schizo, we propose that Abi might be part of a dosage-dependent regulatory feedback mechanism following Arp2/3-dependent actin polymerization.

The data presented in this study point towards a novel model for the regulation of the Arf1 GEF Schizo (Fig. 8). First, we suggest that amounts of N-cadherin are regulated by the Sec7 GEF Schizo and that the N-terminal region of Schizo is crucial for the regulation of its activity (Fig. 8A). The Abl-interacting partner Abi might compete with N-cadherin for Schizo binding and disturbs the removal of N-cadherin. However, since the expression of the constitutive-active form of UAS-Sec7-PH decreases amounts of N-cadherin in the absence of the N-terminal region, we propose that Abi acts rather as an allosteric modulator by either inhibiting the binding of myristyolated Arf1 (Fig. 8B) or by inhibiting the binding of the Sec7-PH module to the lipid bilayer (Fig. 8C). A second regulatory mechanism to control Schizo activity might involve the nuclear localization sequence within the GEF domain in the absence of the N-terminal region. The data reported in our study will be valuable for future structural analyses to determine how Abi prevents GEF activity.

Generation of pUASt-attB-eGFP-siz1-1313, pUASt-attB-eGFP-siz753-1313 flies and siz-Sec-PH

The siz1-1313 and siz753-1313 construct were amplified from the LP01489 cDNA by PCR, subcloned into the pENTRTM/D-TOPO vector and recombined into the Gateway vector pUASt-attB-rfa-eGFP. We used the following primer pairs:

siz753-1313_f: 5′-CACCATGGAGACGATACGCAAG-3′

siz753-1313_rev: 5′-TTAGACCTCCGTCGACCGT-3′

sizSec7-PH_f: 5′-CACCATGTCGGAGAC-3′

sizSec7-PH_rev: 5′-GAGATGGAGTCGTGA-3′

y1cho2v1; graf mutants were generated using a single target construct in the CRISPR/Cas9 system as described by Kondo and Ueda (2013). For cloning of the target construct, primers 5′-CTTCGGTCAAAGATCTTATGAGTG-3′ and 5′-AAACCACTCATAAGATCTTTGACC-3′ were used, and the product was ligated to pBFv-U6.2. The target construct was injected into y2cho2v1P{nos-phiC31\int.NLS}X;attP2(III), and established transgenic flies were crossed with y2cho2v1;attP40{nos-Cas9}/CyO flies for mutagenesis. Twenty-three founder flies were crossed with y2cho2v1; Sco/CyO flies to establish potential mutants. Genomic regions next to the target sequence of homozygous offspring were analyzed by PCR and screened for deletions. We identified three different deletions and an additional sequence insertion.

The graf-ΔBAR and pUASt-ΔBARΔSH3 constructs were amplified from the full-length graf LD28528 cDNA obtained from DGRC. It should be noted that the cDNA clone reported as fully sequenced in flybase, contains an additional nucleotide at position 467, which leads to a shift of the open reading frame. As a consequence, the deduced protein from this cDNA lacks the BAR domain. To generate full-length Graf, we performed a site-specific mutagenesis on the LD28528 cDNA to remove the additional nucleotide.

abi germline clones were produced by heat-shock in hs-Flp; FRT82B ovo^D/FRT82B abi^Δ20 larvae. The resulting females were crossed to males of the abi deficiency Df(3R)BSC617/TM3-Dfd-lacZ. This allowed the detection of a wild-type parental contribution on the basis of β-galactosidase.

The hypomorphic schizo EMS allele siz^C1-028 was generated by Hummel et al. (1999) and the CNS and muscle phenotype of siz^C1-028 was characterized and described in Önel et al. (2004) and Dottermusch-Heidel et al. (2012). UAS-Arf1T31N-eGFP was generated and described in Dottermusch-Heidel et al. (2012). abi^Δ20 mutants were kindly provided by Sven Bogdan (Department of Molecular Cell Physiology, Institute for Physiology and Pathophysiology, Philipps-Universität Marburg, Germany) (Stephan et al., 2011). The twist-GAL4 driver line SG24 was obtained from the Bloomington Stock Center. As abi deficiency we used the deficiency line Df(3R)BSC617 (BL25692) from the Bloomington Stock Center. Further GAL4 driver lines that were used in this study are Mef-GAL4 (Ranganayakulu et al., 1996) and TGX twist-GAL4 (from A. M. Michelson, National Institutes of Health (NIH), National Heart, Lung, and Blood Institute (NHLBI), USA). As blue balancers we used Dr/TM3 Dfd-lacZ and If/CyO hg-lacZ. All crosses were performed at 25°C using standard methods.

Yeast two-hybrid experiments were carried out using the Matchmaker GAL4 Two-Hybrid System 3 (Takara Clontech) according to the manufacturer's instructions. For construct generation, the schizo cDNA LP01489 from DGRC was used. The following primers were used to generated the different schizo and graf constructs:

siz1-323 5′-CATATGATGTCCAGGTGTGA-3′ and 5′-GGATCCTCGCACTCCGC-3′

siz1-758 5′-CATATGATGTCCAGGTGTGA-3′ and 5′-

siz324-753 5′-CATATGATGGCCCGTAACG-3′ and 5′-GGATCCTATCGTCTCCGAC-3′

siz754-993 5′-CATATGATGCGCAAGCGAC-3′ and 5′-GGATCCCACACCAGGTCG-3′

siz994-1313 5′-CATATGATGCATCAGCGCG-3′ and 5′-GGATCCTTAGACCTCCGTC-3′

siz324-993 5′-GAATTCATGGCCCGTAACGCA-3′ and 5′-CTCGAGCACACCAGGTCG-3′

graf 5′-CATATGATGGGCGGCGGCAAAAAT-3′ and 5′-GGATCCTAATGGT GCGGCTTCAAAT-3′

GrafΔBAR 5′-CATATGATGTCAACTAAAAAGCCCGAA-3′ and 5′-GGATCCCTAATGGTGCGGCTTCAAAT-3′

GrafΔBARΔPH 5′-CATATGCTGGCTCCCGGCA-3′ and 5′- GGATCCCTAATGGTGCGGCTTCAAAT-3′

grafΔBARΔSH3 5′-CATATGATGTCAACTAAAAAGCCCGAA-3′ and 5′-GGATCCGGTGCCCGTTGA-3′

grafΔBARΔPHΔRhoGAP 5′-CATATGAGCGCCGATATCAA-3′ and 5′-GGATCCCTAATGGTGCGGCTTCAAAT-3′

The products were cloned into the pCRII-TOPO^® vector (Invitrogen). The bait vector pGADT7 was digested with NdeI and BamHI. siz was cloned with EcoRI and XhoI into the pGADT7.siz was cloned with NdeI and EcoRI into the pGADT7.

The pGADT7-T and pBGKT7-p53 pair was used as a positive control. The candidate interaction pairs were co-transformed into yeast strain AH109, and the transformed yeast cells were selected using synthetic dropout (SD/-Leu/-Trp) medium, and then further selected on SD/-Leu/-Trp-His/-Ade selective medium with X-α-Gal (80 mg/l). Results were obtained after 2 days of growth at 30°C.

Drosophila S2R+ cells obtained from the Drosophila Genomic Rescource Center (DGRC) were propagated in 1× Schneider's Drosophila medium (Invitrogen) containing 10% fetal bovine serum at 25°C, and transiently transfected described by Kaipa et al. (2013) by using the FuGENE^® HD Transfection Reagent (Promega).

Embryos were fixed and immunohistochemically analyzed as described by Schäfer et al. (2007). The following antibodies were used at the noted dilutions: rat anti-CadN-Ex#8 (Iwai et al., 1997) 1:50 (Developmental Studies Hybridoma Bank), guinea pig anti-β3Tubulin (Buttgereit et al., 1996; Leiss et al., 1988) 1:10,000, rabbit anti-β-Gal 1:5000 (Biotrend), rabbit anti-GFP 1:500 (abcam ab5665). Primary antibodies were detected using the fluorescent labeled antibodies Alexa-Fluor-488, 568- or 647-conjugated anti-guinea pig, anti-rabbit and anti-rat IgG at a dilution of 1:500 (Invitrogen). DNA was stained with Hoechst reagent (5 g/ml; Sigma-Aldrich), and F-actin was stained with Alexa-Fluor-647–phalloidin (1:100, Invitrogen). For all stainings, specimens were embedded in Fluoromount-G™ (Thermo Fisher Scientific) and observed under a Leica TCS Sp2, TCS Sp5 or TCS Sp8 confocal microscope.

Statistical analyses were performed with GraphPad Prism software (PRISM7). For Figs 4, 5 and 7 significance was determined by one-way ANOVA with comparison.

A Leica TCS Sp2 (Fig. 1B–E″), TCS Sp5 (Fig. 3A–F) and TCS SP8 (Figs 1F–G′, 2, 4, 5 and 7) confocal microscope was used for fluorescence imaging. The same parameter settings were used to image all samples of the same type. Embryos were embedded in Fluoromount-G^TM (Thermo Fisher Scientific) and scanned using the 20× objective with the galvo scanner at 400 Hz. For the fluorescence intensity measurement, we used the counting modus of the TCS SP8. Laser intensity for imaging β3-Tubulin stainings with Alexa 488 were always set to 0.8%. Laser intensity for imaging N-cadherin stainings with Alexa 568 were always set to 4%. The raw data of the embryo images were analyzed using the FIJI software (Schindelin et al., 2012). A SUM-stack was generated in Fiji and the mean fluorescence intensity of lateral imaged embryos were measured by manually drawing a circle around the embryo to select the area to be measured. Additionally, three nearby areas were also measured to analyze the background fluorescence level. The corrected total relative intensity of each embryo was calculated as follows:

(Areaembryo×Mean–Intensityembryo) – (Areaembryo×Background)

To raise cytofluorograms of single embryos we imaged embryos as we did for fluorescence intensity measurements. We employed the JACoP plugin to generate the cytofluorograms (Bolte and Cordelières, 2006), which was downloaded from https://imagej.net/plugins/jacop.

For time-lapse imaging, the Spinning disc microscope from Zeiss was used with the 63× objective.

Embryos were collected on apple juice plates, dechorionated (using 50% bleach solution in water) and selected under the GFP stereomicroscope from Leica for GFP expression. For each protein extract 10 embryos of stages 15 to 16 were selected and punctured with an inulin syringe. The resulting lysate was mixed with 10 µl of 1x Lammeli buffer (1 µl per embryo lysate) and transferred into an Eppendorf tube as described by Prudêncio and Guilgur (2015). The samples were heated for 5 min at 98°C and subsequently loaded to a 12% SDS-PAGE gel. Following separation, proteins were transferred to a nitrocellulose membrane (Whatman) by semi-dry blotting (Mini PROTEAN Tetra System; Trans-Blot-turbo Transfer System, BIO-RAD). The membrane was blocked in TBST-blocking buffer [50 mM Tris–HCl pH 7.5, 150 mM NaCl and 0.5% (v/v) Tween 20] containing 5% nonfat dried milk for at least 30 min. The following primary antibodies were used: anti-CadN-Ex#8 (1:100), anti-Actin (Millipore MAB1501, 1:2000). The antibodies were removed by washing three times for 10 min each in blocking solution and the secondary antibodies anti-mouse-HRP (sc-516 102, Santa Cruz Biotechnology, 1:5000) and anti-rat-HRP (Thermo Fisher Scientific #31407, 1:10.000) were applied for 1 h at room temperature. Chemiluminescence was detected using SuperSignal West Dura Extended Duration Substrate (Thermo Fisher Scientific) and Odyssey Fc Imaging System (LI-COR Bioscience) and analyzed by Image Studio Software (LI-COR). The quantitative analysis was performed by Image Studio Software (LI-COR) analysis tool. A volume box of same area was drawn around each protein band to measure the signal intensities respectively. The final signal intensity was calculated by subtracting the Median Local background signal. The ratios of signal intensity values for N-Cadherin to Actin for both wild-type and Mef>>siz-Sec7 PH were calculated and normalized to represent them as a bar graph.

All GMR-GAL4>UAS-siz-expressing flies were raised at 25°C. Eyes were fixed in 6% glutaraldehyde and 1% formaldehyde in 0.2 M Hepes buffer for 16 h. Samples were then dehydrated in a 25%, 50%, 70% and 96% ethanol series for 12 h each and finally transferred into acetone by three 10-min changes with 100% acetone. The samples were critical-pointdried by using a Polaron E 3000 (Balzers Union). Samples were attached to sample stubs (Plano GmbH) and sputtered with gold under vacuum using a sputter coater (Balzers Union, Lichtenstein). Scanning electron micrographs of adult fly eyes were taken using a Hitachi S-530 SEM.

We are grateful to Sven Bogdan for providing Abi full-length, AbiΔSH3 and AbiΔN (Bogdan et al., 2005) and to Seungbok Lee for providing the Graf antibody. We thank Renate Renkawitz-Pohl, Anne Holz and Joanne M. Britto for fruitful discussions and carefully reading the manuscript. We thank Sabina Huhn for technical assistance and Lars Kneifert for perfoming the site-pecific mutagenesis on the graf cDNA LD28528 during his bachelor work. Furthermore, we thank the students of the lab2venture project from the Herder school in Gießen that performed a Schizo modifier screen in the Drosophila eye and discovered Abi as a Schizo suppressor: Sophie Dönges, Inga Interwies, Anne Mack, Fiona Metsch, Tobias Post and Katrin Stör.