Extramacrochaetae promotes branch and bouton number via the sequestration of daughterless in the cytoplasm of neurons

The Class I basic helix–loop–helix (bHLH) proteins are highly conserved transcription factors that are ubiquitously expressed. A wealth of literature on Class I bHLH proteins has shown that these proteins must homodimerize or heterodimerize with tissue‐specific HLH proteins in order to bind DNA at E‐box consensus sequences to control tissue‐specific transcription. Due to its ubiquitous expression, Class I bHLH proteins are also extensively regulated posttranslationally, mostly through dimerization. Previously, we reported that in addition to its role in promoting neurogenesis, the Class I bHLH protein daughterless also functions in mature neurons to restrict axon branching and synapse number. Here, we show that part of the molecular logic that specifies how daughterless functions in neurogenesis is also conserved in neurons. We show that the Type V HLH protein extramacrochaetae (Emc) binds to and represses daughterless function by sequestering daughterless to the cytoplasm. This work provides initial insights into the mechanisms underlying the function of daughterless and Emc in neurons while providing a novel understanding of how Emc functions to restrict daughterless activity within the cell.

synaptic branching and synapse number (D'Rozario et al., 2016). We have also demonstrated that the Da mammalian homolog, transcription factor 4 (Tcf4), functions similarly (D'Rozario et al., 2016). In order to mediate this synaptic restriction, Da and Tcf4 bind to cis regulatory regions of the neurexin gene and repress neurexin transcription in neurons and not in proliferating precursor cells.
Here, we report that part of the molecular logic that specifies how Da functions in mitotically active cells is conserved in neurons via the binding and subsequent repression of Da function by the Type V HLH protein extramacrochaetae (Emc). In mitotic cells, Emc serves as a negative feedback regulator of da expression and Da activity (Campuzano, 2001;Garrell & Modolell, 1990;Spratford & Kumar, 2015b;van Doren, Ellis, & Posakony, 1991). Emc lacks a basic domain, allowing it to heterodimerize with Da, but prevents Da from binding DNA and dimerizing to other transcription factors (Cabrera, Alonso, & Huikeshoven, 1994;Ellis, Spann, & Posakony, 1990). Recently, Spratford and Kumar have shown that Emc does this by sequestering Da away from DNA, thus preventing Da-mediated transcription (Spratford & Kumar, 2015b). Here, we expand upon this observation, and show that Emc accomplishes this sequestration predominantly by sequestering Da within the cytoplasm of neurons. We have determined that Emc is present in differentiated motor neurons of the Drosophila neuromuscular junction (NMJ), where it functions to promote axonal arborization and synapse number in these neurons. Emc accomplishes this by binding and inhibiting Da from repressing neurexin expression. Taken together, these data lend insight into the molecular functions of bHLH proteins in mature neurons and begins to tease apart the molecular logical that bHLH proteins use to control novel functions in postmitotic cells. Further, these data may help shed light on diseases associated with HLH proteins in humans.

| Emc is present in neurons and promotes axonal branching and bouton number
In order to begin our analysis, we first wanted to determine whether Emc is expressed within differentiated neurons. We expressed a membrane-tagged GFP (UAS:mCD8-GFP) with the elav-Gal4 driver to exclusively label differentiated neurons at this stage of development (Robinow & White, 1988Yao & White, 1994). Within the ventral nerve cord (VNC) of third instar larvae we observed extensive Emc protein staining by immunohistochemistry in both neurons (labeled with GFP, arrow in Figure 1a) and nonneuronal cells (arrowhead in Figure 1a). These data indicate that Emc is present in both the nucleus and cytoplasm of neurons.
We next wanted to determine the function of Emc in neurons. To begin to do this, we knocked down and overexpressed Emc specifically in neurons driving an overexpression reagent (UAS:emc) to overexpress Emc and RNA interference (RNAi) to knockdown Emc expression by driving two different short hairpins targeting emc under UAS control (UAS:emcRNAi 1 ; and UAS:emcRNAi 2 ) with Dicer elav- Gal4 (Dcr2;;. In order to validate these reagents, we performed western blot analysis of Emc protein in third instar larval brains (Figure 1d). We observed a significant decrease in Emc protein level using both independent RNAi lines targeting Emc (~35% using UAS:emcRNAi 1 and ~50% using UAS:emcRNAi 2 ) compared to wild-type controls (Figure 1d,e). Additionally, we observed an increase in Emc protein level (~150%) when overexpressed compared to wildtype controls (Figure 1d,e).
To determine Emc function in neurons, we used the welldefined innervation of the third instar NMJ, Muscles 6 and 7 of abdominal Segment 3, as a model to assess axonal branching and bouton number. We found that Emc overexpression in neurons leads to increased branching and bouton number compared to controls (Figure 2a,b, and e), while knockdown of Emc in neurons leads to decreased branching and boutons compared to controls (Figure 2a,c, and e). This is the exact opposite phenotype we previously observed with knockdown and overexpression of Da in neurons, respectively (D'Rozario et al., 2016), and is consistent with Emc operating to repress Da function in neurons as it does in mitotic cells. Based on these data, we decided to continue our analysis using UAS:emcRNAi 2 .

| Emc represses Da function in differentiated neurons
Given the striking NMJ phenotypes observed in larvae expressing Emc knockdown and overexpression constructs, we next wanted to determine the interplay between Emc and Da in the control of branch and bouton number in differentiated neurons. We hypothesized that if Emc is functioning to inhibit Da in neurons, then we would predict that altering Emc levels would have the opposite effect as the same alteration of Da levels on axonal branching and synapse number. For example, phenotypes associated with decreased Emc levels should mimic phenotypes associated with increased Da levels. To further explore this, we co-expressed four combinations of transgenes using the pan-neural driver to alter Emc and Da expression: (a) Emc knockdown with Da knockdown, (b) Emc overexpression with Da overexpression, (c) Emc knockdown with Da overexpression, and (d) Emc overexpression with Da knockdown. We observed that compared to the decreased branches and boutons in Emc knockdown alone (Figure 2c (Figure 2f). Interestingly, these animals did not display significantly altered branch number compared to controls (Figure 3g), indicating that this manipulation can rescue branching phenotypes associated with the knockdown of either Emc or Da alone.
Conversely, we observed that compared to the increased branches and boutons we observed in Emc overexpression alone (Figure 2b), or the decreased branches and boutons in Da overexpression alone (Figure 3b and D'Rozario et al., 2016), Emc and Da double overexpression animals displayed no significant difference compared to control animals in either the number of boutons or the number of branches (Figure 3e,g). This suggests that increasing Emc and Da together is not as deleterious as increasing each one individually.
Interestingly, Da overexpression coupled with Emc knockdown was lethal. However, Da knockdown coupled with Emc overexpression produced animals that mimicked both an Emc overexpression alone and Da knockdown alone phenotype ( Figure 3f) suggesting a potential saturation of Emc levels on axonal branching and bouton number. These animals showed increased branches and boutons compared to controls, which is consistent with Emc overexpression and Da knockdown alone (Figures 2f and 3g). Taken together, these data suggest that the relative balance between Emc and Da in neurons is important for proper branch and bouton number and that either too much Da or Emc affects neuronal wiring at the NMJ negatively.

| Altered Emc or Da expression affects motor behavior
To determine if altered Emc protein levels in differentiated neurons has any effect on motor function (the behavior controlled by VNC motor neurons), we analyzed the behavioral output of the transgenic larvae via measurement of larval contraction rate. Larval body wall contraction relies on proper glutamatergic neurotransmission and is significantly affected by mutations that affect transmission (Sandstrom, 2004). Further, we have previously shown that altered Da expression in differentiated neurons leads to synaptic transmission defects (D'Rozario et al., 2016). Based on these data, we hypothesized that altered Emc levels would also lead to a defect in larval contraction rate. We observed a significant decrease in contractions when we knocked down or overexpressed Emc in differentiated neurons compared to controls ( Figure 4). We were able to completely rescue this defective behavior when we simultaneously knocked down and overexpressed Emc ( Figure  4). As previously reported, we also observed a significant decrease in motor behavior when we knocked down and overexpressed Da in neurons ( Figure 4) (D'Rozario et al., 2016). However, in each of the Emc/Da double transgenic animals (Emc knockdown/Da knockdown; Emc overexpression/Da overexpression; Emc overexpression/Da knockdown), we observed a significant decrease in motor behavior relative to controls. However, we observed a significant increase in contractions in Emc knockdown/Da knockdown and Emc overexpression/Da overexpression relative to Emc knockdown and Emc overexpression alone, respectively ( Figure 4). Taken together, these data suggest that proper Emc and Da function is required in neurons for proper motor behavior. Further, although alteration of both Da and Emc together showed minimal effects on NMJ bouton and branch number, there is still a significant negative effect on motor behavior when these proteins are altered.

Da primarily in the cytoplasm of differentiated neurons
Emc belongs to the Class V family of HLH proteins which lacks a basic DNA binding domain, but still contains an HLH dimerization domain, and therefore is thought to repress Da function by preventing Da from binding DNA and controlling transcription. Previous work from multiple labs Scale bar represents 10 μm. Error bars represent standard error. *p < .05 compared to wild-type control. NS is not significant compared to wildtype control | 809 have shown that Da and Emc interact with each other in a number of mitotic cells (Alifragis, Poortinga, Parkhurst, & Delidakis, 1997b;Georgias, Wasser, & Hinz, 1997;Murre, McCaw, Vaessin, et al., 1989;Spratford & Kumar, 2015b). Based on these data, we hypothesized that Emc is inhibiting Da function in neurons via a physical interaction. To assess whether Da and Emc physically interact in differentiated neurons specifically, we utilized the proximity ligation assay (PLA) with established Da and Emc antibodies (Brown, Sattler, Paddock, & Carroll, 1995;Cronmiller & Cline, 1987;D'Rozario et al., 2016;Spratford & Kumar, 2015a, 2015b. We expressed each transgene using the elav-Gal4;UAS:mCD8-GFP driver to mark differentiated neurons. Surprisingly, we observed a majority of PLA signal in the cytoplasm of these neurons in wild-type conditions (arrows in Figure 5a-d and n) compared to no primary antibody control (Figure 5e-h). We observed decreased PLA signal upon knockdown of both Emc and Da ( Figure 5m), but no significant effect on PLA signal upon overexpression of either Emc or Da ( Figure 5m). We also observed that while a majority of PLA signal was cytoplasmic in a majority of the genotypes we tested ( Figure   5n), a significant amount of PLA signal became nuclear upon overexpression of Da (Figure 5i-l and n). Taken together, these data suggest that Da and Emc also physically interact in neurons, but primarily do so in the cytoplasm of cells in wild-type neurons.

| Emc subcellular localization is dependent upon Da expression
Given that we observed an alteration of PLA signal upon Da overexpression, we next wished to determine if the subcellular localization of Emc was dependent upon the expression levels of Da, and vice versa. We quantified the amount of Da present in the nucleus versus the cytoplasm in wild-type neurons and compared this ratio to neurons where Emc was knocked down or overexpressed. We did not observe any significant differences in the amount of Da in the nucleus versus the cytoplasm compared to controls when Emc was overexpressed (Figure 6a,c, and g). However, there was a significant increase in the amount of Da in the nucleus when Emc was knocked down ( Figure  6a,b, and g) suggesting that Emc may play a role in Da Responder corresponds to UAS-mediated expression. Gal4 Driver corresponds to Dcr2;;elav-Gal4 (Pan-Neural). Scale bar represents 10 μm. Error bars represent standard error. *p < .05 compared to wild-type control. NS is not significant compared to wild-type control localization to the cytoplasm. Additionally, we did detect a significant increase in the amount of Emc that is present in the nucleus when Da was knocked down compared to controls (Figure 6d,e, and h). This altered nuclear/cytoplasmic ratio was even more pronounced when Da was overexpressed in differentiated neurons (Figure 6f,h), consistent with our observation from the PLA analysis. Taken together, these data suggest that Da expression is capable of altering Emc subcellular localization in neurons and that Emc expression is capable of altering Da subcellular localization in neurons.

| Emc promotes neurexin expression in differentiated neurons
We had previously shown that Da mediates the repression of NMJ boutons and branches in motor neurons via the repression of the cell surface adhesion molecule neurexin (D'Rozario et al., 2016). Using yeast two-hybrid and transcriptional assay studies, Spratford and Kumar have shown that Emc is capable of sequestering Da away from DNA, thus preventing Da-mediated transcription (Spratford & Kumar, 2015b). Our work here suggests that this may be accomplished by Emc sequestering Da in the cytoplasm of neurons. If Emc is also capable of controlling neurexin expression in neurons via the repression of Da function, then we hypothesize that decreased Emc levels would decrease neurexin mRNA levels, and increased Emc would increase neurexin mRNA levels. We first wanted to determine whether Emc altered da mRNA levels in neurons. We knocked down and overexpressed Emc specifically in differentiated neurons using Dcr2; elav-Gal4. As a control, we also examined da transcript levels in flies expressing the UAS:da-RNAi construct. While we observed a significant decrease in da mRNA when we knocked down Da (Figure 7a), we did not observe any effect on da mRNA levels when we either knocked down or overexpressed Emc (Figure 7a), suggesting that Emc does not alter da mRNA levels in differentiated neurons. We next analyzed the level of neurexin mRNA. We observed a significant increase in neurexin mRNA upon Da knockdown (as we have previously reported D' Rozario et al., 2016). We also observed a significant decrease in neurexin mRNA levels when we knockdown Emc (Figure 7b), and a significant increase in neurexin mRNA levels when we overexpress Emc (Figure 7b). These data suggest that Emc is required to control the levels of neurexin mRNA in differentiated neurons.
Given that Emc knockdown decreases neurexin expression, we wanted to determine if we could effectively rescue the Emc knockdown NMJ and motor behavior phenotypes by overexpressing neurexin in an Emc knockdown background. Overexpression of neurexin produced a significant increase in both bouton number and branches compared to the wildtype control (Figure 7c and e-f), consistent with what has been previously shown (Li, Ashley, Budnik, & Bhat, 2007). Based on the available data, we hypothesized that overexpression of neurexin in animals where Emc is knocked down would rescue the axonal branching and bouton phenotypes normally observed in Emc knockdown neurons.
When we simultaneously knocked down Emc and overexpressed neurexin, we observed no significant difference in the number of branches between this genotype and the control (Figure 7d,f), and an intermediate phenotype in bouton number compared to neurexin overexpression alone or Emc knockdown alone (Figure 7d-e). We next determined whether neurexin overexpression could also rescue the behavioral phenotypes we observed with decreased Emc. We observed that individually both Emc knockdown and neurexin overexpression exhibited a significant decrease in motor ability compared to wild-type controls (Figure 7g). However, we observed a significant rescue of this defect when we simultaneously knocked down Emc and overexpressed neurexin (Figure 7g). Taken together, these data indicate that Emc also mediates its effect on branch and bouton number by working in opposition of Da to control the levels and function of neurexin in differentiated neurons.

| DISCUSSION
Class I bHLH proteins require homo-or heterodimerization in order to bind to DNA at E-box consensus sequences and regulate the transcription of target genes (Cabrera & Alonso, 1991;Cronmiller & Cummings, 1993;Massari & Murre, 2000;Smith and Cronmiller, 2001). Here, we show that the Class V HLH protein Emc promotes axonal arborization and bouton number in neurons by binding Da, and sequestering Da in the cytoplasm of differentiated neurons, thereby preventing Da from binding to and repressing the Da target gene neurexin. This work provides novel insights into the molecular functions of Class I bHLH and Class V HLH proteins in neurons by suggesting that subcellular localization of these proteins may be important in achieving these proteins' functions.
While we show that Emc sequesters Da away from the DNA by binding to Da predominantly in the cytoplasm, the exact mechanisms of this interaction are unclear. A potential mechanism for this cytoplasmic sequestration would be Emc heterodimerizing with Da in the nucleus and exporting Da out of the nucleus. This hypothesis is supported by our data showing that Emc localization increases in the nucleus of neurons when Da is overexpressed, which suggests that upon Da overexpression, more Emc is required in the nucleus for inhibition of Da. Additionally, knockdown of Emc increases F I G U R E 5 Emc interacts with Da mostly in the cytoplasm. (a-l) third instar VNC neurons labeled with PLA analysis using α-Da and α-Emc (a-d). (e-h) PLA analysis with no primary antibody. (i-l) PLA analysis in Da overexpression background. (b, f, j) are PLA channel alone from (a, e, i). (c, g, k) shows differentiated neurons marked with GFP expression from (a, e, i). Membrane-bound GFP labels postmitotic neurons. (d, h, l) shows DNA from (a, e, i). Scale is 10 µm. (m) Shows average relative fluorescence of PLA channel for each genotype indicated. (n) Shows average relative fluorescence of PLA channel in nuclear/total and cytoplasmic/ total fluorescence for each genotype listed. Error bars represent standard error. *p < .05. NS is not significant Da localization to the nucleus suggesting that Da is able to remain in the nucleus in order to affect target gene transcription. Further, it was recently shown that Da protein levels help regulate Emc protein levels through stabilizing Emc in heterodimers, adding an additional layer of complexity to this interaction (Li & Baker, 2018). Finally, Da has been shown to regulate Emc expression via posttranslational regulation and transcriptional regulation during adult peripheral neurogenesis (Li & Baker, 2019). Further work will be needed to fully understand the mechanisms underlying the interplay between Da and Emc subcellular localization we have identified in neurons.
An alternative hypothesis is that the sequestration of Da to the cytoplasm is the result of Emc heterodimerizing with Da in the cytoplasm and preventing Da from entering the nucleus. This hypothesis is supported by literature showing that the mammalian homolog of Da, Tcf4, is able to enter the nucleus via heterodimerization, even when the Tcf4 protein contains a mutated nuclear localization sequence (Sepp, Kannike, Eesmaa, Urb, & Timmusk, 2011). Thus, blocking the ability of Da to heterodimerize with partners other than Emc may be an important step required to maintain Da in the cytoplasm. Emc may also play a role in NMJ branch and bouton number through mechanisms that are independent of Da. For example, Emc has been shown to heterodimerize and negatively regulate Class II bHLH proteins during development (Benezra, Davis, Lockshon, Turner, & Weintraub, 1990). Additionally, the mammalian homologs of Emc (Id proteins) can promote neurite elongation through heterodimerizing with the bHLH transcription factor NeuroD and repressing its function (Iavarone & Lasorella, 2006). Upon Da knockdown in neurons, we observe an increase in Emc localization to the nucleus, which may suggest that since Da is now limited, Emc is freed up to interact with additional binding partners in neurons. Literature showing that Emc requires heterodimerization in order to remain stable also supports this hypothesis (Li & Baker, 2018). Furthermore, while we see an effect on axonal branching and bouton number when we overexpress Emc, we see no effect on Da localization upon Emc overexpression. This could further suggest that Emc affects axonal branching and bouton number through both Da-dependent and -independent mechanisms.
Our work identifies Emc as one binding partner of Da function in differentiated neurons. However, Da is known to have multiple binding partners in mitotic cells and will most likely have additional binding partners that are required for the restriction of axonal branching and bouton number in differentiated neurons as well. Additionally, the overall abundance of Da has a much larger effect on Emc localization, which may suggest that Emc is the limiting protein in the Da-Emc relationship and that Emc is just one protein dictating the subcellular localization of Da in neurons. Our previous work has shown that Da homodimers also mediate neurexin expression in neurons (D'Rozario et al., 2016). However, this does not eliminate the potential for other HLH dimerization partners to be required for normal Da function in neurons. Further, Da has also been shown to affect transcription of target genes through interacting with proteins other than those in than the bHLH family, such as the Notch intracellular domain and suppressor of hairless in the Notch transcriptional complex (Cave, Xia, & Caudy, 2009). Additionally, in order for Da homodimers to bind to DNA at E-box consensus sequences, additional proteins are required to provide specificity and bind at sites adjacent to E-box sequences (Murre et al., 1994;Tanaka-Matakatsu, Miller, Borger, Tang, & Du, 2014). Overall, further investigation is required in order to determine the exact proteins required for DNA binding by Da in neurons.
Emc is not the only protein identified that binds to Da and negatively regulates its function. Since Da is heavily regulated posttranslationally, other proteins in differentiated neurons could be regulating Da function. Transcription factors such as Nervy, Hairy, Stripe, and Enhancer-of-split proteins have all been shown to bind to Da and repress Da activity (Alifragis, Poortinga, Parkhurst, & Delidakis, 1997a;Fischer & Gessler, 2007;Giagtzoglou, Koumbanakis, Fullard, Zarifi, & Delidakis, 2005;Usui, Goldstone, Gibert, & Simpson, 2008;Wildonger & Mann, 2005;Zhang, Kalkum, Yamamura, Chait, & Roeder, 2004). Though we have not investigated other potential binding partners here, these proteins (as well as others) may also be involved in controlling Da function in differentiated neurons.
We have shown that Emc controls the levels of neurexin mRNA in neurons. However, it is likely that there are more genes whose expression Emc is controlling that regulate bouton and branch number in neurons. Emc could therefore be affecting the expression of additional target genes by regulating Da and/or other neuronal-specific bHLH transcription factors. Our previous work identified potential Da target genes in neurons through the analysis of two large-scale genome-wide data sets involving Da and TCF4 (D'Rozario et al., 2016). A list of 44 candidate genes specifically expressed in the Drosophila NMJ was generated from this analysis, with neurexin being one of the potential targets (D'Rozario et al., 2016). These other 43 genes would make attractive candidates to explore futher to help elucidate the molecular mechanisms underlying how Da and Emc affect axonal arborization and bouton number in neurons.
Haploinsufficiency of TCF4 has been associated with Pitt-Hopkins syndrome (Amiel et al., 2007;Tamberg, Sepp, Timmusk, & Palgi, 2015), while common small nucleotide polymorphisms in TCF4 have been associated with schizophrenia (The Schizophrenia Psychiatric Genome-Wide Association Study (GWAS) Consortium, 2011). Defects in human ID protein function are associated with a number of disorders in other cell types including eye disease (Fan et al., 2018;Guo et al., 2015), hydronephrosis (Aoki et al., 2004), and bone disease (Fiori, Billings, de la Pena, Kaplan, & Shore, 2006). Further investigation into the mechanisms underlying Da function in neurons and how Emc affects Da function could have a profound impact on our understanding of the etiology of these diseases and may provide insights into additional targets for therapeutic avenues.
In summary, we propose that Emc functions in neurons to promote axonal arborization and bouton formation through binding to and inhibiting the function of Da. The interplay between Emc and Da is important for stable and appropriate innervation by neurons via the regulation of neurexin expression. Alterations in the normal balance between Emc and Da in neurons leads to defects in neural circuit formation and behavior, as well as altered Emc/Da protein subcellular location. This research identifies the first binding partner of Da in differentiated neurons and opens the door for further research into elucidating the molecular mechanisms underlying Damediated restriction of axonal branching and bouton number.
Imaging of NMJs and larval brains was performed using an Olympus FluoView FV1000 laser scanning confocal microscope. Image analysis and quantification was performed using ImageJ software. The number of axonal branches, the type of branching, and the number of synaptic boutons, were quantified as previously described (D'Rozario et al., 2016;Mhatre et al., 2014). Tissue fluorescence intensity was calculated as previously described (D'Rozario et al., 2016). GFP intensity was used to normalize tissue fluorescence in all samples.

| Larval behavioral assay
Wandering third instar larvae from both sexes were briefly rinsed with PBS and allowed to acclimate for 1 min on a 4% agar 100-mm Petri dish. Individual larvae were then transferred to a clean 4% agar plate. All experiments were performed on a clean flat agar surface under constant white light as previously described (D'Rozario et al., 2016). Body wall contractions were counted for 30 s for each larva using a Leica Mz 125 stereomicroscope.

| Proximity ligation assay
Duolink PLA kit was obtained from Sigma-Aldrich. Larval brains were made accessible via larval filet preparations. Larvae were dissected in PBS and then fixed in 4% paraformaldehyde for 25 min. Tissues were washed in 0.1% PBT for 15 min, 0.5% PBT for 15 min, and then blocked in Duolink Blocking Solution for 30 min at 37°C. Tissue samples was then incubated in primary antibodies overnight at 4°C (mouse anti-daughterless [1:40]] and rabbit anti-extramacrochaetae [1:1,000]). After incubation, samples were washed in Duolink 1× Wash Buffer A twice for 5 min at 25°C. Samples were then incubated in Duolink PLA secondary probes in a 1:5 dilution for 1 hr at 37°C. Samples were then washed in Duolink 1× Wash Buffer A twice for 5 min at 25°C. After washing, tissue samples were incubated in Duolink Ligation stock (1:5) containing ligase (1:40) for 30 min at 37°C. Samples were then washed in Duolink 1× Wash A Buffer twice for 2 min at 25°C. Tissue samples were then incubated in Duolink Amplification stock (1:5) with polymerase (1:80) for 100 min at 37°C. Samples were then washed in Duolink 1× Wash Buffer B twice for 10 min and then in Duolink 0.01× Wash Buffer B for 1 min. Tissue samples were mounted in Duolink Mounting Medium. Images were obtained using an Olympus FluoView FV1000 laser scanning confocal microscope. Image analysis and quantification was performed using Adobe Photoshop software. Mean tissue fluorescence intensity was measured, and background fluorescence was subtracted to calculate a mean corrected fluorescence intensity for each sample. GFP intensity was used to normalize tissue fluorescence in all samples.

| Western blot
Wandering third instar larval brains were collected to prepare lysates. Lysates were prepared from 35 larval brains that were immediately homogenized in RIPA buffer (50 mM Tris, 150 mM NaCl, 1% SDS, 1% NP-40, and 0.5% deoxycholate, pH 8.0) supplemented with protease inhibitor cocktail (EMD Millipore) and 1 mM EDTA. The homogenate was centrifuged at 13,000 rpm at 4°C for 30 min to collect supernatant for analysis. Samples were stored at −80°C. Samples were prepared using the 4× NuPage LDS sample buffer (Invitrogen, Inc.) containing 0.2% BME (β-mercaptoethanol, Sigma-Aldrich) and heated to 95°C for 10 min. Equal lysate volumes were loaded into each well of NuPAGE 4%-12% Bis Tris Gel (Invitrogen). Gels were run using MES running buffer and transferred to PVDF membrane (Immobilon Millipore) using a semi-dry transfer apparatus (Owl Scientific) and NuPage transfer buffer (Invitrogen). Membranes were then blocked with 2% BSA PBS blocking buffer for 1 hr. The membrane was probed overnight at 4°C with blocking buffer containing rabbit anti-extramacrochaetae antibody and an antibody to β-actin (Sigma-Aldrich). The membrane was then washed 4 times with 1× PBST for 5 min each. The membrane was probed for 1 hr at room temperature with goat anti-mouse secondary antibody (IRDye 680LT; LiCor) and goat anti-rabbit secondary antibody (IRDye 800CW; LiCor). Band intensities were quantified using the ImageStudiolite software.

| Quantitative reverse transcriptionpolymerase chain reaction (RT-qPCR)
Fifty wandering third instar larval brains per genotype were dissected in ice-cold phosphate buffer and immediately transferred to RNA Later (Abion) and stored at −80°C. RNA isolation was performed using phenol:chloroform extraction followed by alcohol precipitation. RNA was stored in DNAse/RNase-free water at −80°C. An adapted version of iTaq Universal SYBR Green One-Step protocol (Bio-Rad) was used (Latcheva, Viveiros, & Marenda, 2019) and samples were run on Bio-Rad C1000 Thermal Cycler CFX96 Real-Time system. Primers for Da, neurexin, Emc, and RP49 mRNA were made using IDT. Cycle threshold (Ct) values were obtained graphically for the targets and housekeeping control. ∆Ct values were calculated by subtracting the Ct value for each primer set from the Ct value of the housekeeping control. ∆∆Ct values were calculated by subtracting the ∆Ct value of the control samples from those of the experimental samples. Fold enrichment in expression was as previously described (D'Rozario et al., 2016). Each experiment was performed in triplicate with a minimum of three biological replicates.

| Statistical analysis
All statistical analyses were performed using Microsoft Excel and SPSS Statistical Software. Significance was determined at the 95% confidence interval. Unpaired Student's t test was used for Figures 1, 5, 6, and 7. One-way ANOVA test with Tukey's post hoc or Games-Howell post hoc analysis (depending on the variance) was used for Figures 2, 3, 4, and 7. Genotype was the independent variable. Statistical significance in figures is represented by *= p < .05. Error bars represent the standard error of the mean (SEM).