dTACC restricts bouton addition and regulates microtubule organization at the Drosophila neuromuscular junction

Abstract Regulation of the synaptic cytoskeleton is essential to proper neuronal development and wiring. Perturbations in neuronal microtubules (MTs) are associated with numerous pathologies, yet it remains unclear how changes in MTs may be coupled to synapse morphogenesis. Studies have identified many MT regulators that promote synapse growth. However, less is known about the factors that restrict growth, despite the potential links of synaptic overgrowth to severe neurological conditions. Here, we report that dTACC, which is implicated in MT assembly and stability, prevents synapse overgrowth at the Drosophila neuromuscular junction by restricting addition of new boutons throughout larval development. dTACC localizes to the axonal MT lattice and is required to maintain tubulin levels and the integrity of higher‐order MT structures in motor axon terminals. While previous reports have demonstrated the roles of MT‐stabilizing proteins in promoting synapse growth, our findings suggest that in certain contexts, MT stabilization may correlate with restricted growth.

These findings suggest a model where increased stabilization of MTs is associated with increased NMJ expansion. Additional factors that are associated with MT stability, such as the formins Diaphanous (Pawson, Eaton, & Davis, 2008) and DAAM (Bartolini & Gundersen, 2010;Migh et al., 2018), have also been found to promote NMJ expansion.
As a counterbalance to MT-stabilizers, MT destabilizers/severing proteins, such as Spastin (Sherwood, Sun, Xue, Zhang, & Zinn, 2004) and Katanin (Mao et al., 2014) restrict NMJ size. Consistently, repression of Futsch mRNA levels by Dfxr (homolog of FMR1) prevents NMJ overgrowth (Zhang et al., 2001), further supporting the notion that MT stability correlates with NMJ expansion. Interestingly, mutation of human spastin is the most frequent cause of hereditary spastic paraplegias (HSP; Solowska & Baas, 2015), while mammalian katanin has been associated with behavioral deficits and intellectual disability (Banks et al., 2018;Bartholdi et al., 2014). Similarly, dfxr is associated with Fragile X syndrome, one of the most common forms of inherited intellectual disability (Penagarikano, Mulle, & Warren, 2007). The synaptic phenotypes and disease relevance of genes such as spastin, katanin, and dfxr has led to the understanding that excessive synaptic growth is highly detrimental. Collectively, these findings suggest that a complex set of factors is responsible for maintaining a precise balance of both synapse expansion and restriction to ensure neurological function and health.
Given (a) prior studies suggesting that increased MT stability correlates with growth and (b) the established roles of TACC proteins in MT assembly and stability, dTACC would naturally be expected to promote synapse growth. Surprisingly, we discovered instead that presynaptic dTACC negatively regulates the growth of the larval NMJ by limiting addition of synaptic boutons during development. We also found that within the motor axon terminal, dTACC associates abundantly along the lattice of MTs and regulates both the integrity and higher-order organization of MTs. Our results suggest that in certain contexts, assembly and/or organization of MTs by proteins such as dTACC may restrict NMJ expansion.

| Presynaptic dTACC is required to restrict NMJ size
The NMJ has a highly stereotyped morphology consisting of a branched motor axon terminal that is decorated by numerous presynaptic boutons (Figure 1a). In this system, boutons are often quantified as a measure of overall NMJ size. To determine the effect of dTACC on synapse morphogenesis, we initially counted mature type I boutons at the muscle 6/7 NMJ in late-stage third instar dtacc mutant larvae labeled with the neuronal membrane marker antihorseradish peroxidase (HRP; Jan & Jan, 1982). To generate a strong dtacc loss background, we raised transheterozygotes using two independently derived alleles: dtacc 592 (d-tacc stella ), a complete null (Lee et al., 2001), and dtacc 1 (d-TACC 1 ), which has been previously described as a very strong allele (Gergely, Kidd, et al., 2000). We combined dtacc 592 and dtacc 1 over Df(3R)110, a deletion at the locus ( Figure S1A) and confirmed that these animals showed phenotypes comparable to dtacc 592 /dtacc 1 transheterozygotes ( Figure 1). However, dtacc 592 /Df(3R)110 and dtacc 1 /Df(3R)110 animals were very weak, suggesting that haploinsufficiencies uncovered by the deletion contributed to pleiotropy. Thus, we focused our remaining analysis on dtacc 592 /dtacc 1 transheterozygotes to avoid additional phenotypes resulting from deletion of flanking genes.
In the dtacc 592 /dtacc 1 background, we also quantified branching within the motor axon terminal, overall length of the NMJ, and bouton size; both branch number and NMJ length confirmed highly significant increases of NMJ size in dtacc 592 /dtacc 1 compared to control ( Figure 1k,l). We found that dtacc 592 /dtacc 1 animals display a~33% decrease in average bouton area ( Figure 1m). This bouton size phenotype, along with the dtacc overgrowth phenotype, raised the question of which step(s) of bouton formation may require dTACC. During normal NMJ expansion (Zito et al., 1999), baseline bouton addition is controlled by developmental signaling cues (Figure 2a,i), which coordinate and balance neuronal expansion with muscle growth (reviewed by Van Vactor & Sigrist, 2017). Bouton addition can also be induced acutely by activity-dependent cues in response to stimuli (Figure 2a,ii;Budnik et al., 1990;Chklovskii et al., 2004). Immediately following baseline or activity-induced addition, nascent boutons lack pre-and postsynaptic markers and thus have a "ghost"-like appearance We investigated the potential involvement of dTACC in each of these steps. We considered the possibility that dTACC promotes bouton pruning (Figure 2a,v) but did not find changes in Dlg footprints F I G U R E 2 dTACC regulates bouton addition during development. (a) Schematic showing the steps of bouton formation, including addition of boutons in response to (i) baseline or (ii) activity-dependent growth, followed by (iii) bouton maturation via recruitment of pre-and postsynaptic components. Boutons may continue to (iv) grow to full size and stabilize or (v) retract instead, leaving "footprints" of postsynaptic material. (b and c) NMJ stained α-HRP and counterstained with α-Brp (Nc82; b) or α-Dlg (4F3; c). Nascent boutons (triangle) can be identified by the lack of postsynaptic markers such as Dlg. (d and e) dtacc 592 /dtacc 1 flies showed a twofold increase in Brp-negative boutons (d) and in Dlg-negative boutons (e) compared to w 1118 controls. (f) Spaced-stimulation paradigm used to induce rapid activity-dependent bouton budding. (g) controls and dtacc 592 /dtacc 1 showed nearly identical fold-changes (2.36-and 2.31-fold, respectively) in the number of nascent boutons following activity compared to mock-treated controls. *p < .05, t-test; error bars indicate ± SEM; number of NMJs quantified indicated on graph; scale bar, 1 μm in dtacc animals, making this explanation unlikely (see Section 3). Furthermore, we found no striking defects in the overall distributions of synaptic cytomatrix antigens in dtacc animals, such as the core active zone component Bruchpilot (Brp; Figure S2A,B) or Dlg ( Figure S2C,D), suggesting that there were no catastrophic effects on initial bouton maturation (Figure 2a,iii). However, the~33% decrease in average bouton area (Figure 1e) did suggest a defect at the stage where maturing boutons grow to full size (Figure 2a,iv). Reduction in bouton growth could reflect some compensation for the effects of overgrowth. Alternatively, this could arise because synthesis and/or transport may not increase at the same rate as bouton number, thus making materials too sparse to form normal-sized boutons. In either scenario, the defects in bouton growth and number raised the possibility that there might be upstream defects in bouton initiation.

| dTACC regulates bouton initiation in response to baseline developmental cues
To investigate the potential role of dTACC in bouton addition, we looked for changes in the incidence of nascent boutons, which can be identified by the lack of maturation markers such as Brp (Figure 2b Moreover, the twofold increase in putative nascent boutons in dtacc animals ( Figure 2) is equal to the twofold increase we previously observed in mature bouton number (Figure 1). This doubling of both nascent and total bouton numbers further supports the model that increased bouton number in dtacc mutants is due to increased baseline bouton addition.

| dTACC colocalizes with the lattice of synaptic MTs
To better understand how dTACC might affect NMJ growth, we asked if dTACC associates with MTs at the NMJ. While TACC has been most frequently reported to localize to MT minus-or plus-ends (Ding et al., 2017;Hood and Royle, 2011;Peset and Vernos, 2008;Thakur et al., 2013), dTACC puncta have been observed along the lattice of spindle MTs (Gergely, Kidd, et al., 2000). Furthermore, Xenopus TACC3, the TACC isoform most highly expressed in the Xenopus embryonic nervous system (Rutherford et al., 2016;Tessmar, Loosli, & Wittbrodt, 2002), abundantly decorates the length of MTs in egg extracts (Peset et al., 2005), and the Schizosaccharomyces pombe TACC homolog Alp7/Mia1p is found along MTs both in vivo (Sato et al., 2004) and in vitro (Thadani et al., 2009).
We investigated the precise nature of dTACC protein localization at the NMJ to distinguish between potential modes of MT interaction in vivo. We generated and validated a novel monoclonal antibody against dTACC (see Section 3). Quantification of dTACC intensity and Western blotting confirmed loss of signal in dtacc mutants ( Figure S3A,B). Using this antibody, we found that the majority (~90%) of dTACC signal in controls was found in polymer lattice-like structures within the axon terminal ( Figure 3A; solid triangles) which strongly resemble the MT lattice at the core of the axon, consistent with prior reports that Caenorhabditis elegans TAC-1 localizes to the axons of sensory neurons (Chen et al., 2015). A smaller fraction of dTACC intensity (~10%) was found in punctate bouton-associated structures (Figure 3a; hollow triangles). When we used our antibody on dtacc 592 /dtacc 1 animals, dTACC signal was dramatically reduced with the minor punctate fraction virtually abolished and the major lattice-like staining markedly decreased (Figure 3b). Although dTACC was not detectable on a Western blot with our antibody ( Figure S3A), there was small residual immunohistochemical signal in the dtacc 592 / dtacc 1 mutant ( Figure S3B) that may reflect residual expression (as much as 5% in dtacc 592 /dtacc 1 ) predicted by previous characterization of the dtacc 1 background (Gergely, Kidd, et al., 2000). Motivated by our finding that dTACC is required throughout early NMJ development ( Figure S1B-D), we also examined dTACC distribution in firstinstar larvae ( Figure S3C) and found, as expected, that dTACC is highly expressed in the ventral nerve cord (VNC) of the CNS ( Figure S3D) and throughout motor/sensory axon tracts ( Figure S3E). Importantly, we tested the colocalization of dTACC with MTs by staining for alpha-tubulin (Ab15246 for all tubulin immunohistochemistry) and found that the distribution of lattice-like TACC and alpha-tubulin was highly coincident (Figure 3c). The Manders' overlap coefficient M1 (percent dTACC colocalizing with tubulin) was found to be 56.3% ± 3.7, while M2 (percent tubulin colocalizing with dTACC) was 58.6% ± 6.8 (n = 15 NMJs; Manders, Verbeek, & Aten, 1993). We also considered the possibility that the punctate fraction of dTACC might be associated with MT plus-ends within the motor terminal. To test this idea, we used 3-dimensional structured illumination microscopy (3D-SIM) to measure the average distance between dTACC and MTs based on a published methodology (Lepicard et al., 2014). Compared to confocal microscopy, 3D-SIM improves resolution by twofold in all three dimensions and can thus resolve objects with up to eightfold smaller volume (Gustafsson, 2000 (Figure 3f). Although this method may not be sensitive enough to detect single MTs or MTs that are highly dynamic and/or F I G U R E 3 TACC colocalizes with tubulin at the synaptic terminal. (a and b) Validation of dTACC antibody in NMJs stained with α-dTACC and α-HRP. Compared to control w 1118 flies (a), dtacc 592 /dtacc 1 flies showed reduced dTACC staining (b). In controls (a), the majority of TACC showed a filamentous distribution highly reminiscent of the MT lattice within motor axon terminals (solid triangles), while a smaller TACC population formed puncta (hollow triangles; see high magnification inset). In dtacc 592 /dtacc 1 animals (b), punctate TACC was virtually absent, and filamentous TACC was dramatically reduced. (c) Colocalization of α-dTACC and α-alpha-tubulin (Ab15246) in the axon terminal. The distribution of filamentous TACC was similar to the distribution of tubulin, while punctate TACC appeared spatially distinct from tubulin. The Manders' overlap coefficients were M1 = 56.3% ± 3.7 (percent TACC colocalizing with tubulin) and M2 = 58.6% ± 6.8 (percent tubulin colocalizing with TACC; n = 15 NMJs). Scale bar, 5 μm for main panels, 1 μm for insets. (d-f) Spatial relationships between punctate dTACC and MTs were examined through 3D-SIM of samples stained with α-dTACC and α-alpha-tubulin (d). Double white lines (d) represent positions from which intensity profile plots were drawn. (e) Representative intensity profile plot showing separation between the peaks of α-dTACC and α-tubulin staining. (f) Quantification of mean distance between peaks of α-dTACC and α-tubulin staining measured from intensity profile plots (n = 71 puncta). Mean dTACC-MT distance was found to be 98.6 nm. Scale bar, 500 nm labile, we were unable conclude that the distal puncta of dTACC in boutons are closely associated with MTs.
Overall, our results suggested that in the NMJ arbor, the majority of dTACC is spatially localized with the lattice of MTs, similar to prior observations in Xenopus, Drosophila and fission yeast (Gergely, Kidd, et al., 2000;Peset et al., 2005;Sato et al., 2004;Thadani et al., 2009).
Interestingly, purified yeast Alp7 localizes to regions of overlap between adjacent (parallel or anti-parallel) MTs where it is thought to mediate cross-linking of bundled MTs, thereby promoting the assembly and stability of linear MT arrays both in vitro and in vivo (Thadani et al., 2009). A role in cross-linking kinetochore MTs is also observed for TACC3 in HEK293 cells (Booth, Hood, Prior, & Royle, 2011). Given that neuronal MTs are organized into polarized bundles that resemble the MT arrays found in S. pombe (Baas, Rao, Matamoros, & Leo, 2016;Bartolini & Gundersen, 2006;Hoogenraad & Bradke, 2009), it seemes possible that dTACC serves a similar function in regulating synaptic MT organization. Interestingly, this possibility would be consistent with findings that Pavarotti, a kinesin that cross-links MTs and stabilizes the mitotic spindle, is also a negative regulator of NMJ size (McLaughlin, Nechipurenko, Liu, & Broihier, 2016).

| TACC is required for normal levels and organization of synaptic MTs
The MT-lattice localization of dTACC at the NMJ is consistent with previous studies of TACC function in MT organization and stability. In Drosophila embryos, loss of dTACC results in short astral and spindle MTs (Gergely, Kidd, et al., 2000), and similar roles in regulating both mitotic and interphase MTs have been observed for TACC proteins in a variety of systems (Ding et al., 2017;Hood and Royle, 2011;Peset and Vernos, 2008;Thakur et al., 2013). Thus, the close association of dTACC with the MT lattice suggested a specific role in regulating MTs within motor axon terminals.
F I G U R E 4 TACC is regulates the architecture and higher-order organization of synaptic MTs. (a-c) Comparison of tubulin intensity in control and dtacc 592 /dtacc 1 NMJs stained with α-tubulin (Ab15246) and α-HRP. Control animals (a) showed robust tubulin staining with clear filamentous structures (triangles). Tubulin staining was most concentrated in the main axonal shaft (solid triangles), while thinner tubulin filaments were observed in terminal branches (hollow triangles). In dtacc 592 /dtacc 1 flies (b), tubulin staining appeared weaker and was often undetected in branches. Quantification (c) indicates that detectable tubulin intensity was significantly reduced dtacc 592 /dtacc 1 flies. Scale bar, 10 μm. (d-g) Analysis of Futsch-labeled MT arrangements in control and dtacc 592 /dtacc 1 NMJs labeled with α-Futsch (22C10) and α-HRP. Futsch-decorated MTs in the three terminal boutons of each branch were categorized into continuous, looped, and diffuse patterns (d-f). Side-by-side comparison of boutons showed that Futsch-MT structures are clearly reproducible between controls and dtacc 592 / dtacc 1 mutants. Quantification (g) shows that in dtacc 592 /dtacc 1 , the distribution of different MT structures is altered, with the frequency of continuous and looped structures very significantly decreased and diffuse staining very significantly increased. *p < .05, ***p < .001, t-test; error bars indicate ± SEM; number of NMJs quantified indicated on graph; scale bar, 1 μm To investigate if dTACC regulates synaptic MTs, we compared the staining intensity and distribution of tubulin in dtacc 592 /dtacc 1 animals to w 1118 controls (Figure 4a,b). Due to the fragility of MTs, we used a specifically optimized fixation protocol (see Section 3). Control tubulin staining was clear and robust, with distinct filamentous structures ( Figure 4a). There was a clear concentration of tubulin in the main axon shaft, with thinner filaments leading out into the branches of the synaptic terminal where bouton addition is more frequent. In contrast to previous demonstrations (Jin et al., 2009;Mao et al., 2014;Sherwood et al., 2004;Trotta, Orso, Rossetto, Daga, & Broadie, 2004) of robust muscle MT staining, we observed comparatively weaker postsynaptic MT signal. This likely reflects differences in our protocol, which was specifically optimized to target the labile, unstable presynaptic MT population.
Compared to controls, dtacc animals displayed diminished or undetectable tubulin staining in terminal branches, where the majority of bouton addition events occur. While tubulin was still detected in the main axon shaft, this population was also visibly reduced ( Figure 4b). Consistently, quantification indicated a~32% overall decrease in tubulin intensity in dtacc animals (Figure 4c). This reduction in staining could reflect a decrease in total tubulin mass, however, no obvious decrease in alpha-tubulin (Ab7291) was detected on our Western blots of dtacc mutant larvae ( Figure S3A). This suggested that total MT polymer must be reduced in dtacc mutants.
Alternatively, since single MTs may be difficult to resolve with in vivo light microscopy compared to bundled MT arrays, a change in the spatial organization of MTs could also reduce detectable tubulin staining. This encouraged us to seek an additional histological probe for MTs and their in situ organization.
To further understand correlation between the NMJ size and MT phenotypes, we next tested the effect of dTACC on the formation of higher-order MT structures that have been associated with different states of bouton growth and division (Conde & Cáceres, 2009;Roos et al., 2000;Ruiz-Cañada & Budnik, 2006 (Conde & Cáceres, 2009;Roos et al., 2000;Ruiz-Cañada & Budnik, 2006). Given that Futsch binds a subpopulation of stabilized MTs (Roos et al., 2000) and is not known to bind individual tubulin dimers, the diffuse appearance of dispersed/splayed MTs likely reflects structures such as short MT fragments, or longer individual MTs that are splayed from the main bundle, as opposed to free tubulin.
We investigated the effects of dtacc loss on Futsch-labeled MTs structures previously categorized as "continuous," "looped," or "diffuse" (Jin et al., 2009;Sherwood et al., 2004). These structures are clearly and reproducibly distinguished in both w 1118 controls and in dtacc 592 /dtacc 1 animals (Figure 4d-f). We focused on the three most terminal boutons of each branch, as branch ends are the sites of most active growth (Zito et al., 1999). Loss of dtacc produced consistent, measurable changes to Futsch-MT structures: compared to controls, mutants showed very significantly decreased frequency of both continuous (40% vs. 24%) and loop structures (23% vs. 11%; Figure 4g).
In contrast, the majority of dtacc boutons (65%) showed diffuse staining, in contrast to controls (37%; Figure 4g). Our findings thus suggest that dTACC promotes the organization of MTs into stable continuous and looped structures in wild-type, whereas dtacc loss nearly doubles the number of boutons containing dispersed/splayed MT structures.
Collectively, our findings suggest that dTACC serves to restrict NMJ overgrowth and regulate MT organization and/or assembly.
These studies support the notion that MT regulator is vital to controlling synapse expansion, as implied by studies of several other MTassociated factors. Interestingly, the frequency of diffuse Futsch staining is increased by 1.8-fold in dtacc mutants (Figure 4g). This is reminiscent of the twofold increase in the number of both mature to bundled and looped distributions in paused growth cones (Dent, Callaway, Szebenyi, Baas, & Kalil, 1999;Kalil, Szebenyi, & Dent, 2000;Tanaka, Ho, & Kirschner, 1995;Tanaka & Kirschner, 1991). Given the resemblance of the MT organizations we observe at the NMJ  (Jin et al., 2009). Intriguingly, the overgrowth phenotype of spastin mutants is also accompanied by a reduction in both tubulin and Futsch staining, despite the function of Spastin as a MT-destabilizer (Sherwood et al., 2004). It has been proposed that the MT severing activity of Spastin generates seeds that nucleate the growth of new MTs (Roll-Mecak & Vale, 2006), thus explaining the attenuated MT network of spastin mutants (Sherwood et al., 2004). Collectively, our findings in dtacc mutants, as well as the previous studies of tbce and spastin, suggest that NMJ growth may not be correlated solely with MT stabilization and levels, but may also be related to the organization of MTs. This possibility is consistent with prior studies that have shown a correlation of displayed/splayed MT structures with actively growing or dividing boutons (Conde & Cáceres, 2009;Roos et al., 2000;Ruiz-Cañada & Budnik, 2006).
In conclusion, we demonstrate that dTACC is a negative regulator of bouton addition during the development of the NMJ and that dTACC associates with and regulates the stability and organization of synaptic MTs. We provide evidence that dTACC promotes MT structures associated with paused bouton growth and division. Further studies may investigate the functional partners of dTACC at the NMJ, and how the roles of dTACC may relate to the roles of factors such as TBCE and Spastin, which show similar overgrowth and MT organization phenotypes.

| Antibody production and purification
dTACC sequence containing amino acids 146-327 was His-tagged, bacterially expressed, and purified. dTACC antibody was raised in mice against and purified by PrimmBiotech, Inc. (Cambridge, MA).

| Western blotting
See Supporting Information.

| Image acquisition and analysis
Synaptic arbors of muscle 6/7 in the abdominal segment A2 were used for all analyses. Imaging was performed on a Nikon A1R point scanning confocal and a Nikon Yokogawa spinning disc confocal with a Hamamatsu ORCA-R2 cooled CCD camera. 3D-SIM was performed on a DeltaVision OMX Blaze microscope (GE Healthcare Life Sciences, Marlborough, MA, USA) with a PCO sCMOS camera. Lasers were adjusted to prevent oversaturation. Images were processed and analyzed with ImageJ and/or MATLAB. Bouton number and size were counted and traced by hand. An HRP mask was used to restrict analysis to neuronal signal for intensity analysis, and MATLAB scripts were used to quantify dTACC and tubulin signals relative to HRP. Line scans were used to create intensity profiles to distinguish different Futsch structures.

| Statistics
All comparisons were done using Welch's t-test for unequal variances using Graphpad.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.