The following Drosophila melanogaster strains were used: Oregon-R (Bloomington Drosophila Stock Center, BDSC) w¹¹¹⁸ (used as a wild-type control, BDSC), UAS-dcr2; D42-GAL4 (BDSC), 24B-GAL4 (BDSC), OK6-GAL4 (BDSC), BG57-GAL4 (Budnik et al., 1996), Tao¹⁶/FM7, dfd-YFP (Pflanz et al., 2015), independent lines of UAS-flag-Tao (lines 8, 61, and 89) and UAS-flag-Tao^Kinase-Dead (Boggiano et al., 2011), UAS-Hpo (Wu et al., 2003), UAS-Myc-wts/CyO, dfd-YFP (Jia et al., 2003), UAS-flag-Yki (Xu et al., 2018), UAS-Tao RNAi-1 (VDRC lines 107645 and 17432 combined in one stock), UAS-Tao RNAi-2 (Transgenic RNAi Project (TRiP) line HMS01226),UAS-Tao RNAi-3 (Vienna Drosophila Stock Center (VDRC) line 17432), UAS-Tao RNAi-4 (VDRC line 107645), UAS-Tao RNAi-5 (TRiP line GL00015), UAS-Hpo RNAi (VDRC line 104169), UAS-wts RNAi (VDRC line 106174), UAS yki RNAi (VDRC line 104523), gbb¹/CyO, act-GFP (Wharton et al., 1999), wit^A12/TM6B (Marqués et al., 2002). Adult male D. melanogaster flies of various genotypes were crossed with female virgin UAS-dcr2; D42-GAL4 or OK6-GAL4 (neuronal expression), or 24B-GAL4 or BG57-GAL4 (muscle expression) flies to drive either overexpression or loss of gene function by RNAi. With the exception of w¹¹¹⁸, gbb¹/CyO, act-GFP, and wit^A12/TM6B, all stocks crossed to the GAL4 lines contained an UAS element. After 2 days, the crosses were transferred onto fresh food daily. For Tao¹⁶ experiments, Tao¹⁶/FM7, dfd-YFP virgin females were mated to w¹¹¹⁸ males in embryo collection cages using yeasted grape juice plates which were changed daily. YFP-negative first instar larvae were transferred to fresh yeasted plates, and were sex selected 4 days later as wandering third instar larvae to distinguish Tao¹⁶ hemizygous males from Tao¹⁶/w¹¹¹⁸ heterozygous female control siblings. For the Tao¹⁶ rescue experiments, crosses were performed in embryo collection cages as described above. Tao¹⁶/FM7, actin-GFP; UAS-Tao/TM3, Ser, actin-GFP virgin females were mated to either D42-GAL4 or BG57-GAL4 males to drive neuronal or muscle Tao overexpression, respectively. GFP-negative first instar larvae were transferred to fresh yeasted plates, and were sex selected 4 days later as wandering third instar larvae to select Tao¹⁶ hemizygous males. All crosses were performed at 25°C.

Larvae were grown at 25°C, and fillet or CNS dissections were prepared from wandering third-instar larvae. Briefly, the fillet dissections were performed in 1X PBS, by pinning the head and tail, and then cutting longitudinally along the dorsal midline and across the short-axis at A1 and A7. Innards were removed and cuticle was pinned down leaving only the body-wall musculature, which was then fixed in Bouin’s fixative (Polysciences) for 10 minutes before being rinsed in 1X PBS. The fillet preparations were then transferred to PT (1X PBS + 0.1% Tx-100) in a microcentrifuge tube and were rinsed and blocked in PTN (1X PBS + 0.1% Tx-100 + 1% normal goat serum). Preparations that were later used for bouton quantification were then incubated in rabbit or goat anti-Horseradish Peroxidase (HRP) conjugated to a Cy3 or Alexa 488 fluorophore (Jackson Immunoresearch) at a 1:500 dilution and in either mouse anti-Discs large (Dlg) (4F3, DSHB) at a 1:500 dilution or mouse anti-Bruchpilot (Brp) (nc82, DSHB) at a 1:250 dilution in PTN overnight at 4°C; a donkey anti-mouse secondary antibody conjugated to Alexa 488 or Alexa 594 (Jackson Immunoresearch) was subsequently used at a 1:1000 dilution for 2–4 hrs at RT or overnight at 4°C to visualize Dlg or Brp. Other antibodies used include rabbit anti-Tao (Pflanz et al., 2015) at a 1:1000 dilution. Preparations were then rinsed in PTN and 1X PBS, prior to mounting in ProLong anti-fade mounting media (Invitrogen). CNS preparations followed a similar immunofluorescence staining protocol after dissection but were dissected in Schneider’s media supplemented with 10% FBS (Sigma), and fixed in 4% paraformaldehyde (Polysciences) for 20 minutes before being transferred to borosilicate glass tubes for staining steps. Dissected CNSs were rinsed in 1X PBS before blocking for 1 hr in PTN at room temperature, as well as between primary and secondary antibodies, and after secondary antibody incubation. Antibodies used for CNS immunofluorescence included rabbit anti-pSMAD (1,5) (Cell Signaling 41D10) at a 1:100 dilution, rat anti-Elav (Elav-7E8A10, DSHB) at a 1:50 dilution, rabbit anti-Tao (Pflanz et al., 2015) at a 1:1000 dilution, and mouse anti-Repo (Repo-8D12, DSHB) at a 1:10 dilution in PTN overnight at 4°C. Donkey secondary antibodies conjugated to Alexa 488, Alexa 594, or Alexa 647 (Jackson Immunoresearch) were used at a 1:1000 dilution in PTN to visualize primary antibodies. CNSs were mounted in Invitrogen ProLong Gold anti-fade mounting media with DAPI (Invitrogen).

Fillet dissection preparations and CNS preparations were imaged on a Zeiss LSM 800 laser scanning confocal microscope using a 40× or 100× objective.

Boutons or Brp-positive puncta from muscle 4 of segments A2–4 of the fillet dissection preparations were counted using an Olympus BX60 epifluorescence microscope or a Zeiss Axioplan2 epifluorescence microscope using a 63× or 100× objective. For boutons from NMJ 6/7, only segment A2 was scored. Muscle surface area was determined by imaging muscles from segments A2–4 of fillet dissections stained with Dlg using a 10× objective with a Zeiss Axioplan2 epifluorescence microscope. ImageJ was then used to measure the length and width of muscles, and surface area was calculated in Excel. NMJ length was determined by tracing the complete NMJ arbor of projected samples labeled with anti-HRP from at least 6 animals and representing at least 20 NMJ 4s using the NeuronJ plugin (ImageJ).

All recordings were performed as in (Menon et al., 2015). Briefly, age matched 3rd instar larvae were dissected in 0.3mM calcium HL3 saline (Stewart et al., 1994) (70mM NaCl, 5mM KCl, 20mM MgCl₂, 10mM NaHCO₃, 5mM Trehelose, 115mM Sucrose, 5mM HEPES). Body-wall muscles were visualized under a Nikon FS microscope using a 40× long-working distance objective and preparations were immersed in HL3 saline containing 0.5mM calcium. Recordings were performed by impaling muscle 6 in abdominal segments 3 and 4 with 15–20 MΩ sharp electrodes filled with 3M KCl. Signals were amplified using a MultiClamp 700B (Molecular Devices (MD)) and digitized using a Digidata 1550B (MD). Data was acquired using pClamp 10 software (MD) and analyzed using Mini Analysis software (Synaptosoft). Cut segmental nerves were drawn into a suction electrode and stimulated using a 1ms suprathreshold stimulus via a Master-9 stimulator (A.M.P.I.). Stimulation produced evoked excitatory junctional potentials (EJPs), while unstimulated preparations were recorded passively for miniature excitatory junctional potentials (mEJPs). Analysis was only performed on muscles cells with resting potentials below −60mV. Quantal content was calculated by dividing the mean EJP amplitude by the mean mEJP amplitude for each animal, and the resulting number pooled for each genotype.

Z-stacks of the ventral nerve cords of Tao¹⁶ and control heterozygotes, and D42>Tao RNAi and D42/+ control animals were imaged on a Zeiss LSM 800 confocal at the same settings. From a single section representing the middle of the majority of midline motor neurons, the nuclei from ~2 commissures of each ventral nerve cord (VNC) (20 nuclei/animal) were circled as regions of interest and fluorescence intensity measured in both channels in ImageJ. A background circle was taken on each CNS in each channel and was subtracted from each measurement for that CNS before averaging the intensities. Dividing the intensity of pMad to that of Elav for the same nucleus produced a ratio of pMad:Elav intensity for each nucleus. These ratios were then average for each genotype. At least 8 animals of each genotype were examined.

Crosses to obtain Tao¹⁶ males and Tao¹⁶/w¹¹¹⁸ females were conducted as described above. Each genotype was dissected in Schneider’s media supplemented with 10% FBS, to obtain 20–50 third instar CNSs. These were homogenized in a 1 mL dounce homogenizer using the RLT-plus buffer from a RNeasy plus kit (Qiagen). Further purification of total RNA followed the manufacturer’s instructions. cDNA was synthesized using the iScript kit (Bio-Rad) and used between 115–250 ng total RNA (varied by experimental repeat but was matched between samples). qRT-PCR was performed using custom-designed Taqman assays (Integrated DNA Technologies) in which the forward primer sequence spans an exon-exon junction. These were utilized in a StepOnePlus Real-Time PCR system (Applied Scientific/Thermo Scientific). Rps17 was used as the reference gene for the delta delta Ct comparison protocol to quantify relative changes in gene expression, all in technical triplicate. No-reverse transcriptase controls were conducted to confirm the purity of each cDNA sample, and lack of gDNA contamination in the qRT-PCR. Four independent biological repeats were performed.

One-way ANOVA followed by Tukey’s multiple comparison test or unpaired student’s t-test were performed on raw data, while a Mann-Whitney test was performed on pMad:Elav ratios. All analyses were conducted in Prism 7 software (GraphPad) and differences were deemed statistically significant when p<0.05.

The Hippo pathway is well-characterized as a tumor suppressor pathway in mitotic tissues. However, its role in restricting growth in post-mitotic tissues is less established. During larval development, epithelial tissues undergo tremendous growth that is sensitive to Hippo pathway signaling, raising the possibility that other tissues growing during the same developmental period may also be sensitive to the Hippo pathway. One such structure is the NMJ which must expand to keep pace with the drastic 100-fold increase in muscle surface area during larval growth (Gorczyca et al., 1993; Guan et al., 1996; Miller et al., 2012). Therefore, we investigated whether Hippo signaling could restrict proportional growth of motor neuron arbors during the larval development period.

In order to address the potential role of the Hippo pathway in larval NMJ growth, we used a targeted RNAi approach to knockdown individual components of the pathway (Tao, hippo, warts, and yorkie) in both motor neurons and in muscles. The RNAi lines in this study are well-characterized and reveal strong growth phenotypes in imaginal discs, comparable to somatic mosaic loss-of-function clones (Boggiano et al., 2011; Vissers et al., 2016). The GAL4/UAS system was used to express the RNAi pre- or postsynaptically (Brand & Perrimon, 1993). We found that knockdown of only one component of the core Hippo kinase cascade affected NMJ growth: presynaptic reduction of Tao lead to an increase in bouton numbers (Figure 1A–E, J). This defect is not due to expansion of the postsynaptic muscle since muscle size is unaltered (Supporting Information Figure S1A–B). The overgrowth observed due to loss of Tao is independent of the presynaptic GAL4 driver used, as OK6-GAL4, another motor neuron driver, elicits a similar overgrowth phenotype (Supporting Information Figure S1C). Although we scored muscle 4 NMJs due to ease of visualization and bouton counting, the overgrowth phenotype is independent of the NMJ studied, as the number of boutons on muscles 6/7 also increased (Supporting Information Figure S1D). Postsynaptic RNAi-mediated knockdown of Tao, and other Hippo pathway components, had no effect on synaptic growth or muscle size (Supporting Information Figure S1E–G), suggesting that Tao function is required in motor neurons. The lack of phenotypes due to postsynaptic loss of Tao is also independent of the GAL4 driver, since using BG57-GAL4 (another muscle driver) results in wild-type bouton numbers (Supporting Information Figure S1H).

As a corollary to knockdown of Hippo pathway genes, we also overexpressed Tao and other components to determine if any may be rate-limiting for normal NMJ growth, since hyperactivity of Hippo pathway components in epithelial tissues leads to tissue undergrowth (Boggiano et al., 2011; Ho et al., 2010; Yu et al., 2010). Presynaptic overexpression of all Hippo pathway genes had no effect on NMJ development (Figure 1F–J). We tested two additional Tao overexpression lines as well as a version of Tao carrying a point mutation that renders the protein kinase-dead (Boggiano et al., 2011; Sato et al., 2007), and none showed significant changes in bouton numbers (Supporting Information Figure S1I), suggesting that while Tao is required for normal NMJ development, it may not be rate-limiting in the process.

To confirm that loss of Tao results in overgrown NMJs, we examined additional Tao RNAi transgenes. The presynaptic expression of four independent RNAi transgenes caused a statistically significant overgrowth of NMJ 4 without affecting muscle size (Supporting Information Figure S2A–I), similar to the initial RNAi line used. Furthermore, we utilized a hypomorphic allele of Tao that survives until the pupal stage, Tao¹⁶ (Pflanz et al., 2015), to complement the results from the RNAi transgenes. We observed a statistically significant increase in bouton number compared to heterozygous sibling controls (Figure 2A–C, F), confirming that Tao is required during NMJ development. Importantly, the Tao¹⁶ phenotype is rescued by expression of a Tao transgene in neurons (Figure 2D,F; Supporting Information Figure S2J–K). We were unable to rescue the phenotype with Tao expression in muscles (Figure 2E,F; Supporting Information Figure S2J–K). These data are consistent with the RNAi screen results showing that NMJ development was perturbed only by loss of Tao in neurons.

Tao expression in the larval nervous system has not been reported, though Tao activity is required in larval epithelial tissues (Boggiano et al., 2011; Poon et al., 2011). Utilizing a rabbit anti-Tao antibody, endogenous Tao is found in puncta in motor neuron boutons, supporting a role for presynaptic Tao in NMJ growth (Figure 2G). In Tao¹⁶ mutants, antibody staining was reduced at the NMJ (Figure 2H). Tao is also found in neuron and glial cell bodies in the VNC (Supporting Information Figure S2L,M). Surprisingly, Tao is also expressed in muscles (Figure 2G,H). However, knockdown of Tao expression in muscle does not seem to impact NMJ development (Supporting Information Figure S1E–H), and thus its role in muscles remains unknown.

Our data support a role for Tao that is independent of its conserved function in the Hippo pathway. Knockdown of hippo and warts had no effect on NMJ growth, despite Tao, Hippo, and Warts functioning together to inhibit Yorkie in epithelial growth. In order to provide further support for this model, we reasoned that knocking down Tao levels in combination with yorkie (yki), which on its own has no effect on NMJ development (Figure 1E, J), might suppress the Tao-dependent NMJ overgrowth. However, simultaneous loss of Tao and yki in motor neurons shows the same phenotype as loss of Tao alone (Figure 3), without affecting muscle size (Supporting Information Figure S3). Together, these results suggest that Tao signals through a Hippo-independent pathway to exert its effect on NMJ development.

Each bouton contains several active zones that are responsible for the synaptic communication between the motor neuron and muscle. Given that Tao¹⁶ mutants have an increased number of boutons, we assessed if there was a corresponding increase in active zones. Bruchpilot (Brp) is an integral active zone protein, and Brp puncta serve as a proxy for active zone number (Barber et al., 2017; Liao et al., 2018; Wairkar et al., 2013). In Tao¹⁶ mutants, there are more active zones at NMJ 4 compared to control animals (Figure 4A–E; Tao¹⁶ mean = 335 puncta vs. Tao¹⁶/w¹¹¹⁸ mean = 272 puncta). However, the density of synaptic contacts appears consistent between control and Tao¹⁶ mutant animals (Figure 4F; Tao¹⁶ active zones per μm = 1.64 vs. Tao¹⁶/w¹¹¹⁸ active zones per μm = 1.78). Additional experiments comparing control animals to presynaptic knockdown of Tao show a similar density of Brp puncta (Supporting Information Figure S4A–F). This suggests that the supernumerary boutons in Tao¹⁶ mutants have normally spaced active zones.

To further characterize phenotypes seen when depleting neuronal Tao, we measured the length of each branch of NMJ 4 and summed the lengths of the individual branches to get a total NMJ 4 length value. Tao knockdown animals have an increase in overall branch length compared to wild-type controls without affecting muscle size (Supporting Information Figure S4G–H). Although Tao overexpression does not affect bouton development, we tested if Tao overexpression could decrease branch length, but found no change compared to controls (Supporting Information Figure S4G–H).

The increase in the total number of active zones upon knockdown of Tao suggests that NMJ function may also be altered. To address a role for Tao in NMJ physiology, we measured spontaneous and evoked activity in Tao mutants. Spontaneous events are calcium-independent, single quantal responses and have been implicated in the development of neural networks (reviewed in Andreae & Burrone, 2015). We recorded miniature excitatory junctional potentials (mEJPs) at muscle 6 and found no difference in either the amplitudes or frequencies between Tao mutants and heterozygous controls (Figure 5A–C). Next, we measured evoked excitatory junctional potentials (EJPs) which represent external stimulus driven, calcium-dependent activity, and found a substantial decrease in EJP amplitudes upon removal of Tao (Figure 5D–E). Since there was a decrease in EJP amplitude and no change in mEJP amplitude, quantal content was proportionately reduced (Figure 5F). Therefore, fewer synaptic vesicles (or quanta) are released with every evoked release event.

NMJ development requires the well-characterized retrograde BMP pathway to promote synaptic growth. Since Tao functions independently of the Hippo pathway in NMJ growth, we examined if Tao had any role in BMP-mediated growth. In the canonical BMP pathway, the ligand, Gbb, is released by muscles and binds to the presynaptic type II BMP receptor Wit. In heterozygous gbb or wit animals, the NMJs are comparable to wild-type controls and as discussed above, presynaptic knockdown of Tao leads to expansion of the NMJ. When the heterozygous BMP pathway mutants are introduced into the presynaptic Tao RNAi background, the supernumerary bouton phenotype is suppressed to wild-type or nearly wild-type levels (Figure 6A–G), without affecting muscle size (Supporting Information Figure S6), suggesting that Tao requires the BMP pathway to exert its effect on NMJ development.

We next examined if the BMP pathway is also required for the synaptic function defect observed in Tao mutants. Decrease in presynaptic Tao displayed normal spontaneous activity similar to Tao mutants, and removal of one copy of gbb in the neuronal Tao RNAi background had no effect (Figure 6H–J). Evoked activity was impaired upon reduction of presynaptic Tao, as in the Tao mutants, confirming that the defect in evoked activity is specific to the Tao locus (Figure 6K–L). Loss of one copy of gbb in the Tao RNAi background restored EJP amplitudes to normal levels (Figure 6K–L). Quantal content was also restored when removing one copy of gbb (Figure 6M), confirming that the BMP pathway is required for Tao control of synaptic function.

The novel interaction between the BMP pathway and Tao raises the question of whether Tao is acting upstream or downstream of the BMP pathway. If Tao is upstream, we hypothesized that manipulating Tao levels would affect BMP activity readouts, including phosphorylation of Mad and expression of BMP transcriptional targets. In agreement with this model, we observed a significant increase of nuclear phosphorylated Mad (pMad) in the larval VNC of Tao¹⁶ mutants compared to heterozygous siblings (Figure 7A–F), suggesting that Tao is a negative regulator of BMP activity. A similar increase in pMad was observed with neuronal knockdown of Tao (Supporting Information Figure S7). To further confirm that Tao acts upstream of BMP signaling, we performed qRT-PCR of Cyp6a17, a cytochrome P450 gene that is strongly downregulated in wit mutants (Kim & Marqués, 2010). In Tao¹⁶ mutants we observed a substantial upregulation of Cyp6a17 (~8-fold) compared to heterozygous sibling controls (Figure 7G) consistent with a role for Tao as a repressor of the BMP pathway. Although we do not know how Tao regulates BMP signaling, this novel function for Tao may explain the supernumerary bouton phenotype in Tao loss-of-function.

In this study, we sought to identify novel pathways that control larval NMJ development. We focused on the Hippo pathway due to its known function in regulating growth in various tissues. We tested several components of the Hippo pathway including the transcriptional co-activator Yorkie whose activity is regulated by three kinases (Tao, Hippo, and Warts). Surprisingly, only Tao is required for proper NMJ development. Upon knockdown of presynaptic Tao, the NMJ arbor expands, while the density of active zones is largely unchanged. Functionally, we found no change in spontaneous activity, although evoked responses were significantly reduced upon loss of Tao. Based on our finding that Tao does not require yorkie during NMJ development, we concluded that Tao functions independently of Hippo signaling to restrict NMJ growth. Instead, the dose-sensitive genetic interactions observed between Tao and components of the BMP pathway are indicative of genes that function in the same genetic pathway and suggest that Tao requires BMP signaling during NMJ development and function. Additionally, we found that loss of Tao leads to elevated BMP signaling as evident by an increase in nuclear pMad and increased expression of Cyp6a17, a known BMP target gene, defining Tao as a new negative regulator of BMP signaling.

Though our data posit Tao as a new inhibitor of BMP signaling, we do not yet know its mechanism of action. Two attractive possibilities include Tao negatively regulating BMP signaling at the level of BMP receptor membrane availability, or perhaps by directly regulating pMad levels. For example, the type I BMP receptor Thickveins (Tkv) is negatively regulated via phosphorylation by S6 Kinase-like (S6KL), which marks it for membrane removal and proteosomal degradation (Zhao et al., 2015). Increasing the membrane availability of Tkv in S6KL mutant animals increases BMP signaling, resulting in phenotypes reminiscent of Tao loss-of-function.

Alternatively, Tao could be a regulator of the downstream effector Mad. R-Smads are activated by C-terminal phosphorylation by the BMP type I receptors (Hoodless et al., 1996; Inoue et al., 1998). However, not all phosphorylations on Mad (or its mammalian homologs) are activating. Cyclin-dependent kinase 8 and Shaggy (GSK3 in mammals) are capable of phosphorylating R-Smads in a central linker domain, thereby promoting Smad degradation (Alarcón et al., 2009; Aleman et al., 2014). Drosophila Mad can also be regulated by Nemo phosphorylation at a distinct N-terminal residue (Merino et al., 2009; Zeng et al., 2007), possibly to regulate trafficking or nuclear localization of Mad. Perhaps more intriguing, Misshapen, a Ste20 family kinase (like Tao), and its mammalian homologs can phosphorylate Mad and inhibit its function at yet another distinct site (Kaneko et al., 2011). Of course, Tao might function in the BMP pathway via a different, undescribed mechanism.

An increase in synaptic outgrowth and neurotransmitter release sites, as observed with Tao knockdown, would suggest elevated synaptic vesicle release. However, single synapse analysis reveals that a majority of NMJ synapses are silent (Newman et al., 2017), thus an increase in active zones does not directly correlate to additional functional synapses. Interestingly, knockdown of specific genes that downregulate BMP signaling at the NMJ by promoting internalization of receptors show similar physiological properties as Tao mutants. For example, spinster mutations show a Tkv-dependent NMJ overgrowth phenotype, a decrease in evoked release, and no change in spontaneous amplitude (Sweeney & Davis, 2002). However, other mutations that upregulate BMP activity, such as in S6KL, show similar overgrowth defects, but no change in evoked release and instead an upregulation of spontaneous release amplitude, suggesting either changes in synaptic vesicle loading or changes in postsynaptic glutamate receptors (Zhao et al., 2015). The punctate localization of Tao at the periphery of boutons might suggest a role for Tao in regulating synaptic activity and BMP signaling through a spinster-dependent pathway and thus could be involved in the internalization of receptors.

NMJ development requires proper microtubule dynamics (Nechipurenko & Broihier, 2012) and cell adhesion molecules (reviewed in Menon et al., 2013), both of which are regulated by Tao in other contexts. Mammalian Tao (also known as MARKK) activates Par-1/MARK to inhibit microtubule stability, notably in neuronal cultures (Biernat et al., 2002; Timm et al., 2006). However, in adult Drosophila brains, Tao inhibits Par-1 and positively regulates the microtubule-stabilizing protein Tau during axon outgrowth and neurodegeneration (King et al., 2011; Wang et al., 2007), and there is also evidence that Tao-family kinases can directly phosphorylate Tau (Giacomini et al., 2018). Interestingly, BMP signaling also destabilizes microtubules via Spartin (Nahm et al., 2013). Thus, Tao may regulate microtubule stability through Spartin or Par-1 to affect NMJ development. Tao also negatively regulates cell surface levels of the neural cell adhesion molecule Fasciclin 2 (Fas2) (Gomez et al., 2012). Fas2 is required in NMJs for the proper formation of a synaptic terminal from an axon growth cone structure (Schuster et al., 1996), raising the possibility that Tao might also be acting to similarly regulate levels of Fas2 in developing NMJs.

Tao was recently implicated in Drosophila tracheal morphogenesis by its phosphorylation and activation of a related Ste20 kinase, GckIII (Poon et al., 2018). A role for GckIII in NMJ development has not been reported, but its mammalian homologs have been reported in a variety of axon outgrowth and synapse development processes (reviewed in Chen et al., 2018). In Drosophila, GckIII’s target, the NDR-family kinase Tricornered (Trc), has been implicated in dendritic tiling (Emoto et al., 2006). If Tao functions through a GckIII-Trc signaling module in the NMJ this could provide further insight into how Tao regulates synaptic growth in a Hippo-independent manner. It is noteworthy that only three Tao substrates, Hippo, Par-1, and GckIII, have been identified. Further analysis of Tao in NMJ development may reveal novel substrates which can be further characterized in other Tao-dependent functions.

In summary, Tao plays a previously uncharacterized role in synaptic outgrowth, through negative regulation of BMP signaling. To our knowledge, this is the first report of a Tao kinase as a negative regulator of the BMP pathway, and it will be intriguing to see if Tao-family kinases can inhibit BMP signaling in other developmental contexts as well.