Rap2 and TNIK control Plexin-dependent synaptic tiling in C. elegans

During development, neurons form synapses with their fate-determined targets. While we begin to elucidate the mechanisms by which extracellular ligand-receptor interactions enhance synapse specificity by inhibiting synaptogenesis, our knowledge about their intracellular mechanisms remains limited. Here we show that Rap2 GTPase (rap-2) and its effector, TNIK (mig-15), act downstream of Plexin (plx-1) to restrict presynaptic assembly and to form tiled synaptic innervation in C. elegans. Both constitutively GTP- and GDP-forms of rap-2 mutants exhibit synaptic tiling defects as plx-1 mutants, suggesting that cycling of the RAP-2 nucleotide state is critical for synapse inhibition. Consistently, RAP-2 activity is locally suppressed by PLX-1. Excessive ectopic synapse formation in mig-15 mutants causes a severe synaptic tiling defect. Conversely, overexpression of mig-15 strongly inhibited synapse formation, suggesting that mig-15 is a negative regulator of synapse formation. These results reveal that subcellular regulation of small GTPase activity by Plexin shapes proper synapse patterning in vivo.


Introduction
During nervous system development, various instructive and repulsive signaling cues cooperatively direct neurons to form chemical synapses with their appropriate targets. Studies have identified some molecules and elucidated their downstream mechanisms that instruct synaptogenesis such as FGF, Ephrin/Eph, Ig-family of cell adhesion molecules (IgCAMs) and synaptic cell adhesion molecules (SynCAMs) (Shen and Bargmann, 2003;Shen et al., 2004).
Several axon guidance cues and their receptors also play critical roles to inhibit synapse formation (Inaki et al., 2007). Semaphorins (Sema) and their receptors, Plexins, are two conserved families of molecules that have a well-established function to repel axons during development (Kolodkin et al., 1993;1992;Negishi et al., 2005;Tran et al., 2007) and prominent roles contributing to immune system, cardiovascular development and cancer regulation (Epstein et al., 2015;Neufeld et al., 2005;Takamatsu and Kumanogoh, 2012).
In addition to its function as a long-range axon guidance cue during neuronal development, Sema/Plexin signaling plays a critical role as a negative regulator of synapse formation. The role of Sema/Plexin signaling in inhibiting synapse formation was first observed in Drosophila, where ectopic expression of Sema-2a causes elimination of specific neuromuscular junctions (Matthes et al., 1995). In mammals, Sema3E/PlexinD1 specifies sensory-motor connections (Pecho-Vrieseling et al., 2009). Secreted Sema3F locally inhibits spine development through its receptors PlexinA3 and Neuropilin-2 in hippocampal granule cells (Tran et al., 2009), and Sema5A/PlexinA2 signaling inhibits excitatory synapse formation in dentate granule cells (Duan et al., 2014). Sema5B diminishes synaptic connections in cultured hippocampal neurons (O'Connor et al., 2009).
However, little is known about the intracellular mechanisms through which Sema/Plexin signaling inhibits synapse formation.
The cytoplasmic domain of Plexin contains a GAP (GTPase-activating protein) domain that inactivates small GTPases (Oinuma et al., 2004;Rohm et al., 2000). Upon activation by Semaphorins, Plexins repel axon outgrowth by inhibiting R-Ras (Negishi et al., 2005). Recent biochemical and structural analyses demonstrated the GAP domain of mammalian PlexinA3 is specific for Rap GTPases, which belong to the Ras family of GTPases and regulate actin cytoskeleton (Wang et al., 2012;2013). PlexinA3 dimerization by Semaphorin activates its GAP domain, thereby inhibiting Rap1 from inducing neurite retraction. Drosophila PlexA and zebrafish PlexinA1 promote remodeling of epithelial cells by inhibiting Rap1 GTPase during wound healing (Yoo et al., 2016). Another Rap GTPase, Rap2, can inhibit neurite outgrowth (Kawabe et al., 2010).
Similar to Sema/Plexin signaling, Rap GTPases regulate synapse formation and function. Rap2 negatively regulate spine number in cultured hippocampal neurons (Fu et al., 2007). Rap1 and Rap2 regulate synaptic activity by removing AMPA receptors from spines during long-term depression and depotentiation, respectively (J. J. Zhu et al., 2002;Y. Zhu et al., 2005). While the GAP domain of Plexin is critical to inhibit synapse formation (Duan et al., 2014;Mizumoto and Shen, 2013a), we still do not know whether Plexin regulates synapse patterning via Rap GTPases at presynaptic sites.
In Caenorhabditis elegans, Sema/Plexin signaling functions in vulva formation and male ray development (Dalpé et al., 2005;Fujii et al., 2002;Ikegami et al., 2004;Liu et al., 2005;Nakao et al., 2007;Nukazuka et al., 2008;. Using this model system, we previously reported that Sema/Plexin signaling in the nervous system mediates a critical inter-axonal interaction for the tiled synaptic innervation of two DA-class cholinergic motor neurons (DA8 and DA9) (Mizumoto and Shen, 2013a). Cell bodies of nine DA neurons in C. elegans reside in the ventral nerve cord, sending dendrites ventrally and axons dorsally to form en passant synapses onto the dorsal body wall muscles. Even though axons of DA neurons show significant overlap, each motor neuron forms synapses onto muscles within specific sub-axonal domains, which do not overlap with those from neighboring DA neurons. This unique synaptic innervation creates tiled synaptic patterns along the nerve cord (White et al., 1986).
Tiled synaptic innervation occurs within most motor neuron classes and may contribute to the sinusoidal locomotion pattern of C. elegans (White et al., 1986). Using a combination of two fluorescent proteins (GFP and mCherry) fused with the presynaptic vesicle protein, RAB-3, and two tissue specific promoters ( Figure 1) (Mizumoto and Shen, 2013a), we can visualize this synaptic tiling between DA8 and DA9 neurons. We reported that PLX-1 localizes at the anterior edge of the DA9 synaptic domain in axon-axon interactions in a Semaphorin-dependent manner, where it locally inhibits formation of the presynaptic specialization via its GAP domain. Loss of Semas or plx-1 causes anterior expansion of DA9 synaptic domain and posterior expansion of DA8 synaptic domain. This result indicates loss of inter-axonal interactions between DA8 and DA9 neurons. Consistently, Tran et al., also observed excess dendritic spine formation, specifically within the region close to the cell body, in the plexin knockout mouse (Tran et al., 2009). These findings suggest a conserved mechanism by which Sema/Plexin locally inhibits synapse formation.
We previously reported that let-60/KRas gain-of-function mutants showed mild synaptic tiling defects. Since mammalian Plexin acts as a RapGAP and very mild synaptic tiling defects of let-60(gf) mutants, we hypothesized that Rap GTPase is the major downstream effector of PLX-1 to regulate synaptic tiling of DA neurons. Here, we report that rap-2, a C. elegans ortholog of human Rap2A, and its effector kinase mig-15 (TNIK: Traf2-and Nck-interacting kinase) act downstream of PLX-1 to regulate synaptic tiling. plx-1 delineates the border of synaptic tiling by locally inhibiting rap-2 along the DA9 axon. We also found an unexpected role of mig-15 in inhibiting synapse formation. Our results reveal the molecular mechanism underlying Plexin signaling to form fine synaptic map connectivity.

rap-2 functions downstream of plx-1 to regulate synaptic tiling
Three Rap genes exist in the C. elegans genome (rap-1, rap-2 and rap-3). To delineate which Rap GTPase functions downstream of PLX-1 in synaptic tiling between DA8 and DA9 neurons, we first examined the expression patterns of all three rap genes (Figures S1). Among them, only rap-2, an ortholog of mammalian Rap2a, was expressed in motor neurons including DA8 and DA9, while rap-1 and rap-3 were not expressed in these cells (Figures S1A-C). In wild type animals, synaptic domains of DA8 and DA9 neurons did not show significant overlap, creating tiled synaptic innervation ( Figures 1A and 1G). In the plx-1(nc36) null mutant, synaptic domains of DA8 and DA9 expanded posteriorly and anteriorly, respectively. As a result, synaptic domains of these neurons overlapped significantly ( Figures 1B and 1G).
Since the intracellular domain of Plexin contains a RapGAP domain, we hypothesized that RAP-2 preferentially exists in a GTP-bound form in the plx-1 mutants. The G12V mutant is widely used as a constitutively GTP-form of small GTPases including mammalian Rap2A and C. elegans RAP-1 (Kawabe et al., 2010;Pellis-van Berkel et al., 2005). Expression of a constitutively GTPbound form of rap-2(G12V) under the A-type neuron specific promoter, Punc-4, elicited a similar synaptic tiling defect as plx-1 mutants ( Figures 1C and 1G). Expression of wild type rap-2 under the unc-4 promoter did not affect the synaptic tiling pattern, suggesting that G12V mutation but not over-expression of rap-2 caused the synaptic tiling defect ( Figure 1G). We then generated rap-2(G12V) mutants using CRISPR/Cas9 genome editing. We observed the same synaptic tiling defects in three independent rap-2(G12V) mutant alleles (miz16, miz17 and miz18) as in plx-1 mutants ( Figures 1D, 1G and S2). We found a comparable level of gene expression among all three rap-2(G12V) mutants to wild type rap-2 using RT-qPCR( Figure S2B). These results confirm that the rap-2(G12V) mutation itself, not changes in gene expression, underlie the synaptic tiling defect in rap-2(G12V) mutants.
These results suggest that the cycling between GTP-and GDP-forms of RAP-2 is critical to regulate the spatial patterning of synapses.

RAP-2 functions cell autonomously in DA neurons
We next determined the cellular location for rap-2 function. Since rap-2(gk11) null mutants showed a synaptic tiling defect, we conducted tissue specific rescue experiments using tissue-specific promoters as previously described (Mizumoto and Shen, 2013a  However, DA neuron-specific expression using the unc-4c promoter strongly rescued the synaptic tiling defect ( Figures 2C and 2G). DA9-specific expression of rap-2 under the mig-13 promoter partially rescued the synaptic tiling defect ( Figures 2E and 2G). DA9-specific expression of rap-2 rescued the phenotype of anterior expansion of the DA9 synaptic domain but not the posterior expansion of DA8 synaptic domain ( Figures 2H and 2I). These results suggest that rap-2 regulates synapse patterning in a cell-autonomous manner. In contrast to the DA9-specific rescue experiment in rap-2 mutants, DA9-specific expression of plx-1 cDNA was sufficient to rescue synaptic defects in both DA9 and DA8 (Mizumoto and Shen, 2013a) (see discussion).
We also observed that expression of human Rap2a in DA neurons rescued the synaptic tiling defect of rap-2 mutants, suggesting the function of rap-2 in synapse patterning is conserved across species (Figures 2D and 2G). Previous work suggested a partial functional redundancy between rap-1 and rap-2 in C. elegans (Pellis-van Berkel et al., 2005). However, we found that rap-1 expression in DA neurons did not rescue the synaptic tiling defect of rap-2 mutants, suggesting functional diversity between rap-1 and rap-2 ( Figures 2F and 2G). Taken together, we conclude that rap-2 functions cell autonomously in DA neurons to regulate synaptic tiling.

RAP-2 activity is spatially regulated by PLX-1
Previously, we demonstrated that PLX-1::GFP is localized at the anterior edge of the DA9 synaptic domain, where it negatively regulates synapse formation through its cytoplasmic GAP domain ( Figures 3A and 3E) (Mizumoto and Shen, 2013a). In the rap-2(gk11) mutant background, we observed no change in PLX-1::GFP localization but did observe ectopic synapses in the axonal region anterior to the PLX-1::GFP domain ( Figures 3B and 3F). This result is consistent with our hypothesis that rap-2 acts downstream of plx-1 to regulate synaptic tiling. Together with our findings that synaptic tiling requires both GTP-and GDP-bound forms of RAP-2, we speculate that PLX-1 acting at the anterior edge of the DA9 synaptic domain regulates the spatial activity of RAP-2 along the axon.
Due to the low expression of C. elegans RAP-2 constructs in HeLa cells, we were not able to test whether the mammalian FRET sensor can detect C. elegans RAP-2 activity (data not shown).
We then expressed EGFP-Rap2A and mRFP-RalGDS(RBD)-mRFP FRET sensors in DA9 neurons in C. elegans. As human Rap2a rescued the synaptic tiling defect of rap-2(gk11) mutants ( Figure 2G), we reasoned that the activity pattern of human Rap2A should recapitulate that of endogenous RAP-2. We indeed observed lower Rap2A activity at the anterior edge of the DA9 synaptic domain compared to within the synaptic domain ( Figures 4D and 4F). This observation is consistent with the localization of PLX-1::GFP at the anterior edge of DA9 synaptic domain ( Figure 3A) (Mizumoto and Shen, 2013a). Local inhibition of Rap2a activity was strongly diminished in the plx-1 mutant background ( Figures 4D and 4F). Higher Rap2 activity in the synaptic region could be simply due to the presence of synapses within the synaptic domain, rather than Rap2 inactivation by Plexin at the anterior edge of the synaptic domain. To exclude this possibility, we examined the Rap2 activity in unc-104/Kif1A mutants in which no synapses are formed in DA9 axon (Ou et al., 2010). We showed previously that PLX-1::GFP localization to the synaptic tiling border was independent of synapses since it was unaffected in unc-104/Kif1A mutants (Mizumoto and Shen, 2013a). In unc-104 mutants, we observed the same local inhibition of Rap2A activity at the putative synaptic tiling border, but not in unc-104; plx-1 double mutants (Figures 4E and 4G), indicating that Plexin controls local Rap2 activity independent of synapses.
To understand that this local Rap2 inactivation at the synaptic tiling border depends on the localized RapGAP activity of PLX-1, we examined the rescue activity of two PLX-1 mutant constructs, PLX-1(RA) and PLX-1(Sema), neither of which rescued the synaptic tiling defect of plx-1 mutants (Mizumoto and Shen, 2013a). PLX-1(RA) is a GAP-deficient mutant but localizes normally at the anterior edge of the DA9 synaptic domain. PLX-1(Sema) contains intact GAP domain but cannot be activated by the ligand and shows diffused localization due to the deletion of the extracellular SEMA domain (Mizumoto and Shen, 2013a). We observed no local Rap2 inactivation in plx-1 mutant animals expressing these mutant PLX-1 constructs in DA9, while expression of wild type PLX-1 cDNA rescued the local Rap2 inactivation at the anterior edge of DA9 synaptic domain ( Figure 4H).
While we do not fully exclude the possibility that PLX-1 indirectly regulates local Rap2 activity, together with the biochemical evidence that mammalian Plexin acts as RapGAP (Wang et al., 2013;, these data strongly suggests that Plexin localized at the anterior edge of the DA9 synaptic domain locally inactivates Rap2 GTPase to delineate the synaptic tiling border in DA9.
The other two nonsense alleles (rh326: Q439Stop, rh80: W898Stop) also showed identical synaptic tiling defects as mig-15(rh148) (Figures S5A-5E). mig-15(rh80) has a nonsense mutation within the highly conserved CNH (citron/NIK homology) domain, which is required to interact with Rap2 in both mammals and C. elegans (Taira et al., 2004). This suggests a physical interaction between RAP-2 and MIG-15 for synaptic tiling. plx-1 or rap-2 mutants did not enhance the synaptic tiling defect in mig-15 mutants ( Figures 5B-5F). This result is consistent with the hypothesis that mig-15 acts in the same genetic pathway as plx-1 and rap-2.
The PLX-1::GFP patch at the putative synaptic tiling border was unaffected in mig-15 mutants, even though the position of the PLX-1::GFP patch has shifted slightly posteriorly compared with wild type (Figures 3C and 3G), suggesting that mig-15 acts downstream of PLX-1 in regulating synaptic tiling.
Interestingly, the degree of overlap between DA8 and DA9 synaptic domains was even larger in mig-15 mutants than those observed in plx-1 and rap-2 mutants (compare Figures 1G  and 5D). Taken together, these results suggest that mig-15 also acts downstream of additional signaling pathways (see discussion).

mig-15 functions in DA neurons
We then sought to determine in which cells mig-15 functions by conducting tissue specific rescue experiments. Since several mig-15 isoforms (wormbase and data not shown) exist, we used We observed that Pmig-13::mig-15 weakly rescued the posterior expansion of the DA8 synaptic domain ( Figure 6H). This is likely due to the leaky expression of mig-15 in DA8, as the mig-15 genomic fragment without promoter showed slight rescue of the synaptic tiling defect in mig-15(rh148) mutants ( Figure 6F). Kinase dead TNIK mutants act as a dominant-negative (Mahmoudi et al., 2009). Expression of mutant mig-15(kd), which carries the same mutation at the corresponding amino acid of the dominant-negative TNIK ( Figure 6E), in DA neurons caused a severe synaptic tiling defect ( Figures 6C, 6D and 6I). Based on these results, we conclude that mig-15 functions cell autonomously in DA neurons.

mig-15 inhibits synapse formation
We observed that DA9-specific expression of mig-15 under the mig-13 promoter in mig-15 mutants often exhibited a shorter synaptic domain compared to wild type ( Figures 6B and 6G).
So, we speculated that an excess amount of mig-15 inhibits synapse formation. We tested the effect of mig-15 overexpression in the wild type background. Strikingly, DA9-specific mig-15 overexpression in wild type (mig-15(OE)) significantly reduced synapse number compared to wild type ( Figures 7A, 7C, 7D and S6). This reduction occurred without affecting overall morphology of the DA9 neuron ( Figure S4). Conversely, DA9 synapse number was significantly increased in Similarly, mig-15 overexpression significantly reduced synapse number in DD-type GABAergic motor neurons ( Figure S7). These results indicate that mig-15 is a negative regulator of synapse formation. Furthermore, pan-neuronal expression of mig-15 under the rab-3 promoter caused severe uncoordinated locomotion in wildtype animals ( Figure S8). These locomotor defects occurred concomitant with significantly reduced GFP::RAB-3 intensity in the dorsal nerve cord in mig-15 over-expressing animals and without causing significant axon guidance defects ( Figure   S8). Taken together, these data indicate that reduced synapse number by mig-15 overexpression disrupted the proper functioning of the motor circuit.
Importantly, we observed no significant increase in synapse numbers in plx-1 or rap-2 mutants ( Figure S6), suggesting that the role of mig-15 in negatively regulating synapse number is independent of its role in PLX-1/RAP-2 -mediated synaptic tiling ( Figure 8G).
Rap GTPase and TNIK are well-known actin cytoskeleton regulators (Lin et al., 2010;Taira et al., 2004). Previous studies demonstrated that ARP2/3-dependent branched F-actin is required for presynaptic development (P. Chia et al., 2012;P. H. Chia et al., 2014). Branched F-actin visualized by GFP::ut-CH (utrophin calponin homology domain) is enriched within the DA9 synaptic domain ( Figure 7E) (P. Chia et al., 2012;Mizumoto and Shen, 2013a). We predicted that mig-15 negatively regulates synapse formation by re-organizing the branched F-actin at the anterior edge of the synaptic domain. Consistently, we observed longer synaptic F-actin distribution in rap-2(gk11) and mig-15(rh148) mutants ( Figures 7F, 7G and 7I). While GFP::utCH was observed in the posterior asynaptic axonal region or in the dendrite of DA9, synapse formation is likely inhibited by Wnt and Netrin signaling as reported previously (Klassen and Shen, 2007;Poon et al., 2008). Conversely, overexpression of mig-15 in DA9 significantly decreased the length of synaptic F-actin ( Figures 7H and 7I). Overexpression of mig-15 also appeared to decrease the overall amount of synaptic F-actin ( Figure 7H). This result suggests that mig-15 inhibits synapse formation by negatively regulating the formation of synaptic F-actin.  Figure 3D). These results strongly suggest that the PLX-1/RAP-2 signaling pathway can specify the position of synaptic tiling border according to the available number of synapses in each DA neuron ( Figure 8G). It is therefore likely that synaptic tiling is a mechanism to maintain the uniform distribution of the synapses from one class of motor neuron in the nerve cord. Consistently, DA8 synaptic domain did not shift posteriorly when mig-15 was overexpressed in DA9 of the synaptic tiling mutants, plx-1 or rap-2 ( Figures 8B-8D). This result suggests that DA8 no longer senses the reduction of DA9 synapse number in the synaptic tiling mutants. Synapse number was not different between mig-15(OE) and in rap-2(gk11); mig-15(OE) animals ( Figure S6), suggesting that the role of mig-15 in inhibiting synapse number is not dependent on Plexin/Rap2 signaling pathway ( Figure 8G).
In summary, we demonstrate that synaptic tiling is a mechanism to maintain a uniform distribution of synapses from one class of motor neurons along the nerve cord. Further, our results indicate plx-1 and rap-2 play critical roles in this process by coordinating the position of the synaptic tiling border.

Discussion
While much is known about the morphogenic triggers for axon guidance and patterned synapse formation, the downstream sequelae of these intracellular effectors has remained unclear.
We discovered the role of Rap2 GTPase and TNIK in synapse pattern formation in C. elegans.
Since Sema/Plexin signaling processes to inhibit synapse formation are well conserved across species, we propose that Sema/Plexin also utilize Rap2 and TNIK to regulate synapse patterning in mammals as well.

Cell autonomous and non-autonomous functions of Sema/Plexin signaling components
Previously we showed that both smp-1 and plx-1 are necessary and sufficient in DA9, which suggests that smp-1 and plx-1 act cell-autonomously in DA9 and non-autonomously in DA8 to determine the synaptic tiling border. We proposed that Sema/PLX-1 in DA9 send a retrograde signal to DA8 through an unidentified signaling molecule (X) to induce the synaptic tiling pattern in DA8 (Mizumoto and Shen, 2013a). However, we found that both rap-2 and mig-15 act cell autonomously, since our DA9-specific rescue experiment only rescued the DA9 phenotype, but not the DA8 phenotype. This conclusion is further supported since the synaptic tiling defects of these mutants were fully rescued when both neurons express functional rap-2 cDNA or mig-15 genomic DNA. We propose that each neuron utilizes a different set of cell surface proteins but share common intracellular mechanisms to specify synapse patterning. Diverse signaling and cell adhesion molecules, such as atrial natriuretic peptide receptor (NPR) and GPCRs, regulate Rap activity (Birukova et al., 2008;Gloerich and Bos, 2011;Weissman et al., 2004). Screening for these potential Rap regulators should identify novel molecules that interact with Sema/Plexin and act in DA9.

Cycling of Rap GTPase activity in synaptic tiling
We showed that both GDP-and GTP-forms of RAP-2 are required for proper synapse patterning. Considering that PLX-1 regulates the spatial distribution of RAP-2 activity and mig-15 acts downstream of rap-2 in synaptic tiling, RAP-2 may also locally regulate MIG-15(TNIK).
While we did not observe a specific subcellular localization of GFP-MIG-15 in DA9 (data not shown), PLX-1 and RAP-2 may instead regulate MIG-15 activity rather than its spatial localization.
Further biochemical characterization of MIG-15 regulation by GTP-RAP-2 or GDP-RAP-2 will be necessary to understand the exact functions of RAP-2 in synapse patterning.

mig-15(TNIK) may integrate multiple inhibitory cues during synapse formation
mig-15 mutants show a greater degree of overlap between DA8 and DA9 synaptic domains than plx-1 or rap-2 mutants. This effect partially occurs from excess synaptogenesis in the posterior asynaptic domain of both DA8 and DA9 neurons. Previously, we demonstrated that Wnt morphogens and their receptors, Frizzled, instruct synaptic topographic patterning by locally inhibiting synapse formation. Indeed, synaptic tiling defects in mig-15 mutants was somewhat similar to the combined effect of plx-1 and wnt mutants (Mizumoto and Shen, 2013b). TNIK can act as a positive regulator of the canonical Wnt signaling pathway in colorectal cancer cells (Mahmoudi et al., 2009). While we do not know whether the canonical Wnt signaling pathway contributes to local inhibition of synapse formation, we propose that TNIK integrates multiple signaling pathways for precise synapse pattern formation.
In addition to its role in synapse pattern formation, our data indicate that mig-15 also plays a role as a negative regulator of synapse number. Since neither plx-1 nor rap-2 mutants showed significant increase in synapse number in DA9, mig-15 seems to inhibit synapse formation in a different signaling pathway ( Figure 8G). Since we observed global reduction of synaptic actin staining in animals over-expressing mig-15, it is likely that mig-15 controls synapse number via regulating synaptic F-actin.
The exact mechanisms of synaptic actin regulation by TNIK remain undetermined. TNIK could activate JNK kinase pathway (Taira et al., 2004). The MIG-15/JNK-1 signaling pathway inhibits axonal branch formation in sensory neurons in C. elegans (Crawley et al., 2017). In contrast to these well-established role of MIG-15/TNIK as an activator of the JNK pathway, we did not observe any synaptic tiling defect in jnk-1 mutant animals ( Figure S5F and S5G). Our result suggests mig-15 does not inhibit synapse formation through the JNK pathway. Further genetic studies of mig-15 in synaptic tiling will elucidate the molecular mechanisms that underlie the role of MIG-15/TNIK in synapse pattern formation and in decreasing synapse number.

Plexin signaling and diseases
Aberrant neuronal wiring underlies many neurological disorders. Not surprisingly, Semaphorin and Plexin genes are associated with various neurodevelopmental disorders and intellectual disabilities, including autism spectrum disorders (ASD) and schizophrenia (Mah et al., 2006). For example, PLXNB1, SEMA3A, SEMA4D and SEMA6C are significantly upregulated in the prefrontal cortices of schizophrenic patients (Eastwood et al., 2003). However, nonsynonymous variations in the Sema3D gene had a significant protective effect against developing schizophrenia (Fujii et al., 2011). More recent work showed that loss of Sema5A/PlexA2 signaling induces excess excitatory synapse formation in granule cells, which causes ASD-like behavioural defects in mice (Duan et al., 2014). Similar to Sema/Plexin signaling, TNIK is also associated with various neurological disorders, including schizophrenia and intellectual disabilities (Anazi et al., 2016;Potkin et al., 2010). TNIK can also physically bind and act with DISC1 (Disrupted in Schizophrenia 1) to regulate synaptic composition (Q. Wang et al., 2011). So, we propose that the Sema/Plexin/Rap2/TNIK signaling pathway plays a critical role to precisely define synaptic connections and its disruption may induce serious neurological disorders. Interestingly, SNPs in Plexin genes are also associated with extremely high IQ. Recent work suggests that loss of PlexinA1 confers better motor control in rodents, due to increased synaptic connectivity in the corticospinal cord (Gu et al., 2017;Spain et al., 2016). Further studies on the Plexin/Rap2/TNIK signaling pathway in synapse map formation, as presented here, will likely reveal the genetic basis of these disorders and conditions.

Confocal Microscopy
Images of fluorescently tagged fusion proteins were captured in live C. elegans using a Zeiss LSM800 confocal microscope (Carl Zeiss, Germany). Worms were immobilized on 2% agarose pad using a mixture of 7.5 mM levamisole (Sigma-Aldrich) and 0.225M BDM (2,3-butanedione monoxime) (Sigma-Aldrich). Images were analyzed with Zen software (Carl Zeiss) or Image J (NIH, USA). Definition of each parameter is as follows (Mizumoto and Shen, 2013a): DA8/9 overlap: a distance between the most anterior DA9 synapse and the most posterior DA8 synapse, DA8 asynaptic domain: a distance from commissure to the most posterior DA8 synapse, DA9 synaptic domain: a distance between the most anterior and posterior DA9 synapses. Middle L4 (judged by the stereotyped shape of developing vulva) animals were used for quantification. Averages were taken from at least 20 samples. For GFP::Utrophin-CH, we measured the length from the posterior end of dorsal axon to the anterior end of GFP::Utrophin-CH domain. For each marker strain, the same imaging setting (laser power, gain pinhole) and image processing were used for comparing different genotypes.

FLIM was conducted 24 hours after transfection. For expression of FLIM markers in the DA9
neuron, each fusion protein constructs were cloned into AscI and KpnI sites of the pSM vector containing mig-13 promoter using SLiCE method.
A custom-made two-photon fluorescence lifetime imaging microscope was used as described elsewhere (Murakoshi et al., 2011). Briefly, EGFP-Rap2a was excited with a Ti-sapphire laser (Mai Tai; Spectra-Physics) tuned to 920 nm. The X and Y scanning galvano mirrors (6210H; Cambridge Technology) were controlled with ScanImage software (Pologruto et al., 2003). EGFP photon signals were collected an objective lens (60×, 1.0 NA; Olympus) and a photomultiplier tube (H7422-40p; Hamamatsu) placed after a dichroic mirror (FF553-SDi01; Semrock) and emission filter (FF01-625/90; Semrock). A fluorescence lifetime curve was recorded by a time-correlated single-photon-counting board Becker & Hickl) controlled with custom software. For construction of a fluorescence lifetime image, the mean fluorescence lifetime values (τm) in each pixel were translated into a color-coded image. We quantified free EGFP-Rap2a and EGFP-Rap2a undergoing FRET (binding fraction) as described elsewhere (Yasuda et al., 2006). Briefly, we calculated the proportion of EGFP undergoing FRET in individual ROIs using the following formula: where τfree and τFRET are the fluorescence lifetime of free EGFP and EGFP undergoing FRET, respectively.

Statistics
Prism (GraphPad) software was used for statistical analysis. one-way ANOVA was done and corrected for multiple comparisons with posthoc Tukey's multiple comparisons tests done between all genotypes. Student's t-test was used for pairwise comparison. Sample numbers were predetermined before conducting statistical analyses.