Neuronal polarity is regulated by a direct interaction between a scaffolding protein, Neurabin, and a presynaptic SAD-1 kinase in Caenorhabditis elegans

The development of a functional nervous system requires the maturation of neurons and the establishment of synaptic contacts between neurons and their target cells. Mature neurons are highly polarized cells with morphologically and functionally distinct axons and dendrites. The process of axon and dendrite specification, best observed and most extensively studied in isolated rat hippocampal neuron cultures, is divided into several stages(Dotti et al., 1988). Initially, multiple short and morphologically undifferentiated neurites develop from embryonic neurons. Then, a single neurite extends rapidly and acquires axonal characteristics, which is followed by the maturation of the remaining neurites as dendrites. The sequential axon and dendrite differentiation events are driven by multiple intrinsic mechanisms (reviewed in Arimura and Kaibuchi, 2005; Wiggin et al., 2005). The Par3-Par6-aPKC `polarity' complex is recruited to the growing axon tip(Shi et al., 2003) where it activates the small GTPase Rac1(Etienne-Manneville and Hall,2001; Menager et al.,2004; Nishimura et al.,2005; Shi et al.,2003). Rac1-driven actin remodeling of cytoskeleton supports the fast extension of the neurite that is required for the specification of axonal fate (Nishimura et al., 2005). Interactions between Par3 and the Rac-specific guanine-exchange factor (GEF)Tiam1 further induce Rac1 activity (Chen and Macara, 2005; Nishimura et al., 2005).

Microtubule dynamics also regulate axon formation. Axons and dendrites display different microtubule organizations and are decorated with different microtubule-binding proteins (MAPs) (Baas et al., 1989). MAP1B and Tau are axon-enriched MAPs(Bouquet et al., 2004; Goold and Gordon-Weeks, 2005; Kempf et al., 1996). Their phosphorylation by kinases, including GSK3β, PAR-1, SAD-A (Brsk2) and SAD-B (Brsk1), reduces their association with microtubules and destabilizes microtubule assembly, which is a process that facilitates the initiation of axon outgrowth and specification (Biernat et al., 2002; Kishi et al.,2005; Trivedi et al.,2005).

Although primary neuronal cultures have been the most widely used system to study neuronal polarity, in vivo systems are essential for the elucidation and functional validation of neuronal-polarity regulators(Rolls and Doe, 2004). The fully elucidated neural-circuit diagrams(White et al., 1986) and the development of fluorescent markers for nerve processes in C. elegansallow for in vivo analysis of neuronal polarity. Neuronal polarity can be observed in both sensory and motor neurons using synaptic components, which are stereotypically restricted to specific regions of nerve processes, as markers to distinguish the axonal and dendritic processes. Recent studies have revealed that the wnt signaling pathway is required for anteriorly-directed axonal extension in mechanosensory neurons(Hilliard and Bargmann, 2006; Pan et al., 2006; Prasad and Clark, 2006). SYD-1, a putative Rho GTPase-activating protein, restricts presynaptic proteins to the axons of both motoneurons and chemosensory neurons(Hallam et al., 2002).

Loss-of-function mutations in the C. elegans sad-1 gene, a member of the conserved SAD-family serine-threonine kinase, lead to axon-termination defects, diffuse synaptic-vesicle clustering and the abnormal accumulation of presynaptic proteins in the dendrites of the DD-class GABAergic motoneurons(Crump et al., 2001),suggesting that SAD-1 regulates both neuronal polarity and synapse formation. Morpholino-induced downregulation of the ascidian SAD-family kinase POPK-1 disrupts the proper translocation of maternal mRNAs in ascidian embryos(Nakamura et al., 2005). Double knockout mice of the two mammalian SAD kinases, SAD-A and SAD-B, fail to develop distinct axons and dendrites in cortical and hippocampal neurons,and they exhibit a reduced level of MAP Tau1 phosphorylation(Kishi et al., 2005),suggesting that they function redundantly to specify neurite identity. A recent report suggests that SAD-B associates with synaptic vesicles and active zones in mature synapses, and may also regulate synaptic transmission(Inoue et al., 2006). The molecular pathways through which SAD kinases function to establish neuronal polarity and synapse formation remain unknown.

To identify genes that regulate or mediate the function of SAD-1, we performed a yeast two-hybrid screen and identified the sole C. elegans homolog of Neurabin (NAB-1) that physically interacts with SAD-1 both in vivo and in vitro. Mammalian Neurabin (NeurabinI) and Spinophilin(NeurabinII) were first isolated as F-actin-binding proteins from the rat brain (Allen et al., 1997; Nakanishi et al., 1997; Satoh et al., 1998). They are scaffolding proteins that interact with multiple partners, including protein phosphatase-1 (PP1), p70 S6 kinase, Rac3, the Rho-specific GEF Lfc and the Rac-specific GEF Tiam1 (Buchsbaum et al.,2003; Burnett et al.,1998; Orioli et al.,2006; Ryan et al.,2005; Terry-Lorenzo et al.,2002a; Terry-Lorenzo et al.,2002b). Spinophilin can also interact with G-protein coupled D2 dopamine receptors and α2 adrenergic receptors(Richman et al., 2001; Smith et al., 1999; Wang et al., 2005). In neurons, both Neurabin and Spinophilin are localized at synapses(Nakanishi et al., 1997),enriched and closely associated with the postsynaptic density in mature neurons (Muly et al., 2004a; Muly et al., 2004b), where they recruit PP1 and Lfc to dendritic spines and regulate their morphology and motility (Ryan et al., 2005; Terry-Lorenzo et al., 2005). The elimination of Neurabin expression by antisenseoligonucleotide blocks neurite formation in cultured neurons(Nakanishi et al., 1997),suggesting a role of Neurabin prior to dendritic-spine maturation. Spinophilin and Neurabin single knockout mice are viable(Allen et al., 2006; Feng et al., 2000) and display altered dopamine-mediated synaptic plasticity; Neurabin and Spinophilin mutants are deficient in long-term potentiation and depression, respectively(Allen et al., 2006; Feng et al., 2000). The viability and mild phenotypes of either single knockout suggest a functional redundancy between these two proteins.

Through biochemical and genetic studies, we now demonstrate that C. elegans Neurabin plays an `earlier' role in neurons, where it physically interacts with, and specifically mediates, the function of the SAD-1 kinase to restrict axonal fate in developing neurites.

All strains were cultured at 22°C. nab-1-deletion mutants, ok943 and gk164 were backcrossed four times against wild-type strain N2 prior to phenotypic and biochemical analysis, and double-and triple-mutant construction.

The nab-1 genomic clone pJH513 contains the 9 kb promoter sequence upstream of ATG, the entire gene and the 1 kb downstream sequence. The NAB-1::GFP clone pJH369 was generated from pJH513 by inserting GFP immediately before the stop codon. The nab-1 mini-gene - which contains a cDNA fragment (encoding the first 378 amino acids) that was combined with a genomic fragment - including the last two introns, was inserted into the C-terminal of Punc-25-GFP and mRFP vectors to create pJH507 (Punc-25NAB-1::GFP) and pJH510 (Punc-25 NAB-1::mRFP), respectively. pJH524(Pmyo-3 NAB-1::mRFP) was created by inserting the NAB-1::mRFP fragment from pJH510 into pPD95.86 (Fire Vector kit 1995). pJH617, pJH636 and pJH841 are deletions of pJH510, expressing NAB-1Δ1-190,NAB-1Δ204-387 and NAB-1Δ286-387, respectively. Punc-25-SNB-1::mRFP (pJH505) was constructed by inserting the SNB-1 sequences into Punc-25 mRFP. pJH439 and pJH470 are N-terminal mRFP-fusion expression plasmids with a sad-1 mini-gene C4EA(Crump et al., 2001) and the unc-10 genomic sequence inserted into Punc-25 mRFP,respectively. pJH101 was generated by inserting the Punc-115 promoter in front of the sad-1 mini-gene C4EA. The SAD-1ΔDKV expression vector (pJH447) was created by mutating K910 to a stop codon and subcloned into pJH101. pJH713 and pJH714 were made by inserting cDNA for the SAD-1 long-and short-isoforms behind Punc-25, respectively. The bait construct for the yeast two-hybrid screen was generated by ligating the sad-1cDNA fragment into the pGKBT7 plasmid (Clontech, Mountain View, CA). pJH164,pJH179, pJH180, pJH181 and pJH186 express LexA fused to SAD-1 amino acids 565-914, 306-584, 581-730, 730-914 and 306-407, respectively. pJH200 is a prey plasmid containing NAB-1 cDNA encoding the PDZ domain in pACT2 (Clontech,Mountview, CA). The unc-30 RNAi plasmid pJH573 was generated by inserting a 0.7 kb unc-30 cDNA fragment into pPD129.36 (Fire Vector kit, 1999).

A yeast two-hybrid screen was performed as described in the Matchmaker protocol (Clontech, Mountain View, CA). 1.8×10⁶ clones were screened on HIS^- plates with 50 mM 3-amino triazole and for the activation of lacZ expression.

Three recombinant proteins consisting of overlapping regions of SAD-1(amino acids 280-565, 406-585 and 565-914) fused to glutathione Stransferase(GST) were used to immunize a goat to generate the anti-SAD-1 antibody(Covance, Denver, PA) and affinity-purify the antibody. Whole-mount staining, C. elegans lysate preparation, western blotting, immunoprecipitation and GST pull-down assays were performed as described previously(Liao et al., 2004).

All GFP and mRFP-tagging constructs were co-injected with the LIN-15 expression vector into lin-15(n765) animals. Stable transgenic lines were obtained after UV irradiation of animals carrying the desired extrachromosomal arrays and backcrossed to N2 four times. All rescuing experiments were performed by co-injecting the rescuing plasmid (20 ng/ml)with the Podr-1-GFP marker into mutant animals.

Double-stranded RNA (dsRNA) was synthesized from pJH573 as described(Fire et al., 1998) and injected at 40 μg/ml dsRNA. Young adult F1 animals that lost GFP signals in all six DD cell bodies but retained all 13 GFP-positive VD-neuron cell bodies were scored for the dorsal and ventral GFP synapse puncta.

In C. elegans sad-1 loss-of-function mutants, a presynaptic vesicle marker, SNB-1::GFP, distributes more diffusely at synapses and accumulates ectopically in the dendritic regions of the DD-type GABAergic motoneurons in the first larval (L1) stage(Crump et al., 2001). This suggests a role for SAD-1 in regulating both synapse morphology and specifying neurite identity in DD neurons. We further examined the role of sad-1in neuronal polarity throughout development.

In wild-type L1 animals, only embryonically born DD-type GABAergic neurons are present and synapse onto ventral muscles. At the end of L1, the ventral DD synapses are removed and new DD synapses are established with dorsal muscles(White et al., 1978). VD-type GABAergic neurons, born at the end of the L1 stage, form synapses with the ventral muscles. This rewiring of GABAergic neurons can be observed using a presynaptic vesicle marker, juIs1, which expresses SNB-1::GFP under the GABAergic neuron promoter unc-25 (Punc-25) that is active in both DD and VD neurons (Hallam and Jin, 1998). In wild-type L1 animals, juIs1 puncta are present along the ventral cord only (Fig. 1A, upper panels), representing synapses by DD neurons along the ventral body muscle. From the second larval stage onward, fluorescent puncta are observed on both the ventral and dorsal sides, representing dorsal synapses by DD neurons and ventral synapses by VD neurons.

Consistent with a previous report on sad-1(ju53) mutants(Crump et al., 2001), we found that ky289, a protein-null allele of sad-1(Crump et al., 2001), and two kinase-defective alleles, hp119 (D187N) and hp124 (G64E),also displayed both dorsal and ventral juIs1 puncta at the L1 stage(Fig. 1A, top panels, not shown for hp119 and hp124). We further examined whether this failure in restricting the localization of presynaptic proteins is accompanied by a similar ectopic accumulation of post-synaptic components. Using oxIs22, a fluorescent GABA-receptor marker (UNC-49B::GFP), we observed a corresponding ectopic accumulation of postsynaptic-receptor clusters on dorsal muscles in L1 sad-1 mutants, suggesting that these ectopic dorsal synaptic-vesicle clusters represent functional synapses(Fig. 1A, lower panels). Therefore, DD neurons fail to restrict axonal fate in neurites, forming synapses with both dorsal and ventral muscles in L1 larval-stage sad-1 mutants.

After the L1 stage, the juIs1 marker is expressed in both DD and VD GABAergic neurons. To examine exclusively the polarity of DD neurons in later developmental stages, we eliminated VD neurons using a lin-5mutation, which specifically abolishes all postembryonic cell divisions,including the events that give rise to VD neurons(Horvitz et al., 1983; Lorson et al., 2000). Adult lin-5(e1348) animals carrying juIs1 displayed 110.6±5.2 (n=15) fluorescent puncta exclusively along the dorsal cord. In adult lin-5; sad-1 animals, synapses were observed only on the dorsal cord (data not shown), suggesting the polarity defect of DD neurons observed at the L1 stage was corrected by the remodeling event. However, in these animals, the number of dorsal synapses was reduced to 76.6±6.0 (n=15, P<0.001; Fig. 1B).

To examine the effect of sad-1 mutations on the polarity of VD neurons that synapse with ventral muscles, we eliminated the expression of juIs1 in DD neurons using a RNA interference (RNAi) method(Hallam et al., 2002). UNC-30 is a GABAergic neuron-specific transcription factor(Eastman et al., 1999) that activates the Punc-25 used for driving juIs1 expression. Injection of double-stranded RNA (dsRNA) against unc-30 at specific concentrations selectively eliminates transcription from the Punc-25promoter in DD neurons and therefore allows for visualization of SNB-1::GFP in VD synapses alone.

In wild-type animals carrying the juIs1 marker, injection of unc-30 dsRNA eliminated GFP signals from DD cell bodies and all synaptic puncta on the dorsal cord (Fig. 2A, wt panels) without affecting the GFP signal in VD cell bodies(Fig. 2B, wt panels). In sad-1 mutants, injection of unc-30 dsRNA at the same concentration also eliminated the GFP signal in DD cell bodies. However,72.5±8.7 (n=15) puncta remained along the dorsal side,suggesting that VD neurons fail to restrict synaptic-vesicle transport to their dendrites (Fig. 2A, sad-1 panels). Similarly, these VD neurons also displayed an ectopic distribution of two active-zone protein markers, UNC-10::GFP and SYD-2::GFP,expressed in GABAergic neurons. After unc-30 RNAi injection,UNC-10::GFP and SYD-2::GFP signals were diminished completely from the DD-neuron cell bodies in both wild-type and sad-1-mutant animals. However, intense dorsal SYD-2::GFP (Fig. 2C, wt and sad-1 panels) and UNC-10::GFP (data not shown)puncta remained specifically in sad-1 mutants. Therefore, VD neurons fail to restrict axonal fate in neurites in sad-1 mutants. In addition, a mild but statistically significant decrease of normal ventral synapses was observed in VD neurons in sad-1 mutants(130.3±8.4 for hp124 and 123.8±10.2 for ky289versus 143.7±13.1 for wild-type, n=15, P<0.01; Fig. 2B, sad-1panels). Together, these data indicate that sad-1 mutations cause polarity defects in both DD and VD neurons.

We examined whether the polarity defects of sad-1 loss-of-function mutants is restricted to GABAergic motoneurons. The wdIs20 marker expresses SNB-1::GFP in the VA and DA classes of cholinergic motoneurons from the unc-4 promoter (Miller, III and Niemeyer, 1995). The ventrally located DA8 motoneuron extends a neurite posteriorly, which turns dorsally to join the dorsal nerve cord,where it adopts the axonal fate and forms synapses with dorsal muscles and VD motoneurons. The posteriorly extended ventral process is postsynaptic to several interneurons (White et al.,1986). In a wild-type genetic background, a majority of the animals (87.9%, n=74) showed no ventral SNB-1::GFP puncta posterior to the DA8 cell body along the ventral dendritic process. By contrast, only 32.1% of sad-1-mutant animals displayed wild-type DA8 morphology,whereas the rest showed 3-4 puncta posterior to the DA8 cell body(n=131; Fig. 1C),indicating an ectopic accumulation of synaptic vesicles to the dendritic region of DA motoneurons.

We also examined neuronal polarity in ASI chemosensory neurons that display morphologically distinct dendrites and axons. The short axon from each ASI neuron forms 7-13 en passant synapses with interneurons in the nerve ring while a single long dendritic process extends from the cell body to the tip of the nose where it ends in a ciliated opening(White et al., 1986). This wiring pattern can be directly visualized by the Pstr-3-SNB-1::GFP(kyIs105) marker (Fig. 1D) (Crump et al.,2001). We found that 52% of sad-1 mutants (n=70)displayed dim and diffuse fluorescent puncta along the dendrite, whereas only 11% (n=80) of wild-type animals displayed a sporadic dendritic GFP signal (Fig. 1D). Taken together, we conclude that, in addition to the previously reported severe diffusion of synaptic vesicles, loss of SAD-1 function also leads to the disruption of neuronal polarity in multiple neuron types.

To investigate the mechanisms through which sad-1 regulates neuronal polarity and synapse morphology, we performed a yeast two-hybrid screen to identify SAD-1-interacting proteins. Although both the kinase domain of SAD-1 and its C-terminal non-catalytic regions are essential for SAD-1 function (Crump et al., 2001),we chose the non-catalytic region (amino acids 283-914) of the predicted SAD-1 protein as the bait for the screen. We isolated 34 clones representing eight genes that code for proteins interacting with different regions of the non-catalytic domain of SAD-1.

One of these genes encodes NAB-1, the sole C. elegans homolog of the Neurabin and Spinophilin scaffolding-protein family. By deletion analysis,we determined that the C-terminal region of SAD-1 (amino acids 730-914)mediates this interaction with NAB-1 (Fig. 3A). Because the SAD-1 C-terminus contains a consensus PDZ-binding sequence (Asp-Lys-Val-COOH or DKV motif), and NAB-1 contains a PDZ domain, we tested the ability of this PDZ domain to bind directly to SAD-1 baits. The PDZ domain alone was sufficient to bind full-length SAD-1 (data not shown). Furthermore, deletion of the DKV motif of SAD-1 completely abolished the bait-prey interaction of SAD-1 to either NAB-1(Fig. 3A) or NAB-1 PDZ domain(data not shown), suggesting that this motif mediates the interaction between SAD-1 and NAB-1 in vitro.

The interaction between NAB-1 and SAD-1 was further confirmed by GST pull-down assays. GST alone and GST fused with either full-length NAB-1(GST-NAB-1) or with the NAB-1 PDZ domain (GST-PDZ) were used to precipitate interacting proteins from C. elegans total-protein extracts(Fig. 3B). In C. elegans lysates, anti-SAD-1 antibody recognizes two protein bands, 100 and 110 kD. These two forms were also observed using an anti-FLAG antibody when a FLAG-tagged SAD-1 mini-gene was expressed from the pan-neuronal promoter Punc-115 (Fig. 3B). Both full-length NAB-1 and the PDZ domain of NAB-1 specifically precipitated the 110 kD form of SAD-1 or FLAG-tagged SAD-1(Fig. 3B). The 100 kD band represents a previously unknown isoform of SAD-1 that lacks the last 89 amino acids, including the consensus PDZ-binding site(Fig. 3D).

To determine whether SAD-1 and NAB-1 interact in vivo, we generated a stable transgenic strain, hpIs66, which carries a fully functional GFP-tagged nab-1 genomic clone (data not shown). Immunoprecipitating NAB-1::GFP from total-protein lysates of hpIs66using an anti-GFP antibody also brought down the 110 kD SAD-1 isoform specifically (Fig. 3C, center panels). Conversely, immunoprecipitation using an anti-SAD-1 antibody precipitated NAB-1::GFP from hpIs66 lysate, but not from sad-1(ky289);hpIs66 lysate (Fig. 3C, right panels). Our data show that SAD-1 and NAB-1 physically interact in vivo as well as in vitro.

In the hpIs66 strain that carries functional NAB-1::GFP, GFP was inserted in-frame in the C-terminus, shared by all predicted NAB-1 isoforms with the exception of C43E11.6c (Fig. 4). In western blot analysis using antibodies against GFP, we consistently detected three major forms of NAB-1::GFP that corresponded to the predicted molecular weight of the two longest isoforms, and one band that migrated slower than any of the predicted isoforms(Fig. 3C, left lanes). A deletion allele, ok943, deletes exons 7 to 9 of the nab-1gene, resulting in a premature stop codon immediately following the PDZ domain in all detectable isoforms (Fig. 4). This allele was used for all our subsequent biochemical,genetic and functional analyses.

We used hpIs66 to determine NAB-1 expression during development. NAB-1::GFP expression is restricted to epithelia and neurons. The earliest expression was observed in the hypodermis of 2-fold-stage early embryos(Fig. 5A,B). Immediately prior to hatching, this expression became restricted to the epithelial excretory canal (Fig. 5C-E) and the nervous system, including the central nervous system (nerve ring, Fig. 5C) and the motoneurons(dorsal and ventral nerve cords, Fig. 5D-G). In L3 and L4 larvae, NAB-1::GFP also localized transiently at the membranes of the developing vulva epithelia(Fig. 5E).

SAD-1 is expressed exclusively in the nervous system, and therefore shares an overlapping expression pattern with NAB-1 in neurons. In hpIs66animals, NAB-1::GFP appears punctate along the dorsal and ventral nerve cords(Fig. 5F,G and Fig. 6A), indicative of enrichment at synaptic regions. We examined the subcellular localization of NAB-1::GFP by co-immunostaining with antibodies against various presynaptic proteins.

We found that NAB-1::GFP puncta partially co-localized with the synaptic-vesicle protein SNT-1 (Fig. 6A, left panels) and the active-zone protein UNC-10(Fig. 6A, right panels),suggesting that NAB-1 is present in presynaptic regions that are associated with vesicle pools and active zones. Similar to NAB-1, SAD-1 also showed co-localization with SNT-1 (Fig. 6B, left panels), and a close association with UNC-10(Fig. 6B, right panels). NAB-1::GFP and SAD-1 also showed partial co-localization, where each NAB-1::GFP punctum was associated with SAD-1 staining(Fig. 6C).

In the C. elegans nervous system, synapses formed by adjacent axons in nerve bundles overlap with each other, preventing examination at single-synapse resolution. To examine SAD-1 and NAB-1 localization patterns at the single-synapse level, we co-expressed the GFP-tagged largest isoform of NAB-1 and mRFP-labeled synaptic proteins in GABAergic neurons using the Punc-25 promoter (Eastman et al.,1999; Liao et al.,2004; Yeh et al.,2005). Consistent with the whole-mount staining pattern, Punc-25-NAB-1::GFP showed partial co-localization with Punc-25-UNC-10::mRFP (Fig. 6D, left panels) and Punc-25-SNB-1::mRFP(Fig. 6D, right panels). Punc-25-SAD-1::mRFP showed complete co-localization with Punc-25-SNB-1::GFP (Fig. 6E, left panels), and partial co-localization with Punc-25-UNC-10::GFP (Fig. 6E, right panels). We observed a complete co-localization of Punc-25-NAB-1::GFP and Punc-25-SAD-1::mRFP fluorescent puncta (Fig. 6F), further supporting the idea of a direct interaction between these two proteins. The increased colocalization of NAB-1::GFP and SAD-1::RFP in GABAergic neurons compared with in whole-mount staining may be partially due to the overexpression of two interacting proteins.

If interactions between NAB-1 and SAD-1 are required for their biological functions, mutations in these two genes might result in similar phenotypic defects. Unlike sad-1 mutants, in which synaptic-vesicle clusters are diffuse, we observed fairly normal-shaped synaptic-vesicle clusters in GABAergic (Fig. 2A, nab-1 panels) neurons. Quantitative comparison of the intensity,shape and density of fluorescent vesicle puncta did not show any significant differences between wild-type and nab-1 animals(Fig. 2D). We did not observe any change in synapse morphology in cholinergic and ASI chemosensory neurons in nab-1 mutants either (data not shown), suggesting that nab-1 is not required for synapse morphology in these neurons.

In contrast to the normal synapse morphology, we observed severe polarity defects in nab-1 mutants. DD motoneurons in L1-stage nab-1mutants formed ectopic synapses on dorsal muscles (100%, n=100; Fig. 1A, right panels). As in sad-1 mutants, the DD-polarity defect in nab-1 mutants was also corrected after L1 rewiring, but fewer synapses were made(73.1±9.7, n=18, P<0.001) compared with wild-type animals (110.6±5.2; Fig. 1C). In VD motoneurons, we also observed an ectopic accumulation of presynaptic-vesicle clusters (50.0±6.2, n=15; Fig. 2A, nab-1panels), as well as an ectopic accumulation of the active-zone proteins SYD-2::GFP (Fig. 2C, nab-1 panels) and UNC-10::GFP (not shown) along the dorsal cords of nab-1 mutants. A decrease in the number of normal synapses along the ventral axonal process was also observed in nab-1 mutants(96.6±11.5, n=15, P<0.001; Fig. 2B, nab-1panels).

The polarity of other classes of neurons is also affected in nab-1mutants. Similar to sad-1 mutations, we observed an accumulation of 3-4 ectopic presynaptic-vesicle clusters in the dendritic process of the DA8 cholinergic motoneuron in nab-1 mutants using the wdIs20marker (98.2%, n=58; Fig. 1C, nab-1 panel). In ASI chemosensory neurons, 86% of nab-1 animals (n=80) accumulated bright presynaptic-vesicle clusters in the dendrites when examined by kyIs105(Fig. 1D). As opposed to sad-1 mutants, in which the ectopic synaptic-vesicle clusters are dim and diffuse, all the ectopic vesicle clusters were discrete and normal in morphology in nab-1 mutants (Fig. 1D), supporting a specific role of nab-1 in neuronal polarity.

To determine whether NAB-1 functions cell-autonomously in regulating neuron polarity, we tested whether the polarity defects in VD neurons can be rescued by the specific expression of NAB-1 in GABAergic neurons. Expression of the largest isoforms of NAB-1 from Punc-25 in nab-1-mutant animals eliminated ectopic VD dorsal-synapse formation to the same degree as rescues by a full-length nab-1 genomic construct or by expression of the largest isoforms of NAB-1 from the pan-neuronal promoter(Punc-115) (Fig. 7). The same NAB-1 isoform expressed by the muscle-specific promoter Pmyo-3 failed to rescue any defects(Fig. 7). Similarly, we observed the same rescues of ventral synapse numbers when NAB-1 was expressed in neurons using different neuron-specific promoters, but not the muscle-specific promoter. Wild-type animals have 141.75±13.1(n=15) ventral synapses, whereas nab-1 mutants displayed 95.5±11.2 (n=17) ventral synapses (P<0.001). Expression of NAB-1 from neuron-specific promoters or from the genomic nab-1 construct (Punc-25, 125.6±6.9, n=15; Punc-115, 125.3±9.7, n=15; Pnab-1,125.6±3.3, n=15) in nab-1 animals can restore the ventral synapse numbers to close to wild-type levels (P>0.05). By contrast, NAB-1 expression from the muscle-specific promoter (Pmyo-3,91.8±7.4, n=15) displayed a similar number of ventral synapses as nab-1 mutants (P>0.05). Therefore, NAB-1 is required in neurons to regulate their polarity.

To examine whether the physical interaction between NAB-1 and SAD-1 is required for establishing neuronal polarity, we generated an in-frame deletion of the PP1-binding site and PDZ domain (NAB-1Δ204-378), and an in-frame deletion of the PDZ domain alone (NAB-1Δ286-378) in the largest isoform of the nab-1 gene. Expression of both truncated forms of NAB-1 in GABAergic neurons by Punc-25 failed to rescue polarity defects or synapse number in VD neurons (Fig. 7). This is in contrast to an in-frame deletion of the non-evolutionarily conserved N-terminal portion of NAB-1 protein(NAB-1Δ1-194), which rescued all defects in VD neurons as effectively as the full-length NAB-1 (Fig. 7). Moreover, we did not find any polarity defects in gk164, a nab-1 deletion mutant that deletes exon 2 of nab-1 gene(Fig. 4), leading to an in-frame deletion of 30 amino acids in the non-conserved N-terminal region(data not shown). Because Punc-25-NAB-1Δ204-378::GFP displayed a similar fluorescent intensity and subcellular localization as the Punc-25-full-length NAB-1::GFP (data not shown), we can conclude that the PDZ domain of NAB-1 is specifically required for establishing neuronal polarity.

Mutations in sad-1 lead to polarity defects, as well as other abnormalities in synapse morphology and axon termination not observed in nab-1 mutants (Fig. 1Dand Fig. 2A,D), suggesting that the interaction between NAB-1 and SAD-1 is specifically involved in establishing polarity. To test this hypothesis, we removed the putative NAB-1-binding motif (DKV) by the insertion of a stop codon immediately before the NAB-1-binding site in a SAD-1 mini-gene (SAD-1ΔDKV) and compared its rescuing activity with the original SAD-1 mini-gene. Driven by the pan-neuronal promoter Punc-115, the full-length SAD-1 mini-gene fully rescued synaptic morphology of DD synapses(Fig. 8A) and significantly reduced the number of ectopic dorsal synaptic puncta by VDs in sad-1(ky289)-null mutants (Fig. 8B and Fig. 7C). By contrast, Punc-115-SAD-1ΔDKV failed to reduce the number of ectopic VD synapses (Fig. 8B,C)while fully restoring the morphology of DD(Fig. 8A) and VD(Fig. 8B) synapses,demonstrating that the DKV-mediated SAD-1 interaction with NAB-1 is specifically required for regulating neuronal polarity. The expression of the long isoform of SAD-1 in the GABAergic neurons of sad-1-null mutants consistently rescued polarity defects, whereas the expression of the short-form failed to reduce the ectopic dorsal synaptic puncta in VD neurons(Fig. 8E,F). Therefore, at least in DD and VD GABAergic neurons, we were able to delimit the domain in SAD-1 specifically required for polarity to the PDZ-binding site.

The physical interaction and subcellular co-localization of NAB-1 and SAD-1 to synapses, as well as the overlapping polarity phenotypes of sad-1and nab-1 mutants, suggest that these proteins function together to control neuronal polarity. To further test this hypothesis, we examined the genetic interactions between sad-1 and nab-1 mutants.

We first quantified and compared the number of ectopic dorsal juIs1 puncta in VD neurons in sad-1, nab-1 and sad-1;nab-1 animals after parallel unc-30 RNAi-treatment. In sad-1 and nab-1 mutants, on average, 72.5±8.7(n=15) and 49.7±6.2 (n=15) ectopic dorsal-vesicle puncta were present per animal, respectively. In nab-1; sad-1 mutants, only 52.5±6.5 (n=15) ectopic puncta were present per animal, which did not reflect an additive or enhancing effect of the two mutations. This lack of enhancement of ectopic dorsal VD synapses is not caused by a `saturation' level of ectopic synapses; mutation in another neuronal polarity regulator gene, syd-1, can further enhance the number of ectopic juIs1 puncta. In syd-1 mutants, after unc-30 RNAi treatment, we observed 56.5±6.0 (n=15)ectopic puncta whereas, in nab-1; syd-1 and syd-1; sad-1animals, we observed 87.3±4.5 (n=15) and 86.1±5.4(n=15) ectopic puncta, respectively. syd-1 mutation can even alter the morphology of synaptic-vesicle clusters in nab-1 and sad-1 mutants. Synapse morphology of either nab-1; syd-1 or syd-1; sad-1 mutants appeared extremely diffuse, unlike any of the single mutants (data not shown). These results are consistent with nab-1 and sad-1 functioning in the same genetic pathway.

We observed similar genetic interactions when quantifying defects in DD neurons. After the L1 stage, although the polarity defect of DD neurons had been corrected, the number of DD synapses in both sad-1(73.2±9.7) and nab-1 (76.6±6.1) mutants was reduced when compared with wild-type animals (110.6±5.2)(Fig. 1B). nab-1; sad-1 double mutants showed a similar number of synapses in DD neurons (70.7±4.0, n=15, P<0.001) as either nab-1- or sad-1-mutant alone(Fig. 1B), further supporting the proposal that nab-1 and sad-1 function in the same genetic pathway to regulate neuron polarity.

SAD-family kinases contribute to diverse cellular processes, probably through participating in multiple signaling complexes and regulating the phosphorylation of multiple targets. C. elegans sad-1 mutants display a variety of defects in neuronal development. Here, we reveal that sad-1 regulates axonal fate and synapse morphology through distinct mechanisms. We provide the first biochemical and genetic evidence that C. elegans neurabin (nab-1) physically interacts and functions together with sad-1 to regulate axon-dendrite identity. Furthermore,the NAB-1-SAD-1 interaction is specifically required for neuronal polarity only, allowing a mechanistic separation of different pathways through which SAD-1 regulates the development of the nervous system.

Consistent with a role of SAD-A and SAD-B in regulating neuronal polarity,our previous and current studies also determine that SAD-1 is required for establishing axon-dendrite polarity in a variety of C. elegansneurons. The role of SAD-family kinases in regulating neuronal polarities is thus evolutionarily conserved. A previous study has shown that the overexpression of SAD-1 also induced neuronal polarity defects in the chemosensory ASI neurons in C. elegans(Crump et al., 2001). In this present study, we also noticed synaptic defects induced by high levels of overexpression of SAD-1 (our unpublished observations). Therefore, SAD-1 level is also crucial for proper synapse formation and neuronal-polarity establishment.

The interaction with NAB-1 is required for the role of SAD-1 in neuronal polarization, but is completely dispensable for synapse morphology. NAB-1 may allow the specific activation of SAD-1 kinase by restricting SAD-1 to specific compartments in developing neurites. It could also facilitate the functional specificity of SAD-1 by recruiting SAD-1 regulators or substrates specific for neuronal polarity through its protein-interacting modules. Although our studies demonstrated that NAB-1 and SAD-1 co-localize at presynaptic termini,NAB-1 and SAD-1 do not appear to be required for the subcellular localization of one another. We did not observe obvious subcellular mis-localization of NAB-1::GFP in sad-1 mutants, or vice versa (data not shown). It remains possible that NAB-1 localizes SAD-1 transiently during early neurite outgrowth or that the interaction further refines the localization of these proteins at subdomains of presynaptic termini, because both SAD-1::mRFP and NAB-1::GFP appeared more punctate when co-expressed(Fig. 5F). However, our ability to detect these potential changes is currently limited by a very narrow time window for axon growth and the extremely small size of synapses.

Besides spatially restricting components of signaling pathways and recruiting their regulators and substrates, some scaffolding proteins mediate the cross-talk between signaling pathways. For example, Paxillin, a scaffold for the Raf-1-MEK-ERK MAPK cascade, can recruit and regulate the activation of the focal adhesion kinase, and subsequently Rac, to mediate cell migration(Ishibe et al., 2004; Ishibe et al., 2003). NAB-1 may function in a similar fashion to recruit SAD-1 to a signaling complex that regulates neuronal polarity. Identification of physiological binding partners of NAB-1 will help determine targets, regulators and the signaling pathways through which SAD-1 regulates polarity.

Studies of mammalian Neurabins have focused on the maturation and motility of dendritic spines. In cultured neurons, the F-actin-binding domain is required to promote dendritic-spine maturation(Terry-Lorenzo et al., 2005; Zito et al., 2004). Consistently, both Neurabin- and Spinophilin-knockout mice displayed defects in synaptic plasticity (Allen et al.,2006; Feng et al.,2000; Stafstrom-Davis et al.,2001). Interestingly, invertebrate Neurabins are highly conserved with their mammalian homologs in all other motifs, with the exception of the F-actin-binding domain. Whereas C. elegans neurons have no dendritic spines, at least some Drosophila neurons show dendritic-spine-like structures that are enriched for actins(Scott et al., 2003). Our studies showed that the C. elegans NAB-1 protein controls axon-dendrite determination in a variety of neurons, and that this physiological role depends only on the conserved protein domains between the vertebrate and invertebrate Neurabins. This suggests a potentially conserved role for the Neurabin-protein family during early neurite differentiation. It is possible that mammalian Neurabins play roles prior to dendritic-spine maturation, because blocking their expression inhibits neurite outgrowth in cultured neurons (Nakanishi et al.,1997; Orioli et al.,2006).

The genetic interactions between sad-1 and nab-1 indicate that they function in the same pathway to regulate neuronal polarity. This functional overlap is absent in the regulation of synapse morphology;therefore, SAD-1 regulates synapse morphology through currently unknown mechanisms independent of NAB-1.

In addition to the difference in their requirement for synapse morphology, sad-1 and nab-1 mutants display some differences in the severity of their polarity phenotypes. In both sad-1 and nab-1 mutants, GABAergic and cholinergic motoneurons accumulate ectopic synaptic vesicles in dendrites, and the number of normal synapses is reduced, which could be secondary to the polarity deficits. In VD neurons, the total number of synapses in sad-1 mutants is very slightly reduced;by contrast, nab-1 mutants display a much larger decrease in synapse number. Moreover, the severity of the polarity defects differs slightly in different neurons of nab-1 and sad-1 mutants. While the penetrance and number of dendritic vesicle clusters in DD and VD neurons are comparable in sad-1 and nab-1 mutants, in ASI and DA neurons, nab-1 mutants display a much higher penentrance of dendritic synaptic-vesicle clusters than in sad-1 mutants. This variability suggests that NAB-1 modulates neuron polarity through SAD-1 as well as through other regulators, and that the level of dependence of NAB-1 function on SAD-1 varies in different neurons.

Previous studies have identified few genes that regulate neuronal polarity in C. elegans. SYD-1, a protein with PDZ and Rho-GAP domains, also restricts presynaptic proteins in DD and VD motoneurons(Hallam et al., 2002). Similar to nab-1 and sad-1, disruption of the syd-1 gene does not lead to lethality or severe locomotion paralysis(Hallam et al., 2002). The genetic interactions between sad-1, nab-1 and syd-1 are most consistent with SYD-1 functioning either in parallel with, or independently of, NAB-1 and SAD-1, because syd-1 mutations further enhance the VD polarity defects in both sad-1 and nab-1 mutants. syd-1 mutants also display different phenotypes in DD neurons, which have normal synapse number after the L1 stage in these mutants(Hallam et al., 2002), whereas both nab-1 and sad-1 mutants have a decrease in synapse number (Fig. 1B).

These interactions suggest that multiple genetic pathways regulate neuronal polarity in C. elegans. Our biochemical and genetic analyses have defined two components of one novel signaling pathway through which the interaction between NAB-1 and SAD kinase specifically mediates the restriction of axon fate in neurite differentiation.

We thank D. Miller and J. Kaplan for wdIs20 and nuIs94markers, respectively, and for sharing unpublished data; M. Nonet for antibody against UNC-10 and SNT-1; K. Shen for the Podr-1-GFP injection marker; Y. Wang, J. Kim and H. Li for technical assistance; C. Mok for developing the software to quantify properties of fluorescent markers; L. Brown for help with confocal microscopy; the Caenorhabditis Genetic Center and The C. elegans Gene Knockout Consortium for ok943and gk164 mutants and strains; R. Tsien for monomeric RFP; K. Matsumoto for the yeast two-hybrid library; G. Boulianne and C. Boone for yeast strains; and Y. Kohara and A. Coulson for cDNA clones and cosmids,respectively. We thank H. McNeil for comments on the manuscript. This work was funded by a NSERC grant awarded to M.Z.