C. elegans STRADα and SAD cooperatively regulate neuronal polarity and synaptic organization

Neurons are polarized cells that receive inputs at dendrites and transmit electrical signals along axons. They undergo a series of morphological transformations to establish axonal and dendritic identities (Dotti et al., 1988). In cultures, a nascent neuron, or a neuron undergoing re-polarization, initially extends multiple neurite projections. The fastest growing neurite develops into an axon and the rest become dendrites. As the axon elongates and contacts its targets, synapses are established to allow for communication in neuronal networks. Recent studies have begun to reveal regulators for axon/dendrite polarization. These include both extrinsic cues, such as morphogens and growth factors, and intrinsic mechanisms, including signalling kinase cascades, their associated scaffolding proteins and factors affecting actin cytoskeleton and microtubule dynamics (reviewed in Arimura and Kaibuchi, 2005; Barnes and Polleux, 2009; Conde and Caceres, 2009; Wiggin et al., 2005).

Although many proposed polarity regulators were characterized in dissociated neuronal cultures, only few have been verified to play a role in the axon/dendrite specification in vivo. One such example is a family of Ser/Thr kinases called SAD (synapses of amphids defective). Initially isolated from a C. elegans screen for mutations affecting synapse development, SAD-1 was the first known regulator for both neuronal polarity and synaptic organization in vivo (Crump et al., 2001; Hung et al., 2007). In sad-1 loss-of-function mutants, both pre-synaptic vesicle and active zone proteins fail to be restricted to the axons of motor and sensory neurons. Synaptic vesicles also fail to form tight clusters in sad-1 mutants, implying aberrant synaptic organization in addition to defects in neuronal polarity (Crump et al., 2001; Hung et al., 2007). SAD-1 kinase activity is crucial during the establishment, but largely dispensable for the maintenance, of neuronal polarity and synaptic organization (Kim et al., 2008).

The SAD kinases display functional conservation in neuronal polarity across species. Two mouse homologs of SAD-1, SAD-A and SAD-B, function redundantly to establish the axon/dendrite polarity in hippocampal and cortical neurons (Kishi et al., 2005). Cortical neurons of the SAD-A^−/−;SAD-B^−/− (Brsk2^−/−;Brsk1^−/− — Mouse Genome Informatics) double knock-out mice show aberrant migration and neurite differentiation, and cultured hippocampal neurons project neurites with no distinct axonal or dendritic properties (Kishi et al., 2005). Reducing the expression of rat SAD-B in the mature central nervous system was also shown to affect neurotransmitter release (Inoue et al., 2006).

The SAD kinases are distantly related to the AMPK kinases (Lizcano et al., 2004). LKB1, a tumour suppressor protein and a master Ser/Thr kinase for an AMPK subfamily of kinases, can phosphorylate the mammalian SAD kinases in vitro (Lizcano et al., 2004). LKB1 was also required for neuronal polarization in vivo (Barnes et al., 2007; Shelly et al., 2007). Similar to the defects observed in the SAD-A^−/−;SAD-B^−/− double knock-out mice, cortical neurons of mice lacking LKB1 failed to differentiate distinct axons and dendrites, and the level of phosphorylation in the activating T-loop of SAD kinases in these neurons was drastically decreased (Barnes et al., 2007). Ectopic co-expression of SAD-A and LKB1 in non-neuronal cells resulted in robust phosphorylation of both SAD-A and its sole known substrate, Tau (Kishi et al., 2005), supporting that LKB1 is a direct activator of the SAD kinases in vivo (Barnes et al., 2007).

LKB1 is the mammalian homolog of par-4, one of the six par (partitioning defective) genes that control cell polarity in C. elegans embryos (Kemphues et al., 1988; Watts et al., 2000). LKB1/PAR-4 has since been established as a general regulator for epithelial cell polarity and cell growth in both invertebrates and vertebrates (Baas et al., 2004; Martin and St Johnston, 2003; Watts et al., 2000). Association of LKB1 with two co-factors, STRAD (Baas et al., 2003) and MO25 (Boudeau et al., 2003; Boudeau et al., 2004; Milburn et al., 2004), was implicated for the stability (Baas et al., 2003; Boudeau et al., 2003), optimal activity (Baas et al., 2004; Lizcano et al., 2004) and nuclear-to-cytoplasmic translocation (Baas et al., 2003; Boudeau et al., 2003; Dorfman and Macara, 2008) of LKB1 in various cell culture studies.

STRAD is a STE-20-related pseudokinase missing some of the essential residues for kinase activity (Baas et al., 2003). The alpha (α) version of STRAD (STRADα) was more effective in enhancing the LKB1 activity in vitro than the beta form of STRAD (Lizcano et al., 2004). Recent studies also implied the importance of STRADα in LKB1-mediated neuronal polarization (Barnes et al., 2007; Shelly et al., 2007). The siRNA-mediated knock-down of STRADα in cultured rat hippocampal neurons abolished the axon-dendrite distinction, similar to knocking down LKB1 (Shelly et al., 2007). Co-transfection of STRADα was also required for the excessive axon formation induced by LKB1 over-expression in cultured mouse neurons (Barnes et al., 2007). However, STRADα was dispensable for LKB1-induced supernumerary axon differentiation in rat hippocampal cultures (Shelly et al., 2007). The effects of STRADα manipulation in vitro could be cell-context dependent, and the analysis of STRADα loss-of-function mutants is important to establish the physiological functions of STRADα.

To investigate the physiological functions of STRADα and its interplay with LKB1 and SAD in vivo, we examined the neuronal phenotypes of the loss-of-function mutants for the sole C. elegans STRADα and LKB1 homologs, strd-1 and par-4, respectively. We also determined their genetic interactions with sad-1. Although both PAR-4 and STRD-1 affect neuronal polarity, surprisingly only STRD-1 shares a linear genetic pathway with SAD-1 to regulate neuronal polarity and synaptic organization. STRD-1 also associates with SAD-1 and modulates its sub-cellular localization. PAR-4, by contrast, also functions through additional effectors to regulate neuronal polarity. These findings suggest that STRD-1 and SAD-1 form a complex that is crucial for neuronal development in C. elegans, whereas PAR-4 and STRD-1 can have distinct effectors to regulate neuronal polarity.

All strains were cultured on NGM plates (Brenner, 1974) seeded with OP50. N2 was used as the wild-type strain. The sad-1(ky289)-null allele (Crump et al., 2001; Hung et al., 2007) was used for all sad-1 analyses. Two strd-1 alleles were used: rr91, identified in a screen for a hyperplasia germline phenotype in dauers, is maintained in a daf-2(e1370) background (P.N. and R.R., unpublished); ok2283 is a deletion- and protein-null allele four times out-crossed against N2. daf-2(e1370) had no effect on the neuronal phenotypes (data not shown). For all par-1 and most par-4 analyses, the temperature-sensitive (ts) alleles zu310ts (K. Kemphues, Cornell University, NY, USA) and it57ts (Morton et al., 1992) were used, respectively. For phenotypic analyses, L4-stage zu310ts, it57ts and control strains were cultured at 15°C and transferred to 22°C when they carried embryos (primarily at 4- to 16-cell stages). Their progenies were examined 24 hours post-L4. For the temperature-shift experiments, it57ts embryos were cultured at 15°C until the desired larval stages were reached before being shifted to 22°C; their phenotypes were examined 24 hours post-L4. Non-conditional alleles of par-4, it33 and it75 (Morton et al., 1992; K. Kemphues) are marked by dpy-21(e428) and maintained as heterozygous animals. Homozygous Dpy animals that laid dead eggs were examined for their polarity phenotypes. nab-1(ok943), rpm-1(ju44) and dsh-2(or302) were used for analyses. Only strd-1 and par-4 allele information was given in figures for simplicity.

pJH21 [P_unc-25-sad-1::gfp (L,S)], pJH731 [P_unc-25-gfp::sad-1(L)] or pJH1090 [P_unc-25-rfp::sad-1(S)] were co-injected with a P_lin-15-lin-15 into lin-15 (n765ts) animals and integrated into the C. elegans genome to generate hpIs1, hpIs99 and hpIs105, respectively. juIs1, oxIs22 and kyIs105 have been previously described (Dwyer et al., 2001; Zhen et al., 2000; Zhen and Jin, 1999). The strd-1 transcriptional reporter, pJH1258 (P_strd-1-nls::gfp), was co-injected with a P_lin-15-lin-15 into lin-15(n765ts). For rescue experiments, pJH73 (P_unc-115-sad-1), pJH1139 (P_unc-25-par-4), pJH1215 (P_unc-25-strd-1) or pJH1369 [P_unc-25-strd-1(D175A)] was co-injected with a P_odr-1-gfp into the indicated strains. For sub-cellular localization experiments, pJH1735 (P_unc-25-rfp::strd-1) was injected into hpIs1 and hpIs99 animals and pJH1673 (P_unc-25-gfp::strd-1) was injected into hpIs105 animals. For punctum counting, pJH1090 [P_unc-25-rfp::sad-1(S)] was injected to N2 animals, and the same chromosomal array crossed into strd-1 mutants. For bi-fluorescent reconstitution experiments, pJH1755 (P_unc-25-nyfp::strd-1) and pJH1756 [P_unc-25-cyfp::sad-1(L)] were co-injected or sequentially injected into N2. Co-injection and sequential injection of pJH1755 with P_unc-25-cyfp, and of pJH1756 with P_unc-25-nyfp, served as controls.

pJH1139 (P_unc-25-par-4) was generated by inserting the 1.8 kb par-4 cDNA (Labbé, McGill University, QC, Canada) into pJH540 (P_unc-25). pJH1215 (P_unc-25-strd-1) and pJH1673 (P_unc-25-gfp::strd-1) were generated by inserting the 3.5 kb strd-1 genomic coding region into pJH540 or pJH401 (P_unc-25::gfp), respectively. pJH1735 (P_unc-25-rfp::strd-1) was generated by inserting wCherry (mCherry optimized for expression in C. elegans; A. Desai, University of California, San Diego, CA, USA) and genomic strd-1 into pJH540, respectively. pJH1258 (P_strd-1::gfp) was generated by inserting the 5.2 kb upstream region of the operon CEOP3708 into pPD95.67 (Fire et al., 1990). pJH21 and pJH73 were previously described (Hung et al., 2007). pJH731 and pJH1090 were generated by inserting long- and short-isoform encoding sequences of sad-1 under the unc-25 promoter with gfp and mCherry, respectively. pJH1369 was generated by mutating the genomic sequence (D175 to A) of pJH1215, and cloning a sequencing-verified mutated region back into pJH1215. pJH1755 and pJH1765 were generated by adding the nyfp and cyfp fragments (Z. W. Wang, University of Connecticut, CT, USA) to the 5′ end of strd-1 in pJH1215, or of sad-1 long isoform in pJH730, which is identical to pJH731 but without gfp.

To analyze juIs1 and oxIs22 in L1 animals, early L1s were hatched from eggs harvested from hypochlorite-treated gravid adults and were fixed as previously described (Kim et al., 2008). To visualize ventral cord (VD) puncta in adults, double-stranded unc-30 RNA was synthesized from pJH573 and injected to juIs1 adults as described previously (Hallam et al., 2002; Hung et al., 2007). The selective elimination of juIs1 expression in dorsal cord (DD) neurons was confirmed by the lack of GFP signals in DD cell bodies, and their progenies were scored 24 hours post-L4. Wholemount staining was performed as described using anti-SAD-1 (Hung et al., 2007).

C. elegans lysate preparation, immunoprecipitation (IP) and immunoblotting (IB) were performed as described previously (Liao et al., 2004). For IP, mouse monoclonal anti-GFP (Roche) and anti-mouse IgG (Santa Cruz Biotechnology) were used. For IB, a rabbit polyclonal antibody against heat-denatured GFP (Nakamura et al., 2008), monoclonal mouse anti-β-tubulin E7 (Developmental Studies Hybridoma Bank) and α-STRD-1, a rabbit antibody against amino acid 1-332, were used.

For the kinase assay, 200 μg of lysates were pre-cleared with mouse IgG and Protein A agarose beads (Roche) and incubated with mouse anti-GFP (Roche) and Protein A beads at 4°C overnight. Washed beads were resuspended in 13 μL of kinase buffer (Kim et al., 2008), to which 1 μg of recombinant human TAU (Sigma) and 1 μL of 1 mM ATP were added, and incubated at 22°C for 30 minutes. 10 μL supernatant was subjected to IB with anti-TAU pS262 (Invitrogen).

Significance testing employed the Wilcoxon rank-sum test (VD phenotypes), the Proportion test (DD and ASI phenotypes) and t-test [SAD-1 punctum counts in R(v2.6.2)]. Analysis with Punctaanalyser was previously described (Kim et al., 2008).

STRD-1 and PAR-4 are the sole C. elegans homologs of the mammalian STRADα and LKB1, respectively. To determine whether they are required for the development of the nervous system, we examined neuronal polarity and synaptic phenotypes in the respective loss-of-function mutants.

Two loss-of-function alleles for strd-1 were employed for analyses. strd-1(rr91) was identified in a genetic screen for mutants with germline hyperplasia (P.N. and R.R., unpublished) and harbours a mutation that leads to a premature stop codon at Q322 (see Fig. S1 in the supplementary material). strd-1(ok2283) deletes from amino acid R174 onward and is a protein-null allele (Fig. 4D; also see Fig. S1 in the supplementary material). In our analyses below, both alleles of strd-1 displayed similar phenotypes, slightly stiff but active locomotion (see Movies in the supplementary material) and reduced brood sizes (data not shown), but had otherwise superficially normal larval development.

A complete loss of function in PAR-4 disrupts the asymmetric partitioning of developmental determinants during the first two embryonic cell divisions and leads to subsequent embryonic lethality (Kemphues et al., 1988; Morton et al., 1992). We therefore employed the strongest temperature-sensitive (ts) allele of par-4, it57ts, that harbours a mutation in the kinase domain and exhibits 100% embryonic lethality at non-permissive temperatures (Kemphues et al., 1988; Morton et al., 1992; Watts et al., 2000). it57ts embryos were cultured at a permissive temperature but shifted during the first few cell divisions to a non-permissive temperature for phenotypic observations. We also examined two non-conditional alleles of par-4, it33 and it75, where some homozygous progenies from the par-4(+/−) heterozygous parents survive embryogenesis owing to maternal contribution of PAR-4 (Kemphues et al., 1988; Morton et al., 1992).

We first examined neuronal polarity in GABAergic motoneurons (Hallam et al., 2002; Hung et al., 2007; Kim et al., 2008). Axons of the six embryonically-born GABAergic DD neurons innervate ventral body wall muscles in the first larval stage (L1) animals (Fig. 1A, top illustration). In wild-type animals, the pre-synaptic juIs1 marker, which expresses a GFP-tagged synaptic vesicle protein in GABAergic neurons (P_unc-25-snb-1::gfp) (Zhen and Jin, 1999), and the post-synaptic oxIs22 marker, a GFP-tagged GABA-A receptor (P_unc-49-unc-49::gfp) (Zhen et al., 2000), appear exclusively along the ventral nerve cord (Fig. 1A,B). By contrast, 100% of sad-1(ky289)-null mutants display numerous ectopic fluorescent signals along the dorsal nerve cord (Fig. 1A,B), representing a fully penetrant polarity defect as previously reported (Crump et al., 2001; Hung et al., 2007). Both par-4(it57ts) and strd-1(rr91) mutants showed polarity defects in L1 DD neurons (Fig. 1A,B). It is noteworthy, however, that although strd-1 mutants displayed a polarity defect as severe as sad-1 mutants, both the expressivity and penetrance of the polarity defect were considerably lower in par-4 mutants (Fig. 1A,B). The phenotypic severity in each mutant was also consistent between the pre-synaptic and post-synaptic markers (Fig. 1A,B), suggesting that the ectopic puncta, probably representing synapses, imply a failure in restricting axonal identity to the ventral neurites.

We further examined neuronal polarity in ventral cord (VD) neurons, the post-embryonically derived GABAergic motoneurons, in adults. As there is no VD neuron-specific marker, we performed unc-30 RNAi on juIs1 animals to selectively eliminate the expression of SNB-1::GFP in dorsal cord (DD) neurons (Hallam et al., 2002; Hung et al., 2007; Materials and methods). In wild-type adults, VD axons innervate ventral body wall muscles (Fig. 1C, top illustration). This is faithfully reflected by the localization of the majority of VD-derived pre-synaptic SNB-1::GFP puncta along the ventral, but not the dorsal, nerve cord in adults (Fig. 1C,D). By contrast, sad-1-null mutants fail to restrict the pre-synaptic marker to VD axons and display ectopic puncta along the dorsal cord (Fig. 1C,D).

Both par-4(it57ts) and strd-1(rr91) mutants also displayed VD polarity defects (Fig. 1C-D). Consistent with the observations made from the L1 DD neurons, although the severity of the VD polarity defect in strd-1 adults was similar to that in sad-1-null animals, par-4 mutants (both ts and non-conditional alleles) displayed a considerably weaker phenotype (Fig. 1C,D; also see Fig. S2A in the supplementary material). The VD defects in both par-4 and strd-1 mutants were fully rescued when PAR-4 or STRD-1 were expressed under a GABAergic neuron-specific unc-25 promoter (Fig. 3B; also see Fig. S4A in the supplementary material), consistent with them functioning cell-autonomously to regulate polarity.

Finally, failure to restrict the pre-synaptic marker to axons was consistently observed in the ASI sensory neurons of par-4(it57ts) and strd-1(ok2283) mutants with an identical trend in their phenotypic severity (see Fig. S3A,B in the supplementary material). Therefore, both par-4 and strd-1, like sad-1, regulate the polarity of multiple C. elegans neurons.

In addition to neuronal polarity, SAD-1 kinase also regulates synaptic organization (Crump et al., 2001; Hung et al., 2007). In wild-type adults, round and discrete juIs1 puncta, representing SNB-1::GFP pre-synaptic vesicle clusters, are observed in the DD axons along the dorsal nerve cord (Fig. 1Ei,iii, left illustration). In sad-1-null mutants, the SNB-1::GFP puncta appear diffuse or smaller (Fig. 1Eiv, right illustration) as previously reported (Crump et al., 2001). strd-1(ok2283) mutants showed diffuse and smaller puncta qualitatively identical to sad-1-null mutants (Fig. 1Eiii-v). This is reflected by the wider distribution curve of the punctum widths in strd-1 mutants when compared with that of wild-type animals, but is similar to the distribution pattern in sad-1 mutants (Fig. 1E, bottom density graph). Therefore, like SAD-1, STRD-1 is also required for both neuronal polarity and synaptic organization. By contrast, all alleles of par-4 animals showed normal SNB-1::GFP punctum morphology (Fig. 1Ei-ii, data not shown), and there was no difference in the distribution of the punctum widths between par-4(it57ts) and wild-type animals (Fig. 1E, top density graph).

Our analyses consistently showed that although strd-1 mutants phenocopied sad-1 mutants, par-4 mutants displayed weaker defects than those of strd-1 or sad-1 mutants. These milder phenotypes could be due to allelic effects: the loss of function in it57ts at non-permissive temperatures might not necessarily inactivate all functions of PAR-4, and the mutations in it33 and it75 are not completely null. In both ts and non-conditional alleles of par-4 mutants, the maternal contribution of PAR-4 activity might also mask the severity of their polarity defects.

To directly test whether par-4 and strd-1 function in the same genetic pathway as sad-1, we assessed the phenotypes of par-4;sad-1(null) and strd-1;sad-1(null) double loss-of-function mutants. We reasoned that if PAR-4 functions solely through SAD-1, the partial loss-of-function of par-4 should cause no further enhancement of the polarity defects in the sad-1-null background. However, for both it57ts and it33 alleles of par-4, the polarity defects in par-4;sad-1 double mutants were more severe than the respective single mutants in VD neurons (Fig. 2A; also see Fig. S2A in the supplementary material). Such an additive effect was also observed in the ASI neuron for it57ts;sad-1 (see Fig. S3C in the supplementary material), suggesting that the polarity phenotypes observed in par-4 mutants were independent of sad-1. We could not test par-4 in the strd-1 mutant background because par-4(it57ts);strd-1(rr91) and par-4(it33);strd-1(ok2283) double mutants were embryonically lethal under all conditions.

To further verify these findings, we examined par-1 mutants. PAR-1, an AMPK-related Ser/Thr kinase, shares the same genetic pathway with PAR-4 to regulate early embryonic cell divisions in C. elegans (Watts et al., 2000). par-1(zu310ts) mutants indeed displayed neuronal polarity defects (Fig. 2A; also see Fig. S3C in the supplementary material) and, similar to par-4, par-1 mutation also enhanced the VD (Fig. 2A) and ASI (see Fig. S3C in the supplementary material) polarity defects in the sad-1-null background. Furthermore, the VD polarity defect in par-1 mutants could be partially rescued by over-expressing PAR-4 in GABAergic neurons (see Fig. S4A in the supplementary material), and the level of VD polarity defect in par-1;par-4 double mutants showed no enhancement over the single mutants (data not shown), further supporting a shared pathway between the two par genes in regulating neuronal polarity.

By contrast, despite having more severe polarity defects than par-4 and par-1, strd-1 mutants showed no enhancement of the polarity defects in the VD or ASI neurons over the sad-1-null background (Fig. 2B; also see Fig. S3D in the supplementary material). These findings suggest that strd-1 and sad-1 act in the same genetic pathway. Over-expressing SAD-1 in the strd-1(rr91) mutant background also partially rescued the VD polarity defect but not vice versa (see Fig. S4B in the supplementary material), consistent with STRD-1 being a pseudokinase that functions through a kinase. Furthermore, no enhancement was observed in the synaptic morphology defect of strd-1;sad-1 double mutants when compared with the single mutants (Fig. 1E, bottom density graph), indicating that strd-1 and sad-1 function together to regulate synaptic organization as well as neuronal polarity.

The co-existence of the separate PAR-1/PAR-4 and STRD-1/SAD-1 functional pathways is supported by all additional genetic interactions we have tested to date. For example, a scaffolding protein Neurabin (NAB-1) functions in the same genetic pathway as SAD-1 to regulate neuronal polarity (Hung et al., 2007). Similar to their genetic interactions with sad-1 mutants, both par-4 and par-1 mutants enhanced the polarity defects of nab-1 mutants (Fig. 2C), whereas strd-1(ok2283) mutants, like sad-1, did not (Fig. 2D). We also examined RPM-1, an E3 ubiquitin ligase required for pre-synaptic development (Liao et al., 2004; Zhen et al., 2000). Mutations in rpm-1 lead to severely reduced and uncoordinated locomotion in the sad-1-null background (our published observation; see Movies in the supplementary material). Similarly, strd-1(ok2283);rpm-1 double mutants were also severely lethargic and uncoordinated, whereas both par-4(it57ts);rpm-1 and par-1;rpm-1 mutants displayed close to normal locomotion similar to their respective single mutants (see Movies in the supplementary material).

These findings and further genetic analyses (see Fig. S4C,D in the supplementary material) indicate that strd-1 functions in the same genetic pathway as sad-1 to regulate neuronal polarity and synaptic organization. Although the embryonic lethality of par-4-null alleles prevented us from revealing the full role of PAR-4, the polarity defects we were able to observe were genetically independent of sad-1, suggesting that PAR-4 is likely to also function through other effectors such as PAR-1. We focused on characterizing the interaction between STRD-1 and SAD-1 henceforth.

To investigate the mechanism through which STRD-1 regulates neuronal polarity and synaptic organization, we first analyzed the expression patterns of STRD-1. A transcriptional reporter was generated by driving GFP expression under the promoter of the operon in which strd-1 resides. In somatic tissues, GFP reporter signals were detected in the pharynx, excretory canal and entire nervous system (Fig. 3A). The expression pattern of STRD-1 overlaps with that of SAD-1, which is specifically expressed in the nervous system (Crump et al., 2001). We further asked whether STRD-1 functions cell-autonomously in neurons. Expressing STRD-1 exclusively in GABAergic motoneurons using the unc-25 promoter fully rescued the VD polarity and the DD synaptic organization defects in strd-1 mutants (Fig. 3B,Ci,ii, density graph). Therefore STRD-1, like SAD-1 (Crump et al., 2001; Hung et al., 2007), functions cell-autonomously in neurons to regulate both neuronal polarity and synaptic organization.

Like mammalian STRADα, STRD-1 lacks some of the essential residues for kinase activity (see Fig. S1 in the supplementary material). These include the DFG motif that is mutated to GFR and the missing threonine and serine residues in the activation loop. The aspartic acid in the catalytic loop for proton transfer that is mutated in STRADα, however, remains intact in STRD-1 (see Fig. S1 in the supplementary material). It raised a possibility that C. elegans STRD-1 might be a functional kinase. To test this, we mutated this aspartic acid to alanine (D175A), and STRD-1(D175A) was expressed in GABAergic motoneurons in strd-1 mutants. Complete rescue of the VD polarity and DD synaptic organization phenotypes was observed (Fig. 3B,Ciii, density curve), suggesting that like its mammalian counterpart, STRD-1 functions as a pseudokinase in C. elegans.

To determine the sub-cellular localization of STRD-1, we expressed fluorescently tagged, functional STRD-1 in GABAergic motoneurons. GFP::STRD-1 localized to the soma and the nucleus (Fig. 3D, top panel) and along the neurites in a punctate pattern (Fig. 3D, bottom panel), similar to many synaptic proteins including SAD-1 (Crump et al., 2001). The sad-1 gene encodes two isoforms with distinct functional specificity (Hung et al., 2007). Whereas the large isoform, SAD-1(L), is sufficient for regulating both neuronal polarity and synaptic organization, the smaller isoform, SAD-1(S), affects only synaptic organization. Both are co-expressed from the hpIs1 marker [P_unc-25-sad-1::gfp(L,S)]. RFP::STRD-1 extensively co-localized with GFP::SAD-1 puncta when expressed in GABAergic neurons (Fig. 4A). Extensive co-localization was also observed between RFP::STRD-1 and a SAD-1(L) marker hpIs99 [P_unc-25-gfp::sad-1(L)] (Fig. 4B), as well as the SAD-1(S) marker hpIs105 [P_unc-25-rfp::sad-1(S)] (Fig. 4C). Therefore STRD-1 is present at synapses and co-localizes with SAD-1.

The extensive co-localization between STRD-1 and SAD-1, in addition to our genetics data, prompted us to examine whether the two proteins interact in vivo. We first addressed this question by immunoprecipitation. From the lysate of hpIs99 [P_unc-25-gfp::sad-1(L)] animals, we immunoprecipitated GFP::SAD-1 with anti-GFP and probed it with anti-STRD-1 (Fig. 4D). Endogenous STRD-1 co-immunoprecipitated specifically with GFP::SAD-1(L) from the hpIs99 lysate, but not from the lysate of strd-1(ok2283);hpIs99 or wild-type animals (Fig. 4D), nor from the IgG pull down of the hpIs99 lysate (data not shown). We further examined whether STRD-1 and SAD-1 associate with each other in vivo via bifluorescent reconstitution, where two split fragments of YFP reconstitute and fluoresce when brought in close proximity by proteins they are fused to (Zhang et al., 2004). The two YFP fragments were tagged to the N-terminus of STRD-1 or SAD-1, respectively, and expressed in GABAergic neurons. YFP fluorescence was detected in animals expressing both fusion proteins (Fig. 4Ei), but not in those expressing either fusion protein with the complementary split YFP vector (Fig. 4Eii,iii), suggesting that STRD-1 and SAD-1 tightly associate with each other in a protein complex in vivo.

Mammalian STRADα affects the cytoplasmic-nuclear translocation, stability and optimal kinase activity of LKB1 in cell cultures. We investigated whether STRD-1 plays similar roles on SAD-1 in vivo. Both GFP- or RFP-tagged SAD-1 isoforms were absent from the nucleus in wild-type animals and strd-1(ok2283) mutants (Fig. 5A; data not shown). Nuclear exclusion of endogenous SAD-1 was also observed in wild-type animals and strd-1 mutants (Fig. 5D, middle panels, arrowheads). Therefore, STRD-1 is unlikely to shuttle SAD-1 from the nucleus to the cytoplasm in mature neurons.

Because both STRD-1 and SAD-1 localized to synaptic regions (Fig. 4A-C), we examined whether STRD-1 affects the localization of SAD-1 at synapses. Compared with wild-type animals, in strd-1(ok2283) mutants fewer puncta and more diffuse fluorescence along the dorsal nerve cord were observed with the hpIs1 marker (Fig. 5Bi-ii,C), indicating that SAD-1::GFP(L,S) was mislocalized in strd-1 mutants. Interestingly, when further examined with isoform-specific SAD-1 markers, the synaptic localization of the long isoform was preferentially affected in strd-1 mutants (Fig. 5Biii-iv versus Cv-vi). A reciprocal examination of the sub-cellular localization of GFP::STRD-1 in sad-1 mutants did not reveal drastic changes (data not shown), suggesting that GFP::STRD-1 localization was largely independent of SAD-1. We did not observe obvious changes of SAD-1::GFP or GFP::STRD-1 localization in par-4(it33) (not shown), suggesting that their subcellular localizations do not critically depend on PAR-4 activity.

We also examined the level and localization of endogenous SAD-1 in wild-type animals and strd-1(ok2283) mutants by antibody staining for SAD-1. There was no gross change in the level of SAD-1 staining (Fig. 5D), suggesting that STRD-1 has little effect on the stability of SAD-1. This finding is consistent with the comparable level of GFP::SAD-1 between wild-type animals and strd-1(ok2283) mutants (Fig. 4D). The dorsal and ventral nerve cords also appeared similar between wild-type animals and strd-1 mutants (Fig. 5D, top two panels). However, as both regions have densely overlaid synapses that may mask subtle localization changes, we examined SAD-1 staining at regions of single synapse resolution. Lateral cords comprise only single or a few axons, allowing the visualization of single synapses. Although SAD-1 staining appeared discretely punctate and evenly spaced along the lateral cords in wild-type animals (Fig. 5Di, bottom two panels), more diffuse staining with gaps was observed in strd-1 mutants (Fig. 5Dii, bottom two panels). Therefore, both marker expression and endogenous staining experiments suggest that STRD-1 is required for the punctate SAD-1 localization at synapses.

To address whether STRD-1 also affects the kinase activity of SAD-1, GFP::SAD-1 that was immunoprecipitated from hpIs99 and strd-1(ok2283);hpIs99 animals (as in Fig. 4D) was subjected to an in vitro kinase assay with the sole known substrate of mammalian SADs, Tau. GFP::SAD-1 from wild-type animals and strd-1 mutants displayed similar kinase activities in phosphorylating Tau (Fig. 5E-F), suggesting that the activity of SAD-1 is not affected by STRD-1 in vitro. This finding, however, does not exclude the possibility that, in vivo, STRD-1 is required for SAD-1 kinase activity on its endogenous substrate(s) upon activation.

The SAD kinases are essential for specifying axon-dendrite identity in developing neurons (Barnes et al., 2007; Kim et al., 2008; Kishi et al., 2005). Recent studies suggest that LKB1 regulates the polarity of developing mouse cortical neurons through activating the SAD kinases (Barnes et al., 2007). Consistent with these findings, PAR-4 also functions cell-autonomously in regulating neuronal polarity in C. elegans (see Fig. S4A in the supplementary material). Although the endogenous expression patterns of post-embryonic PAR-4 are unknown, we observed its neuronal expression using a transcriptional reporter (data not shown). Our temperature-shift experiments further indicated that PAR-4 activity is crucial during the establishment of neuronal polarity (see Fig. S2B in the supplementary material).

STRADα was implicated in LKB1-mediated neuronal polarization (Barnes et al., 2007; Shelly et al., 2007), but the analysis of STRADα loss-of-function mutants has yet to be described. Lacking some of the essential residues for kinase activity, C. elegans STRD-1, like STRADα, functions through other kinases. We found that strd-1 loss-of-function mutants phenocopied the neuronal defects of sad-1 mutants, and strd-1;sad-1 double mutants mimicked the single mutants. strd-1 mutants also displayed slightly shallower and more stiff sinusoidal waves in locomotion, a subtle behavioural deficit that was identical to and characteristic of sad-1 mutants (see Movies in the supplementary material). By contrast, par-4 mutants not only exhibit milder polarity defects when compared with sad-1 and strd-1 mutants, but these defects also appeared additive in par-4;sad-1 mutants. Our findings establish STRD-1 as a key regulator of neuronal polarity and synaptic organization functioning through the SAD-1 kinase in vivo.

STRADα is a component of the LKB1/STRAD/MO25 trimeric complex that affects the stability, localization and optimal activity of LKB1 in various cell culture studies (Baas et al., 2003; Boudeau et al., 2003; Lizcano et al., 2004). It was therefore unexpected to identify consistent differences in the phenotypic severity between par-4 and strd-1 mutants, and, moreover, between their genetic interactions with sad-1. The simplest explanation for the milder phenotypes of par-4 mutants is that the residual PAR-4 activity is sufficient to activate SAD-1 in the presence of STRD-1. Nevertheless, the enhancements in the polarity defects of par-4;sad-1 double mutants unambiguously revealed an additional, sad-1-independent genetic pathway through which par-4 regulates neuronal polarity.

This SAD-1/STRD-1-independent PAR-4 function is likely to be through the conserved PAR-4/PAR-1 kinase cascade that regulates the polarity of non-neuronal tissues (Amin et al., 2009; Kemphues et al., 1988; Martin and St Johnston, 2003). Apart from the evidence presented above, the co-existence of the PAR-1/PAR-4 and STRD-1/SAD-1 functional pathways is supported by additional genetic interactions. For example, consistent with previous studies that placed mammalian and Drosophila PAR proteins downstream of the wnt signalling (Spicer et al., 2003; Sun et al., 2001), mutations in dsh-2, a C. elegans Disheveled gene, did not enhance the VD polarity defects of either par-4 or par-1 mutants (see Fig. S4C in the supplementary material). By contrast, dsh-2 significantly enhanced the defects of both strd-1 and sad-1 mutants (see Fig. S4D in the supplementary material). A completely normal localization of GFP::SAD-1(L) in par-4 mutants (see Fig. S5 in the supplementary material) is also consistent with the presence of additional effectors of PAR-4 in neuronal polarity.

Future studies are needed to identify additional activators or regulators of SAD-1. We have not determined whether C. elegans homologues for MO25, MOP-25.1, MOP-25.2 and a more divergent MOP-25.3, participate in SAD-1-mediated neuronal development. Aside from LKB1, other candidate SAD activators include a Ca²⁺/calmodulin-dependent protein kinase, recently shown to be a potent activator for SAD-B in vitro (Fujimoto et al., 2008), and the Tuberous Sclerosis Complex Tsc1-Tsc2, also found to modulate the activity of SAD kinases in cultured neurons (Choi et al., 2008). Regardless of which kinase(s) activate SAD-1, the identical neuronal phenotypes among strd-1, sad-1 and strd-1;sad-1 mutants predict that STRD-1 is involved in this process. Although STRD-1 does not affect SAD-1-mediated Tau-phosphorylation in vitro, it does not dismiss the possibility that STRD-1 is required for SAD-1 activation and/or the phosphorylation of its endogenous substrates in vivo.

We demonstrated by both immunoprecipitation and YFP bifluorescent reconstitution that STRD-1 forms a stable protein complex with SAD-1. This is intriguing, as STRADα has only been previously described in the context of the LKB1 complex. About fifty pseudokinases have been identified in the human genome (Boudeau et al., 2006). Together with our findings, this implies that the formation of complexes between a pseudokinase and a kinase could prove to be more prevalent and promiscuous for many other kinases.

STRD-1 also has functions distinct from SAD-1. It is expressed in several tissues besides the nervous system, to which SAD-1 expression is restricted. Indeed, strd-1 mutants display germline defects similar to par-4 mutants, whereas sad-1 mutants do not (P.N. and R.R., unpublished). This observation supports our premise that STRD-1 is a promiscuous pseudokinase capable of functioning through multiple binding partners in different cellular contexts (e.g. early embryogenesis versus neural differentiation). Interestingly, human genetic studies have noted that unlike LKB1, mutations in STRADα are not associated with the cancer-developing Peutz-Jeghers syndrome (Alhopuro et al., 2005; de Leng et al., 2005); instead, a deletion in STRADα leads to distinct clinical features including abnormal brain and craniofacial development (Puffenberger et al., 2007). This might also provide an alternative explanation for the observation in dissociated mouse cortical neurons where the supernumerary axon phenotype induced by LKB1/STRADα co-expression was only partially blocked by siRNA-mediated knock-down of SAD-A and SAD-B (Barnes and Polleux, 2009).

In summary, instead of having only dedicated functions through LKB1, C. elegans STRADα/STRD-1 forms a complex with and functions through the SAD-1 kinase to regulate neuronal polarity and synaptic organization. Moreover, the first direct genetic analyses between the C. elegans STRADα, LKB1 and SAD mutants revealed separate genetic pathways through which LKB1 and SAD regulate neuronal polarity. It will be important to examine STRADα knock-out mutants and also other AMPK kinases including PAR-1 as additional effectors of LKB1 in establishing neuronal polarity in vertebrates.

We thank K. J. Kemphues for strains, J.-C. Labbé for par-4 cDNA, Z. W. Wang for split-yfp tags, A. Desai for the wCherry construct, C. G. C. for strains, T. Kaneko for antibodies against GFP, C. Mok and T. Kawano for Punctaanalyser, P. C. Boutros for assistance in statistical analyses and J. Calarco, B. N. Lilley, H. McNeill and J. R. Sanes for comments on the manuscript. This work was supported by an NSERC fellowship to J.S.M.K., a CIHR grant to R.R. and a CIHR and an NSERC grant to M.Z.