The role of regulatory domain interactions in UNC-43 CaMKII localization and trafficking

An important objective in neuroscience is to determine how molecular and cell biological changes in the nervous system play a role in acquiring and storing memories. One of the key regulatory molecules involved in memory acquisition, consolidation and retrieval is the calcium and calmodulin-dependent kinase type II (CaMKII), which was originally identified as an abundant brain protein found at the postsynaptic density (reviewed by Hudmon and Schulman, 2002; Rongo, 2002). Pharmacological blockers of CaMKII or mutations in the mouse alpha subunit of CaMKII reduce long term potentiation (LTP) in the hippocampus (Hinds et al., 1998; Ito et al., 1991; Lisman and Goldring, 1988; Malenka et al., 1989; Malinow et al., 1989; Silva et al., 1992b). Moreover, knockout mice show impaired spatial learning, suggesting that CaMKII regulates the molecular events underlying synaptic plasticity (Silva et al., 1992a). One of the mechanisms by which CaMKII regulates plasticity is by regulating the subcellular localization of AMPA-type glutamate receptors (AMPARs) in response to neural activity and subsequent calcium influx (Hayashi et al., 2000; Rongo and Kaplan, 1999).

CaMKII activity is regulated by calcium through its interaction with calmodulin (Meador et al., 1993). The kinase comprises multiple subunits that associate to form a holoenzyme (Brocke et al., 1999; Gaertner et al., 2004; Hoelz et al., 2003). Each subunit consists of a bi-lobed catalytic domain, which includes an N-terminal lobe containing the catalytic core and a C-terminal lobe containing multiple α helices and an autoinhibitory domain. In the absence of calcium signaling, CaMKII is inactive because the autoinhibitory domain binds to the catalytic domain along two hydrophobic pockets, thereby preventing access to substrate (Yang and Schulman, 1999). In the presence of calcium signaling, calcium-bound calmodulin binds to specific residues within the autoinhibitory domain, causing its displacement and exposing a critical threonine, T286, on the autoinhibitory domain. The T286 residue is phosphorylated by adjacent subunits in the holoenzyme upon displacement by calcium/calmodulin (Hanson et al., 1994; Mukherji et al., 1994; Mukherji and Soderling, 1994; Rich and Schulman, 1998). Once phosphorylated at T286, the autoinhibitory domain can no longer interact with the catalytic domain, thus allowing the catalytic domain to access and thus phosphorylate substrates. Threonine to aspartate (T286D) substitutions can mimic autophosphorylation of CaMKII by introducing charge on the autoinhibitory domain, resulting in the displacement of the domain and calcium-independent activity (Fong et al., 1989; Waldmann et al., 1990). Interestingly, the conformational changes induced by autophosphorylation can allow the catalytic domain to interact with other proteins that are similar in sequence to the autoinhibitory domain. For example, sequences within the NR2B NMDA receptor subunit can bind to CaMKII by mimicking the autoinhibitory domain, resulting in a CaMKII that is locked into an active state (Bayer et al., 2001).

CaMKII is found in dendrites, where it colocalizes at or near the postsynaptic densities (PSDs) of synapses, as well as at non-synaptic clusters (Ouimet et al., 1984; Petersen et al., 2003; Strack et al., 1997; Tao-Cheng et al., 2002). Studies using both dissociated neurons and hippocampal slices indicate that CaMKII localization is regulated in response to activity and autophosphorylation (Leonard et al., 1999; Ouyang et al., 1997; Shen and Meyer, 1999; Shen et al., 2000; Strack et al., 1997). However, CaMKII localization is complicated by the presence of multiple isoforms, which can be targeted to distinct sites of subcellular localization (Gaertner et al., 2004; Shen and Meyer, 1999). Holoenzymes can consist of a mixture of different subunit types, which can alter CaMKII localization. An additional complication is that ischemia can affect the localization of CaMKII, resulting in clustering of the kinase during the production of hippocampal slices and dissociated hippocampal neurons (Tao-Cheng et al., 2002).

An alternative approach to understanding CaMKII function and subcellular localization is to use the genetics and simple nervous system of the nematode C. elegans, as the clear cuticle of C. elegans affords the opportunity to observe protein trafficking and localization to synapses in the intact central nervous system of a living animal. C. elegans possesses only a single isoform of CaMKII, called unc-43, which is alternatively spliced into multiple transcripts (Reiner et al., 1999). Null mutants for unc-43 show increased neuromuscular activity in locomotory, egg-laying and defecation circuits, whereas gain-of-function mutants show decreased activity for these same circuits. However, it is unclear which of the specific unc-43 splice forms conducts each of these different functions. Moreover, the specific neuronal circuits in which unc-43 conducts each of these functions is currently unknown.

UNC-43 has also been shown to regulate the synaptic trafficking of GLR-1, an AMPA-type glutamate receptor, in a specific circuit of about a dozen C. elegans interneurons (Rongo and Kaplan, 1999). Although the exact isoforms that conduct this function are not known, the UNC-43G isoform (WormBase nomenclature) in particular can rescue the unc-43 mutant phenotype with respect to GLR-1 localization (Chen et al., 2005; Rongo and Kaplan, 1999). These GLR-1-expressing interneurons project single unbranched processes with both axonal and dendritic character (henceforth referred to as `neurites' for simplicity) into the nerve ring and ventral cord fascicles, where they receive input from glutamatergic sensory neurons and output to motorneurons through en passant synapses (Chalfie et al., 1985; Hart et al., 1995; Maricq et al., 1995; White et al., 1986). A chimeric GLR-1::GFP protein is localized to sensory-interneuron and interneuron-interneuron synapses; however, in unc-43 null mutants, GLR-1::GFP accumulates in perinuclear puncta in the cell bodies of the interneurons. Cell autonomous expression of the UNC-43G isoform in these neurons is sufficient to rescue these defects in GLR-1 localization. However, the relevant targets of UNC-43 in these neurons are unknown.

Here we show that a GFP-tagged chimeric UNC-43 protein is found to be localized both in clusters along the ventral cord neurites, and in a diffuse unlocalized pool within neurites. Mutations known to mimic autophosphorylation of the kinase result in an increase in the abundance of the kinase in neurites, particularly in the unlocalized neurite pool. We find that two key residues of the catalytic domain that stabilize its interaction with the autoinhibitory domain are also required for the kinase to accumulate in neurites and for its function with respect to GLR-1 localization. Our results suggest that although kinase activity itself does not regulate UNC-43 localization, it nevertheless changes the conformation of the kinase, exposing critical residues in the catalytic core required for it to accumulate in neurites.

Standard methods were used to handle and maintain C. elegans (Wood, 1988). We used the following strains: N2, unc-43(n1186), nuIs25[glr-1::gfp], nuIs68[unc-43::gfp], odIs1[snb-1::gfp] and unc-2(e55).

Transgenic strains were isolated by microinjecting various plasmids (typically at 50 ng/μl) using either lin-15(+) (a gift from J. Mendel, Caltech, Pasadena, CA) or rol-6dm (a gift from C. Mello, University of Massachusetts Medical School, Worcester, MA) as a marker. The resulting transformants were followed as extrachromosomal arrays. The unc-43::gfp transgene was generated by subcloning sequences encoding GFP in-frame at the N-terminus of unc-43 cDNA. Sequences for the resulting chimeric protein were placed into the glr-1 promoter vector pV6 (a gift from Villu Maricq, University of Utah, Salt Lake City, UT). Truncated versions of unc-43::gfp were generated by using a corresponding set of primers to make a PCR product of the partial unc-43 cDNA, which was then used to replace sequences in-frame, encoding the full-length unc-43 cDNA in the unc-43::gfp transgene. Mutated versions of unc-43::gfp were generated by using oligonucleotide-mediated mutagenesis to introduce amino acid substitutions into an unc-43 cDNA, which was then used to replace the unc-43 cDNA in the unc-43::gfp transgene. The P_glr-1::unc-43 transgene was generated by subcloning sequences encoding unc-43 cDNA into the glr-1 promoter vector pV6. Mutated versions of P_glr-1::unc-43 were generated using oligonucleotide-mediated mutagenesis as above. All described constructs were confirmed by DNA sequencing.

Plasmids for expression in S2 insect cell lines were generated by subcloning sequences encoding the unc-43 cDNA into pMT/V5, an expression vector containing the Drosophila metallothionein promoter and a V5 epitope (Invitrogen). Mutated unc-43 cDNA sequences were similarly placed into pMT/V5. Rat CaMKII (a gift from Mary Kennedy, Caltech, Pasadena, CA) and CaMKII(T286D) (a gift from Howard Schulman, Stanford University School of Medicine, Stanford, CA) cDNA sequences were also introduced into pMT/V5. Expression constructs were transfected into growing S2 cells and induced by the addition of copper. Cells were harvested 24 hours after transfection. Cells were lysed in lysis buffer [50 mM PIPES, pH 7.0, 1 mM EGTA, 10% glycerol, 2 mM DTT, 1 mM PMSF, antiprotease cocktail (Roche)] using a dounce homogeniser. The lysates were then centrifuged at 16,000 g for 5 minutes, and the supernatant was collected for analysis. The quantity of wild-type and mutant UNC-43 and CaMKII kinases present in the lysates was determined by immunoblotting using anti-V5 antibodies (Invitrogen). Immunoblot analysis demonstrated that the kinase and its mutant forms were at or near full length.

Kinase activity was measured using the CaMKII-specific substrate AC-2 (KKALRRQETVDAL). Kinase activity was measured at 30°C in reactions containing 50 mM PIPES, pH 7.0, 10 mM MgCl₂, 0.1 mg/ml BSA, 0.1 mM CaCl₂, 50 μM ATP containing [γ-³²P]ATP, 1 μM calmodulin and 20 mM AC-2. The average kinase activity of untransfected S2 cell lysates was subtracted prior to plotting the activities shown in Fig. 4. The calcium/calmodulin-independent activity was determined in the absence of calcium and calmodulin, and with the addition of 5 mM EGTA.

GFP- and RFP-tagged fluorescent proteins were visualized in nematodes by mounting larvae on 2% agarose pads with 10 mM levamisole at room temperature. Fluorescent images were observed using a Zeiss Axioplan II and 100× 1.4 NA PlanApo objective and imaged with an ORCA CCD camera (Hamamatsu) using ImagePro v4.1 and VayTek v6.2 software. Exposure times were chosen to fill the 12-bit dynamic range without saturation. Animals were optically sectioned (0.4 μm) and out-of-focus light was removed with a constrained iterative deconvolution algorithm (VayTek).

To quantify the fluorescence levels of GFP-tagged proteins, images of nematodes were captured by CCD as above, using a constant gain and exposure time (filled the 12-bit dynamic range) for all samples. A fluorescent standard (Labsphere) showed less than 1% drift in signal throughout the imaging session. Images were corrected for coverslip fluorescence by subtracting a background image. Pixel intensity was measured for each animal by quantifying an area of interest that collected the fluorescent ventral nerve cord for each sample. The peak amplitude of clusters (localized signal) was calculated for each two-dimensional projection as the fractional increase over the diffuse background of ventral cord fluorescence (unlocalized baseline). Cluster outlines were automatically calculated for fluorescent signals that were two standard deviations above the unlocalized baseline for wild-type UNC-43::GFP using a macro written for ImagePro. Cluster number was calculated by counting the average number of clusters per 10 μm of neurite length. The amount of unlocalized UNC-43 was calculated by measuring the average fluorescence of the diffuse background of the ventral cord. Statistical comparisons were performed using Prism 4.0 (GraphPad). Images of nematodes expressing fluorescent proteins from extrachromosomal arrays were only captured if the array was present in all the GLR-1-expressing cells previously reported.

An integrated array (nuIs68) of the unc-43::gfp transgene was generated by UV-mediated integration (Rongo and Kaplan, 1999). The nuIs68 integrant was backcrossed four times and was inherited in a mendelian fashion. RNA was prepared from worms carrying two copies of the nuIs68 array using strains homozygous for the array. RNA was prepared from worms carrying one copy of the nuIs68 array by mating homozygous nuIs68 males into wild-type N2 hermaphrodites, then selecting fluorescent cross progeny. Total RNA from mixed stage worms of both genotypes was prepared using Trizol (Invitrogen). First-strand reverse transcription was performed with 800 ng total RNA using iScript (Biorad) as per the manufacturer's instructions. Real-time PCR reactions were performed with 5 ng of template using iQ SYBR Green (Biorad) and a Rotorgene 3000 as per the manufacturer's instructions. Primer sets detected either GFP or dlg-1, a control transcript of an adherens junction protein (Firestein and Rongo, 2001). No product was detected for the GFP primer sets in reversed-transcribed RNA from N2 nematodes. No products were detected for either primer set for RNA that was processed in the absence of reverse-transcription. Standard curves of reversed-transcribed RNA from homozygous nuIs68 were generated and analyzed in triplicate by dilution over three orders of magnitude. Calibration curves were linear over this complete range (R²=0.990 for GFP primers, R²=0.992 for dlg-1 primers). Concentrations were derived by comparing Ct (threshold cycle) values of samples (performed 4-8 times/genotype) to the standard curves. For each genotype, mean GFP expression measurements were normalized to those for dlg-1.

To determine whether the mutant versions of UNC-43 could rescue the GLR-1 localization defects of an unc-43 null mutant, we generated the P_glr-1::unc-43 transgenes either with or without the unc-43 mutations. Individual transgenes were co-injected with rol-6dm as a marker into nuIs25[glr-1::gfp]; unc-43 mutants, and transgenic lines were obtained. Rescue was scored by measuring the cell body levels of GLR-1::GFP that accumulated, using the same methodology used to measure UNC-43::GFP cell body levels described above.

We previously observed UNC-43 subcellular localization in neurons by generating a chimeric protein in which UNC-43 sequences are fused to GFP sequences at the GFP C-terminus (Rongo and Kaplan, 1999). The resulting transgene, which contains the glr-1 promoter, was introduced into nematodes where it expressed UNC-43::GFP specifically in the small number of neurons that also express GLR-1 AMPARs. To further explore UNC-43 localization, we generated an unc-43::cfp transgenic strain using the same approach as for unc-43::gfp. To avoid assaying holoenzymes of mixed composition (e.g. mixes of wild-type and mutant subunits, or GFP-tagged and untagged subunits), we introduced the transgene (and subsequent transgenes) into unc-43(n1186) null mutants that lack the endogenous subunit. The resulting UNC-43::CFP protein was detected in neuron cell bodies (Fig. 1A) and out at neurites (Fig. 1B), where it formed clusters along the ventral nerve cord. We previously showed that the unc-43::gfp transgene is functional as it rescues an unc-43 null mutant phenotype for the locomotion defects when expressed in most neurons using a heat-shock promoter (Rongo and Kaplan, 1999).

UNC-43 regulates the translocation of GLR-1 AMPARs to synaptic clusters in the fascicle of neurites found in the ventral nerve cord (Rongo and Kaplan, 1999). To determine if UNC-43 colocalizes with GLR-1, we introduced the unc-43::cfp transgene into nematodes expressing glr-1::yfp, a transgene containing a full-length GLR-1 tagged with YFP (Rongo et al., 1998). GLR-1 colocalizes with Golgi-resident proteins at punctate structures within the cell body (Glodowski et al., 2005; Shim et al., 2004) (Fig. 1C and data not shown). We detected UNC-43::CFP at punctate structures in the cell body (Fig. 1A), which colocalized with GLR-1::YFP puncta (Fig. 1E). We also found that 46±4% (mean±s.e.m. of 42 animals) of UNC-43::CFP puncta colocalized with GLR-1::YFP puncta in neurites (Fig. 1F). There was also an unclustered, non-synaptic pool of UNC-43::CFP found throughout the ventral cord neurites. We obtained similar results through double-label analysis of UNC-43::GFP and GLR-1::RFP, a version of GLR-1 containing mRFP instead of YFP (Glodowski et al., 2005).

We wanted to test whether the activation of UNC-43 could regulate its translocation within neurons. We engineered the T286D mutation, which mimics autophosphorylation and results in calcium-independent kinase activity, into our UNC-43::GFP transgene (Fong et al., 1989; Waldmann et al., 1990). Like wild-type UNC-43::GFP (Fig. 1G), we found that UNC-43(T286D)::GFP could be detected in the neuron cell body and out at the neurites (Fig. 1H and data not shown). Interestingly, we observed higher levels of fluorescence in the ventral cord neurites for UNC-43(T286D)::GFP than for wild-type UNC-43::GFP (Fig. 1M). By contrast, we observed fewer clusters of UNC-43(T286D)::GFP in the ventral cord than wild-type UNC-43::GFP (Fig. 1O).

One explanation for the different levels found in the ventral cord could be the difference in expression levels for UNC-43::GFP and UNC-43(T286D)::GFP. As our UNC-43::GFP transgenes only generate protein in a small number of neurons (the GLR-1-expressing interneurons), we have been unable to detect the protein by western analysis (data not shown). As an alternative, we measured the levels of different GFP-tagged chimeric proteins by quantitative epifluorescence. To calibrate our measurements, we measured the relative levels of UNC-43::GFP from nematodes that carried either one or two copies of a genome-integrated UNC-43::GFP transgenic array. We confirmed that the transcript levels of these strains reflected the array copy number by measuring unc-43::gfp mRNA using real-time PCR (Fig. 1I). As expected, we detected a twofold difference in unc-43::gfp mRNA (Fig. 1I), and in fluorescence levels both in the cell body (Fig. 1J) and in the neurites (Fig. 1K). We next measured UNC-43::GFP and UNC-43(T286D)::GFP cell body levels and detected no significant difference in expression (Fig. 1L).

To determine how the T286D mutation affects UNC-43::GFP translocation, we next examined and measured the levels of wild-type UNC-43::GFP and UNC-43(T286D)::GFP in neurites. We measured their amounts in the ventral cord by summing both the unlocalized pool of fluorescence distributed throughout the neurites and the fluorescence localized into clusters (defined as peaks of fluorescence that rise above a threshold two standard deviations above the mean unlocalized baseline measurements; Fig. 1M). We also measured the mean level of the unlocalized pool of fluorescence in the cord alone, which excludes the signal in fluorescent clusters (Fig. 1N) and found that UNC-43(T286D)::GFP accumulated within the unlocalized pool at roughly twice the level of UNC-43::GFP. As this difference cannot be explained by differences in expression level, our results suggest that the T286D mutation results in an increased accumulation of UNC-43::GFP throughout the ventral cord neurites, the bulk of which is found in an unlocalized pool within these neurites.

We noticed that there was a twofold reduction in the number of UNC-43(T286D)::GFP clusters relative to UNC-43::GFP clusters (Fig. 1O), and we reasoned that the dimmest population of UNC-43(T286D)::GFP clusters might not be detected above the now higher baseline of unlocalized UNC-43(T286D)::GFP (i.e. ∼59 A.U. compared to ∼27 A.U. for UNC43::GFP; Fig. 1N). To examine this possibility, we plotted frequency distributions (Fig. 1P) for the mean fluorescence levels of clusters for UNC-43::GFP (black solid line) and UNC-43(T286D)::GFP (red dotted line). Consistent with our hypothesis, about 50% of wild-type UNC-43::GFP clusters fall below the baseline level (∼59 A.U.) found for UNC-43(T286D)::GFP (Fig. 1P). If UNC-43(T286D)::GFP had similar cluster fluorescent intensities as wild-type UNC-43::GFP, then we would expect to see about half as many clusters as wild-type UNC-43::GFP, which is consistent with our observations (Fig. 1O). Taken together, our results suggest that by mimicking activation of UNC-43 by the T286D mutation, we can increase the amount of UNC-43 present in the unlocalized pool in neurites.

As a first step to connecting the activity-dependent autoregulation of UNC-43 to its mechanism of subcellular localization, we examined the critical amino acids involved in keeping the kinase in its basal state. The regulation of CaMKII kinase activity occurs through an autoinhibitory domain that blocks the ability of the catalytic site to bind to substrate (Yang and Schulman, 1999) (Fig. 2A,B). A number of charged residues in the CaMKII autoinhibitory domain are thought to help mediate its interaction with the catalytic domain. For example, R283 in the autoinhibitory domain is thought to form an electrostatic interaction with D238 in the catalytic domain (Goldberg et al., 1996; Yang and Schulman, 1999). Another charged residue in the autoinhibitory domain, H282, is thought to form an electrostatic interaction with K148 in the catalytic domain (Goldberg et al., 1996; Yang and Schulman, 1999). Two hydrophobic regions on the catalytic domain of CaMKII are thought to help mediate its interaction with the autoinhibitory domain, and are conserved in UNC-43 (Goldberg et al., 1996; Yang and Schulman, 1999) (Fig. 2A). These residues are all conserved in the autoinhibitory domain of UNC-43 (Fig. 2A).

The T286D mutation, which constitutively activates the kinase, can mobilize UNC-43 to the neurites. One explanation could be that kinase activation per se results in the increased translocation of UNC-43 to the neurites, perhaps by allowing the protein to bind to a mobile substrate. Alternatively, kinase activation might change the conformation of UNC-43, thereby allowing an unknown transport factor to interact with the protein and translocate it to the neurites. For example, the NR2B subunit of the NMDAR is thought to interact with the CaMKII catalytic domain in a similar fashion as the autoinhibitory domain (Bayer et al., 2001). To explore these possibilities, we introduced the previously described mutations into UNC-43::GFP to test whether all or only a subset of mutations that result in calcium-independent activity also result in increased UNC-43::GFP translocation.

Like wild-type UNC-43::GFP (Fig. 3A,B), we found that our mutant UNC-43::GFP proteins could be detected in the neuron cell body and out at the neurites (Fig. 3C-H, and data not shown) when introduced into nematodes. We measured the cellular levels of UNC-43::GFP and its mutant forms, and for most mutants we found no statistically-significant difference when compared to wild-type UNC-43::GFP (Fig. 3I). Three mutant proteins, however, showed a lower cellular level, which could be explained if the corresponding mutations (F98K, E109K and K148E) made the UNC-43::GFP protein less stable.

To determine how the mutations affected UNC-43::GFP localization, we next measured the levels of wild-type UNC-43::GFP and its mutant forms in neurites (Fig. 3J). We measured the amount of UNC-43::GFP in the ventral cord by summing both the fluorescence localized into clusters and the unlocalized pool of fluorescence distributed throughout the core of the neurites. Surprisingly, although the mutations result in increased calcium-independent CaMKII kinase activity (Yang and Schulman, 1999), only a subset altered UNC-43::GFP localization (and in very disparate ways). Mutations in F98K, which reduce the affinity of CaMKII for substrate (Yang and Schulman, 1999), show no effect on UNC-43::GFP localization. Mutations in the two hydrophobic regions of the catalytic domain (I205K and V208E) that stabilize its interaction with the autoinhibitory domain (Yang and Schulman, 1999) behave in a similar manner to T286D with respect to localization: increased levels of unlocalized UNC-43::GFP are found in neurites (Fig. 3J). All three of these interactions occur through the α7 helix of the autoinhibitory domain. The autoinhibitory domain also contains a loop of residues N-terminal to the α7 helix (Fig. 2B). Mutations in this loop (H282K and R283E) result in increased calcium-independent CaMKII kinase activity (Yang and Schulman, 1999), and resulted in increased UNC-43::GFP neurite levels (Fig. 3J). The residues in this loop (H282 and R283) are thought to form electrostatic interactions with their partners in the catalytic domain (K148 and D238, respectively) (Yang and Schulman, 1999). Although mutations of K148 and D238 result in a calcium-independent CaMKII kinase activity (Yang and Schulman, 1999), surprisingly, these same mutations resulted in decreased UNC-43::GFP in neurites (Fig. 3J), which is the opposite phenotype from mutations in their salt bridge partners. These results suggest that although mutations that impair the interaction between the catalytic domain and the autoinhibitory domain result in calcium-independent CaMKII kinase activity, they do not automatically result in increased levels of UNC-43 in neurites. Thus, kinase activation itself does not stimulate UNC-43 neurite accumulation.

We next explored the idea that as yet unknown proteins might interact with UNC-43 by mimicking part of the autoinhibitory domain, thereby forming electrostatic interactions with some of the same residues on the catalytic domain. We generated an UNC-43(1-270)::GFP transgene that expresses a truncated UNC-43 containing only the catalytic domain. We found that UNC-43(1-270) accumulates to the same high levels observed for the α7 mutations (Fig. 3J). We also expressed the autoinhibitory domain and remaining C-terminal residues, called UNC-43(271-520), as a GFP chimera. UNC-43(271-520)::GFP does not accumulate to high levels in neurites (Fig. 3J). We also examined the electrostatic interactions between D238, which is in the catalytic domain, with R283E, which is in the autoinhibitory domain. D238R changes the side chain charge and results in a failure of UNC-43::GFP to accumulate in neurites (Fig. 3C,D). R283E replaces the side chain charge on the other side of this salt bridge, but results in the excessive neurite accumulation of UNC-43::GFP (Fig. 3E,F). We reasoned that a D238R R283E double mutant might partially restore the salt bridge, but would reverse the charge polarity in the bridge. When we introduced both of these mutations into UNC-43::GFP, we found that the mutant protein does not accumulate in neurites (Fig. 3G,H). Our results suggest that the D238R localization phenotype can mask the phenotype of R283E. We suggest that mutations in the autoinhibitory loop and the α7 helix, as well as in the hydrophobic regions of the catalytic domain with which α7 interacts, result in displacement of the autoinhibitory domain and exposure of regions on the catalytic domain required for translocation. Consistent with this idea, mutations in this key region of the catalytic domain can intragenically suppress mutations in the autoinhibitory domain.

We next wanted to confirm that UNC-43 has a similar calcium-dependent kinase activity as CaMKII. To test this possibility, we introduced an unc-43 cDNA into an S2 expression vector, and transfected the resulting construct into S2 cells. For comparison to its vertebrate ortholog, we also transfected a plasmid containing the rat αCaMKII cDNA into S2 cells. We lysed the S2 cells and released similar amounts of UNC-43 and CaMKII protein as detected by western analysis against a C-terminal V5 epitope tag present on both proteins (Fig. 4A). Western blots of the lysates detected proteins of 54 kDa and 50 kDa for UNC-43 and CaMKII, respectively, indicating that the products produced by the insect cell lines were similar to the predicted sizes based on conceptual translation for each protein. To assay kinase activity, we incubated the proteins with autocamtide-2 (AC-2), a peptide substrate that contains the CaMKII R-X-X-S/T consensus phosphorylation site (Hanson et al., 1989). We incubated the kinase reactions either in EGTA or varying concentrations of calcium, and found that both UNC-43 and CaMKII calcium-dependent kinase activity was present (Fig. 4B and data not shown). We found that the kinase activity increased in response to increasing levels of free calcium, and that the maximum activity of both UNC-43 and CaMKII was observed at approximately 100 μM calcium. CaMKII and UNC-43 both displayed K_m values of ∼20 μM for calcium, respectively. Lysates from mock-transfected S2 cells had low levels of kinase activity, and transfection of the UNC-43 and CaMKII cDNAs into S2 cells resulted in a ∼20- to 100-fold increase in calcium-stimulated kinase activity in cell lysates.

To examine calcium-independent activity, we transfected either a wild-type UNC-43 cDNA or a cDNA containing the UNC-43(T286D) variant into S2 cells and obtained lysates. We tested the resulting UNC-43 proteins for kinase activity either in the presence or absence (using EGTA) of free calcium. For comparison, we also examined the calcium independence of rat αCaMKII and CaMKII(T286D). Rat CaMKII lysates contained kinase activity in the presence of calcium, and little activity in the absence (Fig. 4C). We measured the percentage calcium independence by comparing the ratio of the kinase activity in the absence of calcium relative to the amount of activity in the presence of calcium, and found that CaMKII from our S2 lysates showed ∼1% calcium independence (Fig. 4C). By contrast, CaMKII(T286D) lysates contained similar levels of kinase activity in both the presence and absence of calcium. Robust kinase activity was present in UNC-43 lysates in the presence of calcium, and a small amount remained in its absence (Fig. 4C). The calcium independence of UNC-43 was ∼5%, which was higher than the ∼1% value observed for CaMKII. UNC-43(T286D) lysates contained similar levels of kinase activity in both the presence and absence of calcium. We also examined the calcium-independent kinase activity of our different mutant versions of UNC-43, and found that mutations reported to increase calcium-independent activity for CaMKII also increase calcium-independent activity for UNC-43 (Fig. 4D). Taken together, our results demonstrate that UNC-43 is a calcium-dependent kinase capable of phosphorylating CaMKII substrates. Like CaMKII, UNC-43 can be made into a calcium-independent form by mutating critical residues in either the catalytic or the autoinhibitory region.

In addition to mutations that were based on the CaMKI crystal structure (Goldberg et al., 1996), there are mutations in UNC-43 that were identified in genetic screens based on the uncoordinated defects of unc-43 mutants. A single gain-of-function mutant was identified, unc-43(n498sd), which has semi-dominant phenotypes (paralyzed muscle contractions, defective egg laying and a decreased rate of defecation) that are opposite to those of the loss-of-function mutations (Reiner et al., 1999). Based on the genetic phenotype, we hypothesized that this mutation might cause constitutive and calcium-independent kinase activity. To test this, we introduced this mutation, E109K, into our UNC-43 S2 expression construct. Although we found that UNC-43(E109K) had reduced kinase activity and was less stable (data not shown), we also found that this change resulted in about 40% calcium-independent kinase activity (Fig. 4D). Interestingly, the E109K mutation, when introduced into UNC-43::GFP, showed no change in UNC-43::GFP translocation to the neurites (Fig. 3J). Our results suggest that the atonic behavioral defects observed in unc-43(n498sd) mutants are probably due to constitutive, calcium-independent activity of UNC-43.

As the K148E and D238R mutations affect the localization of UNC-43 within neurites, we reasoned that they might affect the ability of UNC-43 to regulate the trafficking of GLR-1. In unc-43 loss-of-function mutants, GLR-1 accumulates to high levels in the cell body in perinuclear structures (Rongo and Kaplan, 1999). Expression of a wild-type unc-43 cDNA in these same neurons using the glr-1 promoter is sufficient to rescue unc-43 mutants for this phenotype. To determine whether our mutations would impair UNC-43 function, we introduced the individual mutations into P_glr-1::UNC-43, the transgene we previously used to rescue unc-43. We introduced these individual transgenes into unc-43 mutants that also express GLR-1::GFP, and we examined the cell body levels of GLR-1::GFP in the transgenic animals. As previously reported, unc-43 mutants accumulated GLR-1::GFP in their cell bodies (Rongo and Kaplan, 1999) (Fig. 5A), and this defect was corrected in animals that also expressed a wild-type unc-43 cDNA (Fig. 5B). Mutants expressing UNC-43(K148E) or UNC-43(D238R) accumulated high levels of punctate GLR-1::GFP in their cell body, indicating that the mutant proteins failed to rescue these unc-43 mutants (Fig. 5C,D,O). Mutants expressing UNC-43 with either the H282K or T286D mutations had wild-type levels of GLR-1::GFP in their cell bodies, indicating that the mutant proteins rescued unc-43 (Fig. 5E,G,O). Mutants expressing UNC-43(R283E) had a small but significant decrease in GLR-1::GFP cell bodies levels compared to the wild type (Fig. 5F,O). These results demonstrate that mutations in the catalytic domain that decrease UNC-43 localization to neurites also decrease UNC-43 function with respect to GLR-1 trafficking, whereas mutations in the autoinhibitory domain that increase UNC-43 localization to neurites do not diminish UNC-43 function.

As the mutations in the catalytic domain impaired UNC-43 function and neurite localization, we also examined whether they might result in more general defects in synapse formation. We introduced P_glr-1::UNC-43 and its mutant variants into nematodes that also express a synaptobrevin-GFP reporter, SNB-1::GFP, previously used to label synapses to GLR-1-expressing interneurons (Jorgensen and Nonet, 1995; Rongo et al., 1998). For both wild-type and mutant transgenes, we observed no defects in general development, morphology, axon guidance or neurite outgrowth (data not shown). In all cases we observed normal SNB-1:GFP levels and normal numbers of SNB-1::GFP-labeled presynaptic boutons (Fig. 5H-N,P). Thus, our results suggest that our mutant transgenes do not affect more obvious aspects of these neurons.

As UNC-43 is a calcium-dependent kinase, we reasoned that the opening of a calcium channel results in the activation and subsequent translocation of UNC-43. The unc-2 gene encodes an α subunit of a voltage-dependent calcium channel and unc-2 loss-of-function mutations have a similar phenotype to unc-43 loss-of-function mutations (Rongo and Kaplan, 1999; Schafer and Kenyon, 1995). To test whether UNC-2 is also required for UNC-43 translocation, we introduced UNC-43::GFP into unc-2 loss-of-function mutants to observe its localization. Both wild-type and unc-2 mutant nematodes contained similar levels of cellular UNC-43::GFP (Fig. 6A,C,E). However, although wild-type nematodes contained abundant UNC-43::GFP in their neurites (Fig. 6B), unc-2 mutants contained low neurite levels of the kinase (Fig. 6F). Interestingly, unc-2 loss-of-function mutants also contained fewer clusters of UNC-43::GFP (Fig. 6G). Taken together, our results suggest that UNC-2 regulates two different aspects of UNC-43 neurite localization: the amount of UNC-43 that accumulates in the unlocalized neurite pool and the amount of UNC-43 that accumulates in clusters.

In an attempt to understand UNC-43 function and localization, we found that a GFP-tagged UNC-43 protein, when expressed in C. elegans neurons, is localized both in clusters and in a diffuse pool throughout neurites. We examined the effect on both kinase activity and subcellular localization of several mutations in amino acids thought to mediate interactions between the catalytic domain and the autoinhibitory domain. The majority of mutations that disrupt these interactions result in increased calcium-independent activity, and most also result in the increased amount of UNC-43 present in the neurites. To our surprise, we identified two residues in the catalytic core (K148 and D238) that do not impair kinase activity but nevertheless are important for UNC-43 localization and function in neurites. Moreover, we have shown that the activity of UNC-2, a voltage-dependent calcium channel, is also required to localize UNC-43 to neurites.

Vertebrate CaMKII has previously been shown to be localized to dendrites and enriched at the PSD (Ouimet et al., 1984; Petersen et al., 2003; Strack et al., 1997). In addition, CaMKII can form non-synaptic clusters, with increasing amounts of these clusters forming in response to ischemic stress (Tao-Cheng et al., 2002). We find that UNC-43 forms clusters both in neurites and in the neuron cell bodies. Interestingly, the clusters found in the cell bodies colocalize with clusters of GLR-1 found at the Golgi. UNC-43 activity is required for the transport of GLR-1 from the cell body to the synaptic connections of the neurites, and UNC-43 kinase activity is required for this function (Rongo and Kaplan, 1999). Thus, the colocalization of UNC-43 with GLR-1 at or near the Golgi would suggest that UNC-43 is phosphorylating target molecules that participate in the secretory trafficking of GLR-1, perhaps at these initial regions of protein translocation. UNC-43 is also found to colocalize with GLR-1 along neurites, although a large fraction of UNC-43 clusters do not colocalize. It is possible that this latter population of UNC-43 clusters are at non-glutamatergic synaptic inputs; alternatively, they might be non-synaptic accumulations of UNC-43. There is also an abundant, unlocalized pool of UNC-43 present throughout the neurites. This unlocalized pool of UNC-43 could be functionally important for synapses even though it is not directly at the PSD. For example, the insertion of new glutamate receptors in response to activity is thought to occur by mobilizing receptors from a subsynaptic pool present beneath the PSD and along the length of the dendrites (for reviews, see Malenka, 2003; Malinow, 2003; Sheng and Hyoung Lee, 2003; Song and Huganir, 2002). Recently, the endocytosis of glutamate receptors has been shown to occur at specialized regions outside of the PSD (Racz et al., 2004). Thus, it is reasonable that UNC-43 could be regulating PSD composition and/or function with its presence in an unlocalized pool in the neurites. Consistent with this idea, two mutations in unc-43 (K148E and D238R) that diminish UNC-43 levels in the unlocalized neurite pool also prevent UNC-43 from regulating GLR-1 localization.

Our findings also demonstrate that the UNC-2 voltage-dependent calcium channel is required for UNC-43 to accumulate in neurites, thereby providing a means for neural activity to regulate the localization of UNC-43. Activity has been shown to have an important role in CaMKII translocation in neurons (Leonard et al., 1999; Ouyang et al., 1997; Shen and Meyer, 1999; Shen et al., 2000; Strack et al., 1997). For example, activated NMDA receptors can mobilize αCaMKII within neurons to dendrites (Bayer et al., 2001; Shen and Meyer, 1999; Shen et al., 2000). In addition, activated CaMKII can become trapped at PSDs, and this form of localization trapping could mark active synapses and place CaMKII near a source of calcium influx (Shen et al., 2000). In unc-43 gain-of-function mutants like T286D and R283E that mimic increased neuronal activity, we observed an increase of UNC-43 in neurites, but did not observe an increase at PSDs. In unc-2 mutants, where neuronal activity is decreased, we observed both a decrease of UNC-43 in neurites and a decrease of UNC-43 at PSDs. We speculate that activity regulates CaMKII localization through two steps, neurite translocation and PSD trapping, and that at least the former step has been conserved evolutionarily in C. elegans interneurons.

Our mutational analysis of UNC-43 suggests a possible mechanism by which the translocation of the kinase might occur in response to calcium (Fig. 7). Calcium/calmodulin binding to the autoinhibitory domain can cause the domain to become displaced from the hydrophobic pocket on the catalytic domain to which it binds. Transport factors that are similar in sequence to the autoinhibitory domain could then insert themselves along this same hydrophobic pocket, thus physically tethering the kinase to such factors as they make their way out to the neurites. In such a scenario, the autoinhibitory domain would be competing against these hypothetical transport factors for access to the catalytic domain; hence, mutations that impair the ability of the autoinhibitory domain to interact with the catalytic domain should result in increased amounts of UNC-43 in the neurites. This was indeed the case for most of the mutations we examined, including a deletion mutant that results in the complete removal of the autoinhibitory domain. Only mutations in key residues of the catalytic domain that not only interacted with the autoinhibitory domain but also the transport factor would be expected to have the opposite phenotype, thus decreasing the amount of UNC-43 in the neurites. We found two such mutations, D238 and K148, suggesting that these residues have an important role in the translocation of UNC-43 to the neurites.

What is the nature of the transport factor? One protein thought to bind to αCaMKII by mimicking the pseudosubstrate domain is the NR2B subunit of the NMDA receptor, which contains a short sequence in its C-terminal tail that matches the CaMKII consensus site (Bayer et al., 2001). The C. elegans ortholog of NR2 lacks this same sequence, however, and NMDA receptor knockout mutants have little effect on UNC-43 localization (Brockie et al., 2001) (data not shown). The UNC-2 α1 calcium channel subunit is required for UNC-43 localization, but also lacks the CaMKII consensus site (Mathews et al., 2003). CaMKII probably associates with other proteins for its trafficking, in addition to NR2B. Thus, it is reasonable to expect that some of these proteins will be common interactors for both CaMKII and UNC-43.

We thank A. Fire, A. Fraser (Sanger Institute), R. Herman (CGC), J. Kaplan, M. Kennedy, Y. Kohara, V. Maricq, J. Mendel, C. Mello, M. Nonet, H. Schulman, T. Stiernagle (CGC) and R. Tsien for reagents and strains. We thank Bonnie Firestein for critical comments on the manuscript. C.R. is a Pew Scholar in the Biomedical Sciences. Additional funding was provided by the N.I.H. (R01 NS42023) and a Grant-In-Aid from the American Heart Association.