unc-69 was identified in a large-scale behavioral screen for uncoordinated (Unc) mutants 18. unc-69 loss-of-function (lf) mutants move poorly, coil ventrally and are phenotypically similar to other coiler Unc mutants, many of which are defective in axonal outgrowth and guidance. Additionally, unc-69 mutant hermaphrodites lay more eggs in the absence of food than wild-type worms do (see Additional data file 1), suggesting a defect in the hermaphrodite-specific neurons (HSNs), which control egg-laying behavior.

Previous genetic data placed unc-69 between lin-12 and tra-1 on chromosome III, 0.12 map units to the left of ced-9 19. Using cosmid rescue, we were able to identify the predicted gene T07A5.6a (previously named T07C4.10b) as unc-69 (Figure 1a). The unc-69 gene encodes a 108-amino-acid protein and contains a short coiled-coil domain in its carboxyl terminus (Figure 1b). Although UNC-69 could possibly form a homodimer via its coiled-coil domain, we failed to detect any homophilic interactions of UNC-69 (see Additional data file 1).

The original alleles of unc-69 , unc-69(e587) and unc-69(e602) , are both nonsense mutations in the carboxy-terminal half of the protein (see Figure 1b). The unc-69(e602) mutation causes a T-to-A transversion and replaces a leucine with an amber stop codon at position 77; unc-69(e587) results in a C-to-T transition, changing a glutamine to an amber stop codon at position 86; both of these mutations lie within the well conserved coiled-coil domain. Both unc-69(e602) and unc-69(e587) are candidate genetic null alleles, as the axon extension and branching defects of the neurons named ALM and AVM were not enhanced significantly when either of these two alleles was placed in trans to the deficiency nDf40 (Table 1, Figure 2).

We also isolated a hypomorphic allele, ju69, which results in a G-to-A transition at the start codon and changes the initiator methionine to an isoleucine. Theoretically, the M-to-I substitution (M1I) should abolish translation initiation and hence synthesis of the UNC-69 protein. As the phenotype of unc-69(ju69) mutants is much weaker than that of the other two alleles, however, we suspect that a small amount of UNC-69 functional protein is still being produced, either by leaky translation initiation at the original site, or through initiation at the internal, in-frame ATG site at residue 49, which would leave the coiled-coil domain intact. Indeed, overexpression of a mutant fusion protein of UNC-69 with green fluorescent protein (UNC-69(M1I)::GFP) or a carboxy-terminal fragment of UNC-69 (residues 41–108) could partially suppress the locomotion defect of the unc-69(e587) mutants (data not shown, and see Additional data file 1).

Finally, we analyzed a small deletion, ok339, which completely eliminates the unc-69 locus. Unfortunately, this deletion also removes the essential neighboring gene T07A5.5 and was therefore not studied further (see Additional data file 1). Expressed sequence tag (EST) analysis suggested that the unc-69 locus encodes two splice variants (see Figure 1a and see Additional data file 1). Northern blot analysis of poly(A)⁺ RNA from mixed-stage worms as well as from embryos revealed a 0.65 kb major transcript (Figure 1c), consistent with the predicted size of the T07A5.6a transcript.

We found that UNC-69 is highly conserved through evolution and encodes the C. elegans homolog of mammalian SCOCO (short coiled-coil protein), a protein recently found to interact with dominant-negative ARF-like 1 (ARL1) protein in a yeast two-hybrid screen 20. The Saccharomyces cerevisiae UNC-69 homolog, Slo1p (SCOCO-like open reading frame protein), has been shown to interact with Arl3p, a homolog of mammalian ARFRP1, another ARF-like protein, which is involved in endoplasmic reticulum-Golgi and post-Golgi transport 2122. Uncharacterized UNC-69/SCOCO homologs can also be found in many other animal species (Figure 3a and Additional data file 1).

All of the UNC-69 homologs are predicted to form a coiled-coil structure near their carboxyl termini (the underlined region in Figure 3a). In an alignment of the S. cerevisiae, C. elegans, C. briggsae, mosquito, fly, Fugu, zebrafish, Xenopus, mouse and human protein sequences, identity over the coiled-coil regions is 32.6% (Figure 3a). The identity in the coiled-coil region jumps to 73.9% if the yeast sequence is excluded. Except for yeast, an acidic region immediately upstream of the coiled-coil domain as well as a serine/threonine-rich region and a basic region downstream appear also to be highly conserved. In contrast, the amino terminus of UNC-69 and its homologs is highly divergent, both in length and in amino-acid sequence. The function of UNC-69 proteins seems to be conserved, since expression of human SCOCO as a transgene under the unc-69 promoter restored locomotion to unc-69 mutants (Figure 3c).

We assessed the tissue distribution of human SCOCO transcripts by probing a human fetal tissue northern blot. This probe detected a single transcript of approximately 2.1 kb in all tissues examined (brain, lung, liver and kidney; Figure 3b). Human SCOCO mRNA appeared to be enriched in fetal brain, possibly hinting at a role for SCOCO in mammalian nervous system development.

We generated transgenic animals expressing either amino- or carboxy-terminally gfp-tagged unc-69 fusion constructs under the control of the endogenous unc-69 promoter. Both translational fusion constructs rescued the Unc phenotype of unc-69 mutants, suggesting that the fusion proteins were correctly expressed and biologically functional. UNC-69::GFP expression was first detectable in embryos (Figure 4a,b). In immature neurons, we observed expression of UNC-69::GFP in the processes and growth cones of developing neurites (arrowhead in Figure 4c). In older larvae and adults, UNC-69::GFP was expressed in neurons of the anterior, lateral, ventral and retro-vesicular ganglia in the head, and in neurons of the preanal, dorso-rectal and lumbar ganglia in the tail. The fusion protein was also present in the ventral nerve cord (VNC), in the dorsal nerve cord (DNC), in the dorsal and ventral sublateral nerve cords, and in commissural axons (Figure 4d–f). The reporter was expressed in the neurons named CAN, HSN, ALM, PLM, AVM, PVM, BDU, and SDQR, as evidenced by its localization to the cell bodies of these neurons. Expression of unc-69 in these latter cells was confirmed using an unc-69::LacZ::NLS fusion (data not shown). Taken together, these results indicate that unc-69 is expressed widely, perhaps ubiquitously, in the C. elegans nervous system.

Expression of UNC-69::GFP was also observed in non-neuronal cells. In larvae and adults, we occasionally observed UNC-69::GFP expression in body-wall muscle (data not shown). We also observed UNC-69::GFP in the excretory canal, in the distal tip cells, in the spermatheca and, less frequently, in hypoderm and gut (Figure 4e, and data not shown). The expression in these non-neuronal cells was variable, however, and might not reflect the endogenous expression pattern of unc-69 .

The ventral coiler phenotype of unc-69 mutants suggests a defect in nervous system development. Indeed, previous studies had reported axonal guidance defects of the D-type GABAergic motor neurons, mechanosensory neurons and the HSN neurons in unc-69 mutants 2324. We confirmed these observations and extended them to other cell types (see Tables 1, 2 and Figures 2, 5a–f). Incorrect targeting of the DD and VD motor axons is likely to contribute to the Unc phenotype of unc-69 mutants. In addition to outgrowth and guidance defects, we also observed ectopic branching of the DD/VD neurons and mechanosensory neurons in unc-69 mutants (Figure 5d,f). In a few cases the axons had unusual large swellings and occasionally meandered along the lateral body wall.

FMRF-amide (Phe-Met-Arg-Phe-NH₂) is a neuropeptide that serves as a neuromodulator, and is co-released together with other neurotransmitters. In examining other neuronal classes in unc-69(e587) mutants, we observed premature termination of axons of the FMRF-amide-positive neurons ALA, RID and AVKR, but not RMG (data not shown, and see Table 2). FMRF-amide-positive neurons are so-called neuropeptidergic neurons and could be sensory, motor or interneurons. We observed that 67% (20/30) of ALA axons terminated prematurely, and ALA axons sometimes branched before termination. AVKR had frequent axonal outgrowth and guidance defects: 85% (17/20) of AVKR axons terminated prematurely or crossed from the left VNC (VNCL) to the right VNC (VNCR). Taken together, these observations support a role for unc-69 in ventral and dorsal axonal guidance as well as in axonal elongation within the fascicles.

As unc-69 mutants have midline crossover defects (see Table 2), it is likely that axons running in the same fascicle lose cell-cell adhesion and fail to stay together. We constructed a series of electron micrograph (EM) cross-sections through the major nerve cords (DNC, VNCL and VNCR) that run antero-posteriorly in adult hermaphrodites. In wild-type animals, the composition of axons in any of these nerve cords is highly stereotyped, with four axons fasciculated to run in VNCL and the other ventral axons running within VNCR (Figure 5g) 25. In unc-69(e587) and unc-69(e602) mutants, many fascicles split into two or more groups and in some cases defasciculated axons could be seen running alone along the hypodermal ridge. Moreover, some axons of both the DNC and VNCL appeared to be mislocalized and can be seen on the wrong side of the hypodermal ridge (Figure 5h and data not shown). Anti-tubulin and anti-GABA staining confirmed the observed fasciculation defects in unc-69(e587) mutants (data not shown).

To determine whether unc-69 expression is required in the growing neurites or in the surrounding tissues, we created unc-69 transgenic lines expressing unc-69(+) specifically in the six touch neurons under the control of a mec-7 promoter. We compared outgrowth and guidance defects of the ALM and AVM neurons in three such lines with those of unc-69(lf) mutants (see Table 1, Figure 2). In all three transgenic lines, the percentage of ALM neurites that failed to extend to full length or send a branch into the nerve ring dropped significantly. Similar observations were made for AVM outgrowth and branching. Note that none of the transgenic lines completely rescued the ALM outgrowth and branching defects. This could be due to loss or silencing of the transgene carried on the extrachromosomal array or could reflect a requirement for unc-69 in other neuronal and/or non-neuronal cells. Nevertheless, we conclude that UNC-69 promotes outgrowth and guidance largely, if not completely, in a cell-autonomous manner.

The C. elegans synaptobrevin/vesicle-associated membrane protein (VAMP) homolog SNB-1 is a vesicular soluble N-ethylmaleimide-sensitive factor attachment protein receptor (v-SNARE) on synaptic vesicles (SVs). Tagged SNB-1 can be used to follow SVs as they are transported to presynaptic regions 26. We isolated an allele of unc-69, ju69, in a visual genetic screen for mislocalization of a SNB-1::GFP reporter in D-type GABAergic motor neurons. In wild-type worms, SNB-1::GFP expressed in the D neurons can be localized to discrete puncta along the VNC and DNC, at sites of neuromuscular junctions (Figure 6a,c). In unc-69(ju69) mutant nerve cords, SNB-1::GFP puncta were irregular in size and position, on average larger than in wild type, and often completely missing for extended stretches (Figure 6b,d,e). In addition, we occasionally observed puncta that abnormally diffused from the nerve cords into the commissures (Figure 6d). Despite the abnormal shape and distribution of presynaptic regions, the overall morphology of DD and VD neurons was grossly normal (Figure 6f–i) and only occasionally (<10%; n = 50) did one commissure fail to exit the VNC. We made similar observations in touch neurons using worms carrying the P_mec-4::gfptransgene zdIs5 (data not shown), a strain chosen for reconfirming findings made on D-type GABAergic motor neurons.

Much more dramatic SNB-1::GFP distribution defects were observed in the strong mutant unc-69(e587) (data not shown). Because of the extensive pathfinding defects observed in strong unc-69 mutants, however, which might complicate interpretation of the SNB-1::GFP distribution defect, we restricted our subsequent analysis to the unc-69(ju69) background, in which axonal guidance is largely normal. Indeed, although unc-69(ju69) mutant worms are Unc, they move much better than strong unc-69 mutants. Thus, the locomotion defect observed in unc-69(ju69) mutants is probably a consequence of a defect in transport or localization of axonal cargos rather than in axon guidance.

To determine whether the outgrowth defects we observed in unc-69 mutants are specific to axons, we examined the morphology of the AWC class of sensory neurons using the kyIs140 [P_str-2::gfp] transcriptional reporter, which is normally stochastically activated in either the right or left AWC neuron 27. The bilaterally symmetric AWC neurons have a distinct bipolar structure, with a dendrite extending to the tip of the nose and an axon extending into the nerve ring (Figure 5i). In unc-69(e587) mutants, the axon of the AWC neuron often stopped prematurely (Figure 5j), and str-2::gfp expression was often silenced (see below). In contrast, the dendrite of the AWC neuron had no outgrowth defect, as 100% (136/136) of the AWC dendrites extended to their full length. In unc-69(e587) mutants, 73% (99/136) of AWC neurons had ectopic bulges or branches protruding from either the cell body or the axon (similar to what we observed in the mechanosensory neurons, Figure 5f,j). Ectopic branches only rarely extended from dendrites, however (data not shown). Dendritic morphology was also normal in the ASI neurons (visualized by the str-3::gfp transgene), the AWB, AWC, ASG, ASI, ASK, and ASJ neurons (visualized by the tax-Δ::gfp transgene) 2829, and the sensory neurons ASJ, ASH, ASI, ASK, ADL, and ADF (visualized by staining with the lipophilic dye DiI; data not shown). Finally, an odorant receptor was still properly localized to the cilia (see below). From these observations, we conclude that UNC-69 is probably not required for either cilia formation or dendritic elongation within the amphid sensilla, a sensory organ within the head of a worm.

In vesicle-trafficking mutants such as unc-16 and unc-116, markers for synaptic vesicles are also mis-sorted into dendrites 7. We wondered whether unc-69 mutants also show such a general sorting defect, or whether unc-69 might be required more specifically for efficient trafficking within the axons. At the L1 larval stage, the thirteen VD neurons are not yet born, and the six DD neurons are the only D-type GABAergic motor neurons present in the VNC. At this stage, the DD neurons receive their synaptic inputs from the DNC and output onto the ventral body-wall muscles. In wild-type L1s, therefore, the SNB-1::GFP puncta can be seen only along the VNC. In unc-69(ju69) mutants, the synaptic GFP was not significantly mislocalized to the DNC (3.4%; n = 59; Figure 6k). In contrast, SNB-1::GFP puncta were frequently seen in the DNC in unc-16(ju146) mutant L1s (90.6%; n = 32; Figure 6k). We also made similar observations in worms carrying a snb-1::gfp transgene expressed in a pair of ASI sensory neurons, in which SNB-1::GFP was not significantly mislocalized to the ASI dendrites in unc-69(ju69) mutants (C-W.S., Y.J. and M.O.H., unpublished data).

We next asked whether UNC-69 has any role in transporting proteins within the dendrites. We used an odr-10::gfp transgene that is expressed in the AWB neurons to answer this question 30. ODR-10 is an odorant receptor for diacetyl, and is actively transported in vesicles from the cell bodies to the cilia at the end of the dendrites, where the GFP fusion is deposited (Figure 6l). In dendritic targeting mutants, such as unc-101 (which encodes the homolog of AP1 μ1 clathrin adaptor protein), ODR-10::GFP is not targeted to the AWB cilia 30 (Figure 6n); in contrast, in both unc-69(ju69) and unc-69(e587) mutants, ODR-10::GFP was still properly targeted (Figure 6m; data not shown). Taken together, our results suggest that dendritic development and transport of proteins into dendrites is not impaired in unc-69 mutants. Thus, UNC-69 is possibly specifically required for axonal transport and outgrowth.

To identify potential UNC-69 interactors, we screened three C. elegans yeast two-hybrid libraries using full-length UNC-69 as bait. From these screens, we isolated at least 34 independent clones of UNC-76, a 385-amino-acid protein that was previously shown to be involved in axonal outgrowth and fasciculation in C. elegans 121314. The Drosophila homolog of UNC-76 was identified as a KHC-binding protein and shown to be a regulator of axonal transport 15. A mammalian homolog of UNC-76, FEZ1, is a substrate for PKCζ 16. Worm, fly and mammalian UNC-76 proteins are not only conserved in amino-acid sequence but also have several conserved regions (Figure 7d) predicted to be capable of forming coiled-coil domains 1415. UNC-76 localizes to axons, and worms harboring mutations in unc-76 have a severe Unc phenotype and coil ventrally, phenotypes very similar to those observed in unc-69 mutants 14.

We used an in vitro glutathione S-transferase (GST) pull-down assay to verify the physical interaction between UNC-69 and UNC-76. As shown in Figure 7a, in vitro translated full-length UNC-76 (UNC-76FL) was pulled down efficiently by GST-UNC-69 but only minimally by GST-CBP, a eukaryotic transcription factor used as a negative control 31. Conversely, in vitro translated adenoviral protein E1A efficiently bound to its cognate partner GST-CBP but not to GST-UNC-69. Therefore, the interaction between UNC-76 and UNC-69 is specific and most likely direct.

To narrow down the regions of interaction, we generated truncated proteins lacking various parts of UNC-76 (Figure 7b,d) and tested for their interaction with GST-UNC-69. We found that amino acids 281 to 299 of UNC-76 were necessary to interact with UNC-69 in vitro . Interestingly, this 19-amino-acid region overlaps with a region predicted to form a coiled-coil structure (amino acids 265–292; purple region in Figure 7d) and lies within a region conserved from worms to humans (gray-shaded region in Figure 7d).

To corroborate the in vitro interactions with the in vivo function of UNC-76, we expressed truncated UNC-76 proteins tagged with yellow or cyan fluorescent protein (YFP or CFP) in unc-76(e911) mutant worms (Figure 7d) and assayed for rescue of the Unc phenotype. Both amino-terminally and carboxy-terminally tagged full-length UNC-76::YFP or CFP::UNC-76 fusion proteins were functional and rescued unc-76(e911) mutants (Figure 7d). The CFP::UNC-76 Δα fusion protein (which lacked the amino terminus of UNC-76) failed to rescue unc-76(e911) mutants, suggesting that the amino-terminal region of UNC-76 is required for its function in vivo. Bloom and Horvitz reported that amino acids 1–197 of UNC-76 are sufficient to direct proteins into the axons in C. elegans 14. As the axonal targeting sequence of UNC-76 includes the region deleted in UNC-76 Δα, we speculated that CFP::UNC-76 Δα fusion proteins were not transported to axons. Indeed, the CFP signal was weak and seemed to congregate more around the soma (data not shown). In contrast, the CFP::UNC-76 Δγ fusion protein was both strongly expressed in soma and axons, but failed to rescue unc-76(e911) mutants, consistent with the hypothesis that binding to UNC-69 is critical for UNC-76 to function in vivo.

If coiled-coil structures are important for the UNC-76-UNC-69 interaction, any mutation that abolishes the coiled-coil structure would possibly also abolish physical interaction between the two proteins. To test this idea, we mutagenized four conserved residues in UNC-76: Glu275, Leu281, Leu287, and Lys291. Both UNC-76(E275A) and UNC-76(K291A) mutant proteins still bound UNC-69 in vitro (Figure 7d). Likewise, YFP fusions of these mutant proteins rescued unc-76(e911) mutants. In contrast, both UNC-76(L281P) and UNC-76(L287P) mutant proteins failed to bind UNC-69 in vitro. Surprisingly, UNC-76(L287P) was still able to rescue unc-76(e911) in vivo (Figure 7c,d; we did not test UNC-76(L281P) for rescue). These data suggest that a single-amino-acid substitution might not be potent enough to destroy the coiled-coil structure when UNC-76 protein is folded in its native state. Finally, we created a mutant protein carrying both L281P and L287P mutations (P2), as well as an internal deletion mutant, Δ19, which deletes amino acids 281–299 of UNC-76. Both P2 and Δ19 mutants largely failed to rescue unc-76(e911) in vivo (Figure 7d; occasionally, mutant hermaphrodites carrying the unc-76 P2::yfp or the unc-76 Δ19::yfp transgenes were slightly rescued as young adults). In summary, amino acids 281–299 of UNC-76 probably contain or overlap with an UNC-69-binding site, and UNC-76 may require interaction with UNC-69 to function in vivo.

As both UNC-69 and UNC-76 are required for axon outgrowth and fasciculation, we asked whether they function in the same genetic pathway to regulate axon extension. We first tested whether overexpression of UNC-69 in unc-76(lf) mutants could bypass the unc-76 mutant phenotype. We overexpressed a functional unc-69::gfp transgene as an extrachromosomal array in unc-76(e911) mutants but did not see any rescue in locomotion (three independent lines, data not shown). Likewise, overexpression of a functional unc-76::yfp transgene failed to rescue the locomotion defect of unc-69(e587) mutants (data not shown).

We also performed a double-mutant analysis to further address the question of whether unc-69 and unc-76 act in the same pathway. In C. elegans, expression of the odorant receptor gene str-2 is randomly turned on in either the left or the right AWC sensory neuron (AWCL/R), but never in both 27. In wild-type worms, this '1 AWC^ON' phenotype is determined by axonal contact and calcium signaling between AWCL and AWCR. In axonal guidance mutants such as unc-76, sax-3 and vab-3, the two AWC axons often fail to meet, and P_str-2::gfp expression is consequently silenced in both AWCs, giving rise to a '2 AWC^OFF' phenotype 27. We used this system to quantitatively score axon extension defects in the nerve ring in different unc-69(lf) and unc-76(lf) mutants as well as in unc-69(lf); unc-76(lf) double mutants.

In both strong loss-of-function mutants, unc-69(e602) and unc-69(e587), 30–34% of animals showed a 2 AWC^OFF phenotype. In contrast, the hypomorphic allele unc-69(ju69) resulted in only 1% of mutant worms (n = 190) having P_str-2::gfp expression silenced in both AWCs (Table 3). This result was consistent with our previous observation that neuronal morphology is largely normal in unc-69(ju69) mutants. In agreement with previous studies 27, 47% of unc-76(e911) mutants (n = 101) had the 2 AWC^OFF phenotype; e911 was the strongest allele among all the nine alleles that we tested. For the other unc-76 alleles, the 2 AWC^OFF phenotype varied from 6% to 30%. Interestingly, the strength of the AWC expression defect (which is an indication of axon extension defects) showed an inverse colinear relationship with the position of each mutation in the open reading frame: the most 5' mutation, unc-76(n2457), showed the least defect in axon extension, whereas alleles located most carboxy-terminally showed greater defects than alleles located close to the amino terminus (Table 3). Interestingly, we did not observe enhancement of axon extension defects in unc-69; unc-76 double mutants: in all cases, the defect in the double mutant was no stronger than in the stronger of the single mutants (Table 3). In contrast, axon extension defects were greatly enhanced in unc-76(e911); sax-3(ky123), unc-76(e911); unc-6(n102) and unc-33(e204); unc-76(e911), and slightly enhanced in unc-76(e911); vab-3(e648) and unc-119(ed3); unc-76(e911) double mutants (Table 3). Because unc-76 alleles failed to show any additivity with the candidate null alleles unc-69(e587) and unc-69(e602), we conclude that UNC-69 and UNC-76 probably act in the same pathway to control axon extension, at least in the case of the AWC sensory neurons.

We showed above that UNC-69 is required for localization of synaptic vesicles in axons. Does UNC-76 also have a role in this process, and if so, does UNC-76 control presynaptic organization together with UNC-69? Unfortunately, all existing unc-76 alleles have severe axonal outgrowth defects, making interpretations of defect in synaptic vesicle localization difficult. To bypass this problem and to reveal possible genetic interactions between unc-69 and unc-76, we looked at the localization of the synaptobrevin SNB-1::GFP puncta in unc-69(lf)/+; unc-76(lf)/+ double heterozygotes (Figure 8). In wild-type adult hermaphrodites, SNB-1::GFP can be seen as evenly distributed puncta along the DNC 7 (Figure 8a,e). The distribution pattern of GFP puncta in DNC was not significantly different in unc-69(e587)/+ heterozygotes (Figure 8b) as compared with wild-type animals. However, in both unc-69(e587)/+; unc-76(e911)/+ and unc-69(e587)/+; unc-76(n2457)/+ double heterozygous hermaphrodites, SNB-1::GFP puncta were occasionally more diffused, larger, or completely absent within a stretch of DNC (Figure 8c,d,f); the absence of SNB-1::GFP puncta may be due to either transport or axon extension defects. In addition, unc-69(e587)/+; unc-76(e911)/+ and unc-69(e587)/+; unc-76(n2457)/+ double heterozygotes occasionally had a slight Unc phenotype in locomotion, resembling weak synaptic transmission mutants. The weak locomotion defect could be a direct or indirect effect of the synaptic vesicle mislocalization defect.

In summary, the unc-69/+; unc-76/+ double heterozygotes show phenotypes that are similar, albeit significantly weaker, to those observed in unc-69(ju69) homozygotes. Haplo-insufficient genetic interactions of this type, commonly known as nonallelic (or unlinked) noncomplementation, are often observed with proteins that form heterodimers or function in a common protein complex (such as α- and β-tubulin; 32). Several other explanations are also possible, however (discussed in 33). Thus, our observations are compatible with, but do not definitively prove, the hypothesis that UNC-69 and UNC-76 act in a common pathway required for proper synaptic-vesicle localization.

To determine the subcellular localization of UNC-69 and UNC-76, we coinjected P_unc-69::cfp::unc-69 and P_unc-76::unc-76::yfp constructs at low concentration (5 ng/μl) into unc-76(e911) mutant hermaphrodites, and selected rescued transgenic animals for examination. At low concentration, both CFP::UNC-69 and UNC-76::YFP often appeared as puncta along the DNC, in CAN neurons, as well as in other neuronal processes that run along the subdorsal and sub-ventral tracts (Figure 9a–f). Less frequently, these puncta could also be found in commissures that connect the DNC to the VNC. The punctate pattern of UNC-76 can also be observed when worms are stained with anti-UNC-76 antisera 14, consistent with this being the endogenous expression pattern of UNC-76. Both CFP::UNC-69 and UNC-76::YFP puncta were of variable size but were usually large and immobile, even in the commissures. Interestingly, CFP::UNC-69 and UNC-76::YFP proteins also colocalized in round, perinuclear dots in the soma (Figure 9j–l). These observations strengthen our belief that UNC-69 and UNC-76 coexist in a protein complex. The molecular nature of the observed UNC-69-UNC-76 puncta (multiprotein complexes or vesicles, perhaps) remains to be determined.

In Drosophila, Unc-76 associates and copurifies with KHC, which is the major component of the conventional kinesin motor Kinesin-1 required for axonal transport towards the plus ends of microtubules 15. A similar biochemical interaction between UNC-76 and the C. elegans KHC ortholog UNC-116 34 has not been reported so far. To determine whether the UNC-69-UNC-76 complex is transported to axons by UNC-116, or by another kinesin, the KIF1A homolog UNC-104 35, we compared the subcellular localization of both CFP::UNC-69 and UNC-76::YFP in wild-type and in different kinesin mutant backgrounds.

In unc-116(rh24) mutants, UNC-76::YFP puncta were occasionally diffuse and sometimes failed to be accompanied by CFP::UNC-69 puncta in a stretch of axon (Figure 9g–i). In addition, both CFP::UNC-69 and UNC-76::YFP proteins often occupied distinct but partially overlapping perinuclear territories in the soma in unc-116(rh24) mutants (Figure 9m–o). Whereas perinuclear CFP::UNC-69 dots increased in size in unc-116(rh24) mutants, perinuclear UNC-76::YFP either split into several smaller dots (as in Figure 9n(i)) or formed an irregular reticular structure (as in Figure 9n(iii)) in unc-116(rh24) mutants. The unc-116(rh24) mutants carry two missense mutations (I304M and E338K) at the end of the motor domain of KHC (amino acids 1–358) 34. Thus, these mutations are likely to affect the processivity of KHC and cargo transport along the microtubules.

We also generated functional integrated UNC-69::GFP and UNC-76::YFP transgenes that were stably overexpressed in the nervous system and studied their subcellular localization in different kinesin mutant backgrounds. The CAN neurons are a pair of bilaterally symmetric neurons that send processes antero-posteriorly along the excretory canal (Figure 9p) 36. In wild-type animals, UNC-69::GFP and UNC-76::YFP could be observed both in the CAN soma and throughout the processes (Figure 9p,s). In worms mutant for unc-104(e1265), the C. elegans KIF1A homolog 35, subcellular distribution of UNC-69::GFP and UNC-76::YFP was not significantly altered (Figure 9q,t). In unc-116(rh24) mutants, overexpression pattern of UNC-69::GFP and UNC-76::YFP were both significantly different from wild-type animals. The CAN neuron accumulated UNC-69::GFP in the vicinity of its cell body, which was swollen and deformed. In addition, there were ectopic branches near the cell body, and UNC-69::GFP also accumulated in these processes (Figure 9r). Unlike UNC-69::GFP, UNC-76::YFP appeared as giant dots along the CAN processes in unc-116(rh24) mutant, as if UNC-76::YFP was removed from the cytoplasm and concentrated in certain subcellular compartments (Figure 9u). Moreover, a CEH-23::UNC-76_1–197 ::GFP fusion protein 37 also appeared as large aggregates along CAN processes in unc-116(rh24) mutants (Figure 9w).

In summary, our data show that the subcellular distribution of both UNC-69 and UNC-76 is altered in unc-116(rh24) mutants. It is striking that the nearly perfect co-localization of UNC-69 and UNC-76 is disrupted in unc-116 mutants. We are still at a loss to explain the molecular basis of this unexpected finding. What is clear, however, is that axonal transport of UNC-69 and UNC-76 is still occurring in unc-116(rh24) mutants. Thus, other kinesin motors and/or additional factors probably contribute to transport of UNC-69 and UNC-76 along the axons.

UNC-69 homologs in S. cerevisiae and mammals have been reported to interact physically with members of the family of ARF-like small GTPases. To investigate whether a similar interaction occurs in C. elegans, we first used yeast two-hybrid assays to study protein-protein interactions between UNC-69 and three closely related but distinct ARF-like small GTPases, ARL-1 (F54C9.10), ARL-3 (F19H8.3), and ARFRP (Y54E10BR.2) 38. Whereas UNC-69 readily interacted with the carboxyl terminus of UNC-76 (UNC-76γ), it did not interact with any of the three ARF-like proteins (Figure 10a). As human SCOCO was isolated as an effector for GTP-bound ARL1 20, we also tested the ability of UNC-69 to interact with GTPase-defective forms of ARL-1 and ARFRP. UNC-69 did not interact with either ARL-1(Q70L) or ARFRP(Q79L) (Figure 10a). Deletion of the amino-terminal myristoylation site 39 also had no effect: UNC-69 did not interact with the amino-terminal deletion ARL mutants, ARL-1Δ16 (with or without the GTPase-defective mutation) or or ARL-3Δ17 (data not shown). In contrast, we readily detected the previously reported interaction between ARL-3 and UNC-119 40, a homolog of human retinal gene 4 (HRG4) 414243 (Figure 10b). Thus, the failure to detect any interaction between UNC-69 and the three ARF-like proteins might not have been due to inappropriate protein folding or subcellular compartmentalization in yeast.

To address directly the question of whether UNC-69 is a Golgi-associating protein, we coexpressed CFP::UNC-69 and a YFP-tagged fragment of the C. elegans Golgi protein mannosidase II F58H1.1 (mansII::YFP) 44. Unlike the colocalization pattern we observed previously for UNC-69 and UNC-76, UNC-69 and mansII only occasionally colocalized (Figure 11). Moreover, we clearly observed regions in which both UNC-69 and mansII occupied non-overlapping subcellular territories, even under overexpression conditions (arrows in Figure 11c,f,i). This mutual exclusion could not simply be explained by the squeezing out of UNC-69 from the mansII-containing territories as a result of spatial constraint, as UNC-69 and mansII territories did sometimes overlap (arrowheads in Figure 11c,i