Genetic screens in C. elegans identified many unc-genes, which lead to impaired (‘uncoordinated’) locomotion (Brenner, 1974). Neuronal polarity defects and disorganized neuronal microtubules have been observed in several unc-mutants (Maniar et al., 2012; Yan et al., 2013). In this study, we analyzed neuronal development and axon-dendrite polarity in the highly polarized PVD neuron (Figure 1A), which is widely used as a model neuron (Sundararajan et al., 2019). We found that the unc-119 mutant ed3, which carries an early stop mutation, has a fully penetrant microtubule orientation defect in the PVD anterior dendrite (Figure 1B and C). Axons were occasionally affected as well (Figure 1—figure supplement 1A). The microtubule orientation defect is not specific to the PVD neuron, as we observed a similar phenotype in the PHC neuron (Figure 1—figure supplement 1B). In contrast, the ciliated URX and PQR neurons did not have obvious microtubule defects (Figure 1—figure supplement 1C and D), as is expected because these neurons have additional mechanisms to organize neuronal microtubules (Harterink et al., 2018). Furthermore, synaptic vesicles were consistently mislocalized to the cell body and dendrites in the unc-119 mutant (Figure 1D and Figure 1—figure supplement 1E–I), and we observed morphological defects for the PVD neuron; the axon and anterior dendrite are shorter and dendrite branching is reduced (Figure 1E and Figure 1—figure supplement 1J–L). The disorganized microtubules and vesicle mislocalization resemble defects previously observed in unc-33 (CRMP) and unc-44 (Ankyrin) mutants (Maniar et al., 2012). unc-119 is mainly expressed in neurons (Maduro and Pilgrim, 1995) and codes for a highly conserved protein (Maduro et al., 2000), homologous to human UNC119A and UNC119B. Indeed, by introducing UNC-119::GFP specifically in the PVD neuron we were able to suppress the dendritic microtubule defect (Figure 1C), which indicates that UNC-119 functions in a cell intrinsic manner. However, the molecular function of UNC-119 in controlling neuronal development is completely unknown.

To study UNC-119 protein localization, we generated a C-terminal GFP knock-in line using the SEC-approach (Dickinson et al., 2015) and observed GFP expression in all neurons (Figure 1F), as previously reported (Lim et al., 1999; Maduro and Pilgrim, 1995). The GFP signal can be detected in axons (ventral nerve cord and nerve ring) as well as dendrites (amphid nerve). Strikingly, the protein is not enriched in the cell bodies, which suggests that the protein is not freely diffusing in the cytoplasm. To determine UNC-119 protein dynamics, we performed fluorescence recovery after photobleaching (FRAP) microscopy (Diegelman, 1998). We found that both in axons and dendrites the fluorescence hardly recovers and that the immobile UNC-119 pool is 70–80% (Figure 1G–I). Taken together, this suggests that UNC-119 is tightly associated with undefined structures in axons and dendrites, where it is important for neuronal development and microtubule organization.

To understand how UNC-119 functions in organizing neuronal polarity, we performed immunoprecipitations experiments from C. elegans using the unc-119::gfp knock-in animals followed by mass spectrometry to identify potential interaction partners of UNC-119 (Supplementary file 1). Interestingly, one of the top-scoring partners of UNC-119, UNC-33 (CRMP), was reported to be important for neuronal development and microtubule organization (Maniar et al., 2012) and mutations in unc-33 lead to similar microtubule defects (Harterink et al., 2018) as the unc-119 mutant (Figure 1 and Figure 1—figure supplement 1). UNC-33 is a putative microtubule binding protein, which has three isoforms in C. elegans: small (UNC-33S), medium (UNC-33M) and large (UNC-33L), but only the large isoform is able to rescue the neuronal polarity defects of the unc-33 mutant (Maniar et al., 2012; Tsuboi et al., 2005). To validate and map the interaction between UNC-33 and UNC-119, we performed pull-down experiments in cultured Hek293T cells and found that the UNC-33L, but not UNC33S, efficiently binds to UNC-119 (Figure 2A and B). Moreover, UNC-119 was able to pull down the UNC-33L specific N-terminus (Figure 2B), further proving that the UNC-119 interacts with the UNC-33L isoform.

Previous work has shown that UNC-33 (CRMP) acts downstream of unc-44 (Ankyrin) and that the unc-44 mutant has very similar phenotypes as the unc-33 mutant (Harterink et al., 2018; Maniar et al., 2012; Meng et al., 2016). However, we failed to detect a direct interaction between UNC-44 and UNC-33 by immunoprecipitation from Hek293T cells (Figure 2D). In a yeast two-hybrid screen for C. elegans interacting proteins, UNC-119 was identified as a possible binding partner for UNC-44 (Li et al., 2004). This, together with our finding that UNC-119 interacts with UNC-33, suggests that UNC-119 bridges UNC-44 to UNC-33. To test the possible interaction between UNC-44 and UNC-119, we cloned UNC-44 into six fragments (Figure 2A). Using these fragments in Hek293T cells we found that the UNC-44 C-terminus specifically was able to bind to UNC-119 (Figure 2C), which is in agreement with the interaction domain identified by the yeast two-hybrid screen (Li et al., 2004). To test if UNC-44/UNC-119/UNC-33 form a ternary complex we performed pull-down experiments with the UNC-44 C-terminus mixed with either UNC-119::GFP, UNC-33L::GFP or both proteins. We again found that UNC-119 efficiently binds to the UNC-44 C-terminus, however UNC-33L only binds to UNC-44 in the presence of UNC-119 (Figure 2D), showing that UNC-33L, UNC-44, and UNC-119 together can form a ternary complex (Figure 2E). Since mutants for each of these three unc-genes have very similar phenotypes and that the corresponding proteins can form a complex strongly suggests that UNC-33L, UNC-44, and UNC-119 have a common function in neuronal development.

To determine whether the ternary complex also exists in living organisms, we visualized the localization and dynamics of the individual proteins. Using CRISPR/Cas9, we inserted a GFP at the C-terminus of the giant AO13 Ankyrin splice isoform, which is essential for neuronal development (Boontrakulpoontawee and Otsuka, 2002; Otsuka et al., 2002). UNC-44::GFP mainly localizes to axons and dendrites and its expression pattern is nearly indistinguishable from the UNC-119::GFP knock-in (Figure 3B). We observed that UNC-44::GFP clearly localizes to the cortex (Figure 3C), and that it is highly immobile by FRAP in axons and dendrites (Figure 3D–F), as expected from a protein that is part of cortical cytoskeleton (Bennett and Lorenzo, 2016), Moreover, this immobility does not depend on either UNC-119 or UNC-33 (CRMP) (Figure 3G and H, Figure 3—figure supplement 1C and D), which is expected if UNC-44 (Ankyrin) acts upstream of these proteins. However, when we analyzed the dynamics of the UNC-119::GFP in the unc-44 mutant, we found that the UNC-119 recovers rapidly and seems to diffuse freely in the absence of UNC-44 (Figure 3I and J, Figure 3—figure supplement 1E and F). Notably, the UNC-119 immobility was not affected in the unc-33 mutant. Taken together, this data suggests that UNC-44 binds and stabilizes UNC-119.

Since UNC-33 exists as three isoforms and that an N- or C-terminal GFP fusion of UNC-33L disrupts protein function (results not shown), it is not possible to endogenously tag specific isoforms. However, inserting an internal GFP resulted in a functional UNC-33L protein (Figure 3K and Figure 3—figure supplement 1G; Maniar et al., 2012). Therefore, we generated single copy (MosSCI) transgenes for each of the UNC-33 isoforms, which we expressed in the PVD neuron, and found that specifically UNC-33L is able to rescue the neuronal microtubule orientation defect of the unc-33 mutant (Figure 3M). We did not detect a specific subcellular localization for UNC-33L, nor did we observe obvious differences in localization between the three UNC-33 isoforms, except for UNC-33S and UNC-33M being a bit enriched in the cell body (Figure 3L; Maniar et al., 2012). However, FRAP analysis did reveal differences in dynamic properties between the UNC-33 isoforms. UNC-33L signal did hardly recover (immobile fraction is ~85%; Figure 3N–P), whereas UNC-33M and UNC-33S partially recovered (immobile fraction is ~60%). This suggests that the two smaller UNC-33 isoforms have a rapidly diffusing pool as well as an anchored pool. Furthermore, we found that the UNC-33L immobility is completely lost in the unc-44 or unc-119 mutants (Figure 3Q and R; Figure 3—video 1), which indicates that UNC-33 acts downstream of UNC-44 and UNC-119. Similarly, GFP::UNC-33S mobility increases in the unc-119 and unc-44 mutant backgrounds (Figure 3S-T). In conclusion, these results together with the biochemical interaction data suggest that UNC-44 (Ankyrin) immobilizes UNC-33L (CRMP) via UNC-119.

Ankyrin proteins are well-characterized component of the cortical actin-spectrin cytoskeleton (Bennett and Baines, 2001), which forms a characteristic periodic cytoskeleton in invertebrate (Krieg et al., 2017) and vertebrate neurons (Xu et al., 2013). We thus sought to investigate if UNC-44, similar to Ankyrin-G in mammalian and fly neurons (Leterrier et al., 2015; Pielage et al., 2008) forms periodic structures as an indicator for its integration into the cortical cytoskeleton. In order to super resolve the cytoskeleton, we resorted to TIRF-SIM on isolated primary neuron cultures (Materials and methods, Figure 4—figure supplement 1). We observed that UNC-44::GFP (Ankyrin) consistently forms a periodic assembly similar to the 180–200 nm pattern observed with YFP::UNC-70 (β-spectrin) (Figure 4A; p=0.17, pairwise Wilcoxon rank sum test). Importantly, we also observed a ~ 200 nm periodic pattern with the UNC-119::GFP, that corresponds well with the observed UNC-44 distribution (Figure 4B; p=0.82, pairwise Wilcoxon rank sum test). Next, we isolated primary neurons from transgenic animals for the full-length, long version of UNC-33L, which we found to interact with UNC-119, and as well the truncated UNC-33M version, which does not bind to UNC-119. Consistent with this, we observed the periodicity of UNC-33L, in striking similarity to UNC-119 and UNC-44 (Figure 4C–E; p=0.82; p=0.97; pairwise Wilcoxon rank sum test, respectively). In contrast, we did not observe a periodic pattern for the UNC-33M isoform (Figure 4E; p=4e-3, pairwise Wilcoxon rank sum test with UNC-119). The partial periodicity could be explained by the immobile fraction we also observed in the FRAP assay, that associates with microtubules. Taken together, the ternary UNC-44/UNC-119/UNC-33L complex shares a stereotypic periodic pattern similar to the cortical spectrin cytoskeleton.

CRMPs are described as cytosolic phosphoproteins that can bind microtubules via their C-terminus (Lin et al., 2011; Niwa et al., 2017; Zheng et al., 2018). To increase subcellular resolution, we ectopically expressed UNC-33S::GFP and UNC-33L::GFP in the hypodermal seam cells and found that both efficiently bind to microtubules and that the C-terminus is indeed essential for this interaction (Figure 5A and B; Figure 5—video 1). Interestingly, we observed that upon loss of microtubule binding, UNC-33L partially relocalizes to the cortex, whereas the shorter UNC-33S does not. In neurons, immobilization of UNC-33L depends on UNC-44 and UNC-119 (Figure 3) and although expression of these proteins is mainly neuronal, low levels are also present in the hypodermal seam cells (Figure 1F; Chen et al., 2017). Indeed, we found that the cortical relocalization of UNC-33L upon loss of microtubule binding depends on UNC-44 and UNC-119 (Figure 5B; Figure 5—video 1).

To address the dynamics of UNC-33 upon loss of microtubule binding in neurons, we expressed the three UNC-33 isoforms lacking the C-terminus (ΔC) in the PVD neuron (Figure 5—figure supplement 1A). With FRAP analysis, we found that UNC-33SΔC and UNC-33MΔC rapidly recover (Figure 5C–G), which indicates that the immobile pool of UNC-33S and UNC-33M (Figure 3N-P) is the result of microtubule binding. UNC-33LΔC however, is still highly immobile (Figure 5B, H and I). This suggests that although binding to microtubules via the C-terminus is essential for UNC-33L functioning (Figure 5—figure supplement 1B), it is not required for its immobilization. The UNC-33LΔC signal rapidly recovers in the unc-119 mutant (Figure 5J and K), which shows that UNC-119 binds to the N-terminus to immobilize UNC-33L to UNC-44. In agreement, in the unc-44 mutant UNC-33L appears to decorate the dynamic microtubules in the PVD cell body (Figure 3—video 1). Together, these results indicate that in neurons UNC-33L stably associates with microtubules via its C-terminus and to the cortex through its N-terminal interaction with UNC-119.

The stability of the UNC-44/119/33 complex and the attachment of the microtubule binding UNC-33L (CRMP) to the cortex, suggests that the complex anchors microtubules in neurons. To visualize the microtubule cytoskeleton we expressed the microtubule binding protein MAPH-1 in the PVD neuron (Harterink et al., 2018) and found that microtubules are evenly distributed over the primary dendrites and the axon in wild type animals (Figure 6A). In the cell body, some microtubule mobility can be observed (Figure 6—video 1). Interestingly, in each of the unc-mutants we observed neuronal development defects and disorganization of the microtubule cytoskeleton; microtubules were no longer restricted to the primary dendrite, but penetrated the side branches as well (Figure 6A and B). Additionally, in these mutants microtubule mobility in the cell body is increased and we observed microtubule buckling at branchpoints (Figure 6—video 1). To examine the peculiar microtubule mobility, we constructed a TBA-1 (tubulin) with an N-terminal photoactivatable-GFP (PA-GFP) (Kahn et al., 2018; Patterson and Lippincott-Schwartz, 2002). Upon activating a small region in the dendrite, the cytoplasmic tubulin rapidly diffuses, while the fluorescent tubulin that is incorporated in the microtubule remains. In wild type worms, the photo-activated region hardly displaced over the course of several minutes (Figure 6C–E; Figure 6—video 2), which indicates that the neuronal microtubule cytoskeleton in a mature neuron is largely immobilized. Strikingly, this immobility was lost in the unc-44 (Ankyrin), the unc-119 and the unc-33 (CRMP) mutants (Figure 6C–E; Figure 6—video 2) and we observed rapid transport of the photo-activated region with a bias towards the cell body: In the unc-mutants: in 12/20 animals the microtubules translocated towards the cell body and in 8/20 microtubules moved in both directions. The signal from the microtubule incorporated PA-GFP-TBA-1 was generally weaker in the mutants, which might suggest the presence of fewer microtubules in the PVD dendrite. Additionally, using the microtubules plus-end marker EBP-2::GFP we observed subtle changes in microtubule plus-end dynamics and growth event frequencies in the mutants (Figure 6—figure supplement 1A–D). These results show that the UNC-44/119/33 complex functions as a cortical anchoring complex to immobilize the microtubule cytoskeleton in neurons.

The displacement of the microtubules in the unc-mutants suggests that forces are exerted on the microtubule cytoskeleton, potentially by microtubule motors. Previously, unc-116 (kinesin-1) was shown to be important for dendritic microtubule organization, possibly by transporting microtubules (Yan et al., 2013). If the forces on microtubules are indeed generated by UNC-116, the unc-116 mutant should not exhibit increased microtubule dynamics, since the microtubules are still anchored to the cortex by the UNC-44/119/33 complex. Furthermore, the microtubule sliding observed in unc-mutants should be suppressed by depletion of unc-116. We used an endogenously floxed unc-116 mutant, which can knock-out unc-116 specifically in the PVD neuron lineage (Harterink et al., 2018). As anticipated, we found that loss of unc-116 did not induce microtubule mobility. More importantly, the depletion of unc-116 (kinesin-1) completely suppressed the microtubule mobility in the unc-33 (CRMP) mutant (Figure 7A–C; Figure 7—video 1), but not the microtubule polarity defect (Figure 6—figure supplement 1D) nor the morphology defects (not shown). Additionally, the partial loss-of-function mutant for the kinesin-1 adaptor, klc-2 (kinesin light chain; KLC), also suppressed the microtubule mobility of the unc-33 mutant (Figure 7A–C). This shows that in the absence of the cortical anchoring complex kinesin-1 slides microtubules through the neurites. This function appears specific to the kinesin-1 motor, as depletion of unc-104 (kinesin-3) did not suppress the microtubule mobility in the unc-33 mutant (Figure 7A–C; Figure 7—video 1).

Altogether we have found that UNC-119 bridges the membrane-associated UNC-44 (Ankyrin) to the microtubule binding UNC-33L (CRMP) to immobilize the microtubule cytoskeleton (Figure 8A). While in a wild type animal the UNC-116 (kinesin-1) motor is essential for neuronal polarity (Yan et al., 2013), in the absence of either of the components of the ternary complex, UNC-116 induces robust microtubule sliding (Figure 7). We propose that the UNC-44/119/33 complex acts as a membrane-associated anchoring complex to curb kinesin-1 motor activity for proper neuronal development.

To confirm this model, we examined whether artificial anchoring of UNC-33S (CRMP) is sufficient to balance kinesin-1 activity in the absence of the cortical anchoring complex. To do this, we fused UNC-33S, which cannot bind to UNC-119, to the GFP-nanobody (vhhGFP) and expressed it in the PVD neuron of unc-44::gfp knock-in animals (Figure 8A). Artificial coupling of UNC-33S to UNC-44 rescued the neuronal microtubule defects of the unc-33 and unc-119 mutants (Figure 8B) as well as the neuronal morphology of the unc-33 mutant (Figure 8C). Remarkably, expression of the vhhGFP::UNC-33S fusion protein in all neurons rescued the animal movement of the unc-119 mutant (Figure 8D and E and Figure 3—figure supplement 1B; Figure 8—video 1), showing that neuronal functioning is restored and that the unc-119 mutant locomotion defect is caused by a loss of UNC-33L anchoring to the cortical UNC-44 (Ankyrin) in neurons.

Strains were cultured using standard conditions at 15°C or 20°C and imaged at room temperature. All strains used for this study can be found in the key recourse table and if generated in this study are available upon request. New transgenic strains were generated by injection of the indicated constructs forming extra-chromosomal lines, single copy insertions using mosSCI (Frøkjaer-Jensen et al., 2008) or by endogenous insertions/deletions using CRISPR. unc-33(hrt7), unc-44(hrt8) were generated by injection of the guide + Cas9 plasmid + Pmyo-2::tdTomato into STR318 and screening for Unc phenotype. Similarly, the deletion of cb-unc-119 form mosSCI lines was performed by injecting cb-unc-119 sgRNA with Cas9 protein and scoring for the Unc phenotype. To generate unc-44(hrt5[GFP]) we used pha-1 co-conversion approach (Ward, 2015) and to generate unc-119(hrt13[GFP]) we used the self-excising cassette-approach (Dickinson et al., 2015). Constructs were generated using multisite Gateway cloning (Invitrogen) or by Gibson assembly using NEBuilder (New England Biolabs) and were sequenced for verification (all cloning details can be found in Supplementary file 2).