We explored the origins of MT organization using the PVD neuron in C. elegans, a sensory neuron with stereotypical morphology of its axon and non-ciliated dendrites (Albeg et al., 2011). Early PVD morphogenesis is temporally and spatially stereotyped: the axon always grows out first, followed by the anterior and then posterior dendrite, with all emerging neurites oriented in the direction of their mature neurite pattern. The elaborate branches develop after the primary dendrites fully extend (Figure 1A, Figure 1—figure supplement 1A and B). We previously showed that the mature anterior primary dendrite contains largely minus-end-out MTs while the posterior dendrite and axon have plus-end-out MTs (Taylor et al., 2015). To investigate the establishment of minus-end-out MTs during dendrite outgrowth (step III in Figure 1A), we visualized the endogenous localization of the MT plus-end tracking protein EBP-2::GFP/EB1 in developing PVD anterior dendrites and found that numerous EBP-2 comets emerged from a single region within the distal neurite, suggesting the presence of a dendritic growth cone MTOC (dgMTOC, Figure 1B and Video 1). To characterize the directionality of these EBP-2 comets, we plotted the frequency of plus-end-out and minus-end-out MTs at interval distances in the anterior dendrite, setting the center of the apparent dgMTOC as zero (Figure 1C, Figure 1—source data 1). Near the most distal region of the dendrite (~3–6 µm from the dendrite tip), almost all the EBP-2 comets move toward the distal tip indicating plus-end-out microtubules, while nearly all the comets along the shaft of the anterior dendrite move toward the cell body indicating only minus-end-out MTs in this region (Figure 1B and C). The emergence of a high frequency of EBP-2 comets from a discrete region near the dendrite tip (mean = 0.44 comets/s±0.11 (SD) in a 3.2 µm region), which also encompasses a transition zone of MT directionality, is consistent with the presence of a local MTOC (Figure 1B and C). This MT organization is apparent as soon as a morphologically distinct growth cone-tipped anterior dendrite has outgrown from the cell body and remains throughout the outgrowth of the anterior dendrite (data not shown). In addition, the number of comets generated near the dendrite tip greatly exceeds comets from any other discrete point or region in the dendrite or from the cell body throughout the morphogenesis of the anterior dendrite, indicating that the dgMTOC is the only apparent MTOC. The axon and posterior dendrite, which contain predominantly plus-end-out MTs, do not contain similar structures (Figure 1B and Figure 1—figure supplement 1C). Significantly, we find evidence for a similar dgMTOC near the tip of outgrowing Drosophila melanogaster Class I sensory vpda primary dendrites (Figure 1—figure supplement 1D and Video 2), suggesting a conserved mechanism for the genesis of minus-end-out MTs in the primary dendrite.

To further confirm the existence and location of the MTOC, we performed time-lapse imaging on worms expressing GFP::TBA-1/α-tubulin in PVD, which allowed us to track MT polymerization and depolymerization events. TBA-1 was enriched in a region immediately behind the growth cone (Figure 1—figure supplement 1E and F, Figure 1—source data 1), consistent with the presence of a MTOC. The relatively low number of MTs in C. elegans neurons (Yogev et al., 2016) allowed us to identify individual MTs on kymographs. This analysis also revealed an apparent dgMTOC from which plus-end-out MTs extended distally toward the dendritic tip and minus-end-out MTs extended proximally toward the cell body (blue and red lines, respectively, in Figure 1D). Consistent with the direction of EBP-2 comets, we observed numerous polymerization events originating from the dgMTOC and directed toward the cell body (red lines in Figure 1D). A similar frequency of depolymerization events was also observed in this region (green lines in Figure 1D and Figure 1E, Figure 1—source data 1), indicating that the majority of dgMTOC-derived minus-end-out MTs are highly dynamic. In summary, both EBP-2 and TBA-1 dynamics suggest an active MTOC localizes to the dendritic growth cone region.

Microtubules are nucleated by the conserved γ-tubulin ring complex (γ-TuRC), a ring of γ-tubulin complex proteins (GCPs) that templates the assembly of new MTs (Oakley et al., 1990; Zheng et al., 1995). γ-TuRCs localize to MTOCs, including the centrosome, the best studied MTOC which generates MTs from within its pericentriolar material (PCM) in dividing animal cells to form the mitotic spindle (Conduit et al., 2015). To understand the molecular components of the dgMTOC, we examined the localization of endogenous GFP::GIP-1/GCP3 and GIP-2::GFP/GCP2 – core components of the γ-TuRC – during dendrite outgrowth. We found that GIP-1 and GIP-2 consistently localized near the dendrite tip in a cluster of punctate structures which track with the growth cone in the outgrowing anterior dendrite (Figure 2A and B). GIP-1 precisely localized to the site from which EBP-2 comets originated (Figure 2C) and quantification of the distribution and location of GIP-1 and GIP-2 clusters showed that their location closely match the dgMTOC regions defined by EBP-2 comets (Figure 2D and E, Figure 2—source data 1). These γ-TuRC clusters are specific to the outgrowing anterior dendrite since no GIP-1 or GIP-2 clusters were found in the cell body, axon, or posterior dendrite during their outgrowth (Figure 2A, Figure 2—figure supplement 1A and B), consistent with the lack of an apparent MTOC in these structures. Furthermore, growth cone-localized GIP-2 was visible as soon as the anterior dendrite emerged (Figure 2—figure supplement 1C) but absent near the tip of mature dendrites (Figure 2—figure supplement 1D), suggesting that the dgMTOC plays a role in the establishment of MT organization during dendrite development. Significantly, GIP-2 also localized close to the dendrite tip during dendrite outgrowth in the C. elegans DA9 motor neuron (Figure 2—figure supplement 1E), further supporting a more general role for the dgMTOC in establishing dendritic minus-end-out MTs.

To investigate the significance of the dgMTOC and directly test if γ-TuRC is required for the dgMTOC activity, we performed conditional knockdown of GIP-1 in the PVD lineage using the ZIF-1/ZF degradation system previously established in C. elegans (Armenti et al., 2014; Sallee et al., 2018). We inserted a ZF tag into the endogenous gip-1 locus, enabling the controlled degradation of endogenous GIP-1 upon expression of the E3 ubiquitin ligase substrate-recognition subunit ZIF-1. Expression of ZIF-1 in the developing PVD lineage led to a dramatic reduction of punctate GFP::GIP-1 signal in ~80% of outgrowing PVD anterior dendrites (GIP-1^PVD(-), Figure 2—figure supplement 1F). We then examined EBP-2 comets in GIP-1^PVD(-) animals, which revealed a striking reversal of anterior dendrite MT polarity leading to plus-end-out orientation in both outgrowing and mature dendrites in about half of the animals (Figure 2F and G, Figure 2—source data 1). Unlike in wild-type (wt) animals where the majority of EBP-2 comets were generated by the dgMTOC, in GIP-1^PVD(-) animals, many dendritic comets originated from the cell body (Figure 2F and Video 3). The incomplete penetrance of this phenotype in GIP-1^PVD(-) animals is likely due to partial knockdown of GIP-1, as shown by the bimodal distribution of the MT polarity phenotype (Figure 2G). These results indicate that γ-TuRC is a critical component of the dgMTOC and that the dgMTOC is essential for establishing minus-end-out MTs in developing and mature dendrites.

To determine the subcellular ultrastructure of the dgMTOC, we performed serial sectioning electron microscopy (EM) reconstruction of developing distal PVD dendrites. We began by reconstructing MTs to identify the dgMTOC region in our sections. In our EM series, numerous short, staggered MTs were observed with their ends unambiguously identified. Consistent with our EBP-2 and TBA-1, the highest MT density was found ~2–4 µm from the dendritic tip, consistent with the presence of an MTOC in this region (Figure 3A–C, Figure 3—source data 1).

Serial EM reconstruction showed no evidence of a centriole or Golgi outpost in the dgMTOC area. Instead, numerous clear- and dense-core vesicles were found in between the MT arrays (Figure 3A and B). In particular, clusters of clear-core vesicles (Figure 3B, white) were found in the region with the highest MT numbers suggesting that GIP-1 and GIP-2 might localize to these vesicles to function in the dgMTOC. To investigate the molecular identity of these vesicles, we expressed GFP-tagged markers to label different vesicles and membrane compartments. Synaptic vesicles are the most abundant clear-core vesicles in neurons; however, the synaptic vesicle marker RAB-3, which robustly localized to the axon (Maniar et al., 2012), was not enriched in the dendritic growth cone region (Figure 3—figure supplement 1A). Consistent with our EM results, the early/medial Golgi cisternae markers RER-1 (Sato et al., 2011) and AMAN-2/alpha-mannosidase 2, the latter of which localizes to Golgi outpost MTOCs at mature dendrite branch sites in D. melanogaster (Ori-McKenney et al., 2012), were only present in the cell body but not in the growth cone (Figure 3—figure supplement 1B and C). Surprisingly, the late-Golgi compartment marker RAB-6.2 (Zhang et al., 2012) was enriched in the growth cone in a localization pattern similar to that of γ-TuRC (Figure 3—figure supplement 1B and C). In addition to localizing to Golgi stacks, RAB-6.2 also plays a role in recycling cargo molecules from endosomes to the trans-Golgi network in neurons (Zhang et al., 2012). Therefore, we further examined recycling endosomes and found that the recycling endosome marker RAB-11.1 co-localized and moved together with RAB-6.2-labeled vesicles in the growth cone (Figure 3—figure supplement 1D and E). Notably, RAB-11.1 showed striking co-localization with GIP-2 (Figure 3D), with 81.7 ± 3.5% (SEM, n = 11) of RAB-11.1 fluorescence overlapping with GIP-2 and 80.6 ± 5.2% (SEM, n = 11) reciprocal GIP-2 fluorescence overlapping with RAB-11.1 in the growth cone (Figure 3E, Figure 3—source data 1). Time-lapse recordings revealed that RAB-11.1 displayed a nearly identical movement pattern to that of GIP-2, with both proteins moving together in the growth cone (Figure 3F and G). A frame-to-frame analysis of RAB-11.1 and GIP-2 distribution showed high correlation in the growth cone (Figure 3—figure supplement 1F), but no correlation in the cell body (Figure 3—figure supplement 1G; Costes et al., 2004; Manders et al., 1993). Co-localization and co-trafficking of RAB11.1 and GIP-2 suggest that Rab11-endosomes may be at the core of the dgMTOC.

Interestingly, Rab11-associated endosomes and even Rab11 itself have been found to influence the MT organization of mitotic spindles (Hehnly and Doxsey, 2014). We therefore interfered with RAB-11.1 function using two different approaches and assessed changes to γ-TuRC localization and MT polarity. Expression of a dominant negative RAB-11.1(S25N) (Ren et al., 1998) caused a loss of GIP-2 puncta from the dendritic growth cone in ~35% of animals (Figure 3H and J, n = 28, Figure 3—source data 1) and a reversed or mixed MT polarity in ~10% of animals (Figure 3—figure supplement 1H, n = 28). We speculate that some undetectable GIP-2 may remain in some growth cones in the 35% of RAB-11.1(S25N) overexpressing worms explaining a less frequent disruption of MT polarity than GIP-2 localization. As an alternative approach to test the function of RAB-11.1 in dgMTOC organization or localization, we generated Lox sites around the endogenous rab-11 gene and expressed Cre specifically in PVD to knock out rab-11 cell-specifically. 20% of Cre expressing worms displayed a loss of GIP-2 puncta in the dendritic growth cone with mislocalization of GIP-2 as either multiple dim puncta dispersed within the cell body (12%, Figure 3J,i1 and i2 in Figure 3I, n = 41) or along the dendrite shaft (7%, i3 in Figure 3I), suggesting transportation or clustering defects.

The somewhat low penetrance effects on GIP-2 localization we found with both approaches to interfering with RAB-11.1 function may be due to multiple factors. Expression in the PVD lineage may not be sufficiently early and a small variable amount of functional endogenous RAB-11.1 may remain. Additionally, Rab11 itself may not be required for the biogenesis of these endosomes but rather play a role in their function, such as transportation (Hehnly and Doxsey, 2014), and may act in combination with other partially redundant proteins to correctly localize the dgMTOC. Taken together, these results suggest that γ-TuRC localization to RAB-11.1-positive endosomes in the growth cone underlies the dgMTOC, which generates short plus-end-out MTs in the distal growth cone and dynamic minus-end-out MTs that populate the growing dendrite.

Given the specific positioning of the dgMTOC, we predicted that the subcellular localization of the dgMTOC to the dendritic growth cone is instructive to populate the proximal dendrite with minus-end-out MTs. We tested this hypothesis by perturbing the localization of the dgMTOC. We previously showed that the mature PVD anterior dendrite loses minus-end-out MTs and gains plus-end-out MTs in unc-116/kinesin-1 mutants (Taylor et al., 2015). Consistent with the MT polarity defect in mature PVD dendrites, unc-116(e2310) mutants showed a large decrease in the percentage of minus-end-out MTs in the anterior dendrite shaft during outgrowth (wt: 93.8 ± 2.7% (SEM, n = 11); unc-116(e2310): 21.4 ± 10.4% (SEM, n = 12); Figure 4A and B, Figure 4—source data 1). This dendritic MT polarity defect in unc-116 mutants was also evident for GFP::TBA-1 dynamics (Figure 4—figure supplement 1B). Additionally, the reduced minus-end-out frequency in unc-116 mutants was partially rescued by expressing a wt unc-116(+) transgene during early neurite outgrowth (using the unc-86 promoter), but not rescued by expressing unc-116(+) after anterior dendrite outgrowth (using a ser-2 promoter), suggesting that UNC-116/Kinesin-1 is required specifically during early neurite outgrowth to establish dendritic minus-end-out MTs (Figure 4—figure supplement 1A, Figure 4—source data 1).

Given that UNC-116 functions during an early developmental stage and that unc-116 mutants display MT polarity defects during dendrite outgrowth, we considered that UNC-116 might be required for dgMTOC localization or function. To test this hypothesis, we examined EBP-2 dynamics in the outgrowing dendritic growth cone where the dgMTOC normally resides. In striking contrast to wt, unc-116 mutants lacked a convergence of plus-end-out and minus-end-out MTs in the outgrowing dendritic growth cone. Instead, the vast majority of MTs originated from the cell body and were plus-end-out throughout the anterior dendrite (Figure 4A, Figure 4—figure supplement 1C and D and Video 4), raising the possibility that the dgMTOC is inactive or mislocalized in unc-116 mutants. By tracing the origin of EBP-2 comets over time in unc-116 mutants, we frequently observed that many EBP-2 comets emanated from a single region within the cell body indicating a putative MTOC (Figure 4C and Video 5) – a stark contrast to the lack of a point-of-origin for EBP-2 comets in GIP-1^PVD(-) animals (Video 3). These results suggested that the dgMTOC remains in unc-116 mutants but is mislocalized to the cell body, a possibility we assessed by examining the localization of GIP-2 and RAB-11.1. Unlike in wt, GIP-2 was largely absent from the outgrowing dendritic growth cone and instead formed a cluster within the cell body in 64% of unc-116 mutants (n = 45, Figure 4D and Figure 4—figure supplement 1E, Figure 4—source data 1). RAB-11.1 also formed a cluster within the cell body rather than in the growth cone region in 71% of unc-116 mutants (n = 28, Figure 4E and Figure 4—figure supplement 1F, Figure 4—source data 1), providing further support that Rab11 endosomes are a component of the dgMTOC. Occasionally, we found animals with a GIP-2 cluster in the posterior dendrite (Figure 4—figure supplement 1G), consistent with a rarely observed posterior dendrite-localized MTOC in unc-116 mutants (Figure 4—figure supplement 1H). Together, these results indicate that the kinesin-1 motor is required for localizing the dgMTOC to the outgrowing dendritic growth cone. Furthermore, while depletion of GIP-1 in PVD showed that the dgMTOC is required to establish dendritic minus-end-out MTs, this unc-116 mutant phenotype suggests that the location of the dgMTOC instructs MT polarity.

Next, we explored the mechanisms by which kinesin-1 localizes the dgMTOC. Time-lapse analyses of GIP-2 cluster movement showed that GIP-2 fluorescence undergoes processive movements toward the distal tip with several features. First, GIP-2 movements are largely unidirectional toward the distal dendrite tip, which is evident in kymograph analysis (f1 in Figure 4F), GIP-2 cluster localization at different time points (left and right panels of f2 in Figure 4F) and quantification of the anterior displacement of the GIP-2 cluster (Figure 4K). Second, GIP-2 localization alternates between a tight cluster and a broader cluster with more dispersed puncta (middle and right panels of f2 in Figure 4F). Third, the movements are interspersed by pauses (middle panel of f2 in Figure 4F), which is also evident by the alternating slopes and plateaus in GIP-2 moving distance analyses (black trace in Figure 4I and Figure 4—figure supplement 2A, Figure 4—source data 1). Most GIP-2 exists in a bright cluster during pauses, while the fluorescence is dispersed into multiple smaller puncta during the mobile phase (f2 in Figure 4F and Video 6). This phenomenon can be measured by the periodical changes of the diameter of the GIP-2 cluster during the movement cycle (Figure 4H, red trace in Figure 4I and Figure 4—figure supplement 2A, Figure 4—source data 1). These results indicate that GIP-2 moves in a saltatory manner, with cycles of aggregation and dispersal, toward the distal dendritic tip. We hypothesized that kinesin-1 moves the GIP-2-positive endosomes toward the distal dendritic tip along the plus-end-out MTs in the growth cone. Consistent with this idea, unc-116/kinesin-1 mutants showed largely reduced movement or dispersal of GIP-2 clusters (Figure 4G and J–L, Figure 4—figure supplement 2B and Video 7, Figure 4—source data 1). These analyses suggest that UNC-116/kinesin-1 moves GIP-2 containing endosomes in the outgrowing dendrite.

With MTs emanating from the dgMTOC in both directions, the unidirectional movements of GIP-2 suggest that kinesin-1 prefers transport on the plus-end-out MTs stretching toward the tip of the dendritic growth cone rather than the minus-end-out MTs growing toward the cell body. To investigate differences between the populations of MTs originating from the dgMTOC, we examined TBA-1 dynamics in the outgrowing dendrite (Figure 5A). Interestingly, we noticed that unlike minus-end-out MTs which transitioned rapidly between polymerization and depolymerization (Figures 1C and D and a3 in Figure 5A, triangle-shaped peaks in kymographs, Figure 5—source data 1), the individual plus-end-out MTs displayed frequent pauses between polymerization and depolymerization that lasted 10.34 ± 0.85 s (SEM, n = 34) (a2 in Figure 5A, plateaus, Figure 5B and C, Figure 5—source data 1), indicating that the distal plus-end-out MTs are transiently stabilized, likely through interaction with the growth cone. We note that the distal plus-end-out MTs originating from the cell body in unc-116 mutants also paused in the tip of the dendrite and showed similar pause frequency and time as wt animals (Figure 5B and C, Figure 5—source data 1), indicating that the pausing of plus-end-out MTs in the growth cone is independent of UNC-116 function. These data reveal an asymmetry in the behavior of dgMTOC-generated MTs, with plus-end-out MTs pausing more than minus-end-out MTs.

Kinesin-1 has been shown to prefer transporting cargoes on stable over dynamic MTs (Cai et al., 2009; Konishi and Setou, 2009; Tas et al., 2017). To assess whether a preference of kinesin-1 for stable MTs could explain the biased dgMTOC movements, we simultaneously recorded TBA-1 dynamics and RAB-11.1 movements and assessed whether endosome movement corresponded with the presence of paused rather than dynamic MTs. No obvious distal directed RAB-11.1 movements were observed when only dynamic plus-end-out MTs were present (d1 in Figure 5D). In contrast, endosome movements toward the distal dendritic tip were observed in the presence of stable plus-end-out MTs (d2 in Figure 5D, orange lines). To correlate the RAB-11.1 movements with MT dynamics, we quantified the displacement of RAB-11.1 over time and determined and labeled time periods which have paused plus-end-out MT(s) (red, Figure 5E and Figure 5—figure supplement 1A) and periods which only have dynamic plus-end-out MTs (black, Figure 5E and Figure 5—figure supplement 1A). We found that endosome movements preferentially occurred during periods with paused MTs (red portion of solid line in Figure 5E and red portion of Figure 5—figure supplement 1A) and there was no obvious RAB-11.1 movement when the plus-end-out MTs were dynamic (dashed line in Figure 5E and black portion in Figure 5—figure supplement 1A, Figure 5—source data 1). By sampling of a population of worms, we found that RAB-11.1 endosome moving distance showed a positive correlation with the total MT pause time overall (Figure 5F, Figure 5—source data 1). These results suggest that distally-directed endosomes move on paused or stable plus-end-out MTs in a kinesin-1-dependent manner.

Taken together, these findings are consistent with a model in which kinesin-1 transports γ-TuRC-positive endosomes toward the dendritic tip as the growth cone advances. The γ-TuRC at the dgMTOC builds plus-end-out MTs toward the growth cone, which serve as tracks for the next bouts of kinesin-1-mediated endosome movements. Individual plus-end-out MTs are stable for about 10 s followed by depolymerization events (a2 in Figure 5A). The transient nature of these ‘stabilized’ plus-end-out MTs is needed to ensure that the majority of MTs are still minus-end-out once the MTOC has passed this area.

While kinesin-1-mediated motility explains the processive movements of the dgMTOC towards the dendritic tip, it is not clear how the multiple γ-TuRC puncta aggregate to re-form a single cluster after their dispersal during movement, which is likely critical to maintain a singular dgMTOC. Cytoplasmic dynein has been shown to concentrate organelles such as Golgi or Rab11 endosomes near MTOCs in non-neuronal cells (Corthésy-Theulaz et al., 1992; Horgan et al., 2010). We therefore tested whether dynein is required for dgMTOC clustering by examining dhc-1/dynein heavy chain mutants. dhc-1(or195), a temperature-sensitive dynein mutant, displayed several types of defects in dgMTOC localization when animals were cultured at the nonpermissive temperature. In ~30% of dhc-1 mutants, GIP-2 clusters were completely absent from the PVD anterior dendrite and the rest of the cell (compare wt localization in Figure 6A to b1 in Figure 6B, Figure 6C, Figure 6—source data 1). In another ~40% of mutant animals, one or several GIP-2 cluster(s) could be found in the cell body and occasionally in the posterior dendrite (b2 in Figure 6B, Figure 6C). In the remaining ~30% of mutant animals, a GIP-2 cluster still localized to the anterior dendrite but showed a variable distance to the dendritic tip which was longer than in wt animals (b3 in Figure 6B, Figure 6C and D, Figure 6—source data 1). Time-lapse imaging analysis of GIP-2 movements in the subset of dhc-1 mutant worms which have an anterior dendrite-localized GIP-2 cluster showed that net GIP-2 movement toward the dendritic tip was reduced compared to that of wt, consistent with its ectopic localization. Strikingly, further analyses of dhc-1 mutants showed that GIP-2 puncta exhibited repeated dispersal and failed to form a relatively stable singular cluster (Figure 6E–I, Figure 4—figure supplement 2C, Video 8 and Figure 4—source data 1). The continuous dispersal of GIP-2 clusters led to a reduction in the maximal intensity of the cluster over time which might explain the lack of GIP-2 clusters in a subset of dhc-1 mutants (Figure 6J and K, Figure 6—source data 1). Endogenously tagged DHC-1 localized to the dgMTOC region as a cluster and like the GIP-2 clusters remained associated with the base of the advancing growth cone (Figure 6L). These dynamic movements closely coincided with those of RAB-11.1 foci (Figure 6M). This striking subcellular localization of DHC-1 further suggests that DHC-1 functions locally at the dgMTOC to move RAB-11.1 endosomes. Consistent with the defect in dgMTOC localization, we found that about 50% of dhc-1 mutants showed complete reversal of MT polarity (n1 in Figure 6N) while another 20% of mutants showed mixed plus-end-out and minus-end-out MTs in the anterior dendrites (n2 in Figure 6N) perhaps due to the dispersed GIP-2 clusters in these mutants (Figure 6N and O, Figure 6—source data 1). These results suggest that dynein uses its minus-end-directed motor activity to re-cluster γ-TuRC-positive endosomes together which had dispersed during translocation towards the distal tip of the dendritic growth cone by kinesin-1, and this clustering is required for establishing dendritic minus-end-out MT polarity.

Worms were raised on NGM plates at 20°C using OP50 Escherichia coli as a food source. Worm strains which contain temperature sensitive mutant dhc-1(or195) were maintained in 16°C and were shift to 25°C for phenotype analysis. C. elegans strains used in this study are listed in the key resources table. OD2509 [gip-2(lt19[gip-2::gfp]::loxP::cb-unc-119(+)::loxP) I; unc-119(ed3) III] was a gift from Dr. Karen Oegema at the University of California San Diego (Wang et al., 2017). The generation of JLF273 ebp-2(wow47[ebp-2::gfp]) II; zif-1(gk117) III; wowEx10 (used to make TV23841), JLF38 gip-1(wow5[zf::gfp::gip-1]) zif-1(gk117) III; wowEx10 (used to make TV24455), and JLF155 zif-1(gk117) has been described (Sallee et al., 2018). Strains are available upon request.

Worms were prepared for conventional EM by high pressure freezing/freeze-substitution. Worms in E. coli containing 20% BSA were frozen in 100 µm well specimen carriers (Type A) opposite a hexadecane coated flat carrier (Type B) using a BalTec HPM 01 high-pressure freezer (BalTec, Lichtenstein). Freeze-substitution in 1% OsO4, 0.1% uranyl acetate, 1% methanol in acetone, containing 3% water (Buser and Walther, 2008; Walther and Ziegler, 2002) was carried out with a Leica AFS2 unit. Following substitution, samples were rinsed in acetone, infiltrated and then polymerized in Eponate 12 resin (Ted Pella, Inc, Redding, CA). Serial 50 nm sections were cut with a Leica UCT ultramicrotome using a Diatome diamond knife, picked up on Pioloform coated slot grids and stained with uranyl acetate and Sato’s lead (Sato, 1968). Sections were imaged with an FEI Tecnai T12 TEM at 120 kV using a Gatan 4k × 4 k camera. TrakEM2 in Fiji was used to align serial sections (Cardona et al., 2012; Schindelin et al., 2012). Modeling of serial sections was performed with IMOD (Kremer et al., 1996).

Plasmids and primers used to generate transgenic or knock in C. elegans strains in this study are listed in the key resources table. Detailed information on plasmid generation and primer sequences are provided in Supplementary file 1. Expression clones were made in the pSM vector, a derivative of pPD49.26 (A. Fire) with extra cloning sites (S. McCarroll and C.I. Bargmann, personal communication). rab-11.1(wy1444[lox]) was generated by co-injecting oligo donors and CAS9 RNPs (Dokshin et al., 2018). Punc-86::mCherry::PLCdeltaPH was generated using Gibson cloning. Transgenic strains (1–50 ng/μl) were generated using standard techniques and coinjected with markers Podr-1::GFP or Podr-1::RFP. Plasmids are available upon request.

Gravid adults were bleached in a hypochlorite solution to obtain embryos which were washed in M9 and allowed to hatch either in M9 or on unseeded NGM plates overnight to obtain a population synchronized in L1 arrest. This L1-arrested population was kept for a maximum of 5 days and transferred to OP50-seeded NGM plates at different times to achieve specifically aged synchronized L2 populations for imaging.

To image early outgrowing PVD neurites, wild type and GIP-1 knockdown L1-arrested animals were grown on OP50-seeded NGM plates at 22°C for 18–19 hr or 25°C for 16–17 hr before imaging. unc-116(e2310) L1-arrested animals were grown on OP50-seeded NGM plates at 22°C for 20–22 hr or 25°C for ~18 hr before imaging.

dhc-1 l1-arrested worms were obtained at 25°C and then were grown on OP50 seeded NGM plates at 25°C for 16–18 hr.

Wild-type GFP::TBA-1 animals were imaged 1.5–2 hr later than other wild-type image to get a more clear MT dynamic kymograph in the growth cone region, as in the later outgrowth stage, the MT number in the growth cone region will slightly reduce.

To image mature PVD neurites, bleached adults were placed on OP50-seeded NGM plates for about 48 hr at 20°C, then the mid L4 worms were picked to image for EBP-2 dynamics and GIP-2 localization.

To image the outgrowing DA9 neuron, some 3-fold stage embryos were transferred to a new plate and then the L1 worms right after hatching were imaged.

Just prior to imaging, C. elegans animals were mounted on 3–5% agarose pads. L2s or L1s were picked and released into a 1 μl M9 or water droplet on an inverted coverslip while minimizing bacterial transfer. Prior to mounting on the freshly made agarose pad, the droplet was surrounded by a 1 μl droplet of 0.05 µm Polysterene Polybeads (Polysciences) and a droplet of levamisole (final concentration of approximately 3 mM) to immobilize worms. L4s were picked to M9 directly on the agarose pad with levamisole added prior to mounting. Slides were sealed with VALAP or Vaseline prior to time-lapse imaging. All imaging was performed within 40 min of mounting.

Imaging of C. elegans was performed on an inverted Zeiss Axio Observer Z1 microscope equipped with a Yokogawa spinning disk, QuantEM:512SC Hamamatsu camera (set to 600 EM Gain), a Plan-Apochromat 100x/1.4 NA objective (Zeiss), 488 nm and 561 nm lasers, and controlled by MetaMorph Microscopy software (Molecular Devices).

Endogenous EBP-2::GFP dynamics in outgrowing and mature PVD dendrites were imaged using 50% 488 nm laser power, 100 ms exposure, and with 200 ms time interval between acquisitions. 200–300 frames were recorded per animal. The same imaging parameters were also used to image GFP::TBA-1 dynamics in outgrowing dendrite.

To assess protein localization within PVD during neurite outgrowth, endogenously-GFP-tagged GIP-1, GIP-2 were imaged using 70–80% 488 nm laser power recorded with 300 ms exposure. The membrane mCherry co-marker was imaged using 70–80% 561 nm laser power with 200 ms exposure. To fully sample PVD in three dimensions, z-sections were imaged at Nyquist resolution, every 0.25 μm from above to below a single PVD in each animal. Micrographs in Figure 2A, and Figure 2—figure supplement 1A are maximum projections of the z-sections that include the PVD dendrite.

For time-lapse imaging of these strains to assess dynamics of complex localization to the outgrowing dendrite tip, 488 nm laser power was reduced to 50% with 200 ms exposures to reduce phototoxicity and photobleaching. Up to 4 z-sections were imaged at 0.5 μm, and acquisitions occurred every 15 or 30 s for up to a total of 60 min. Note that Figure 2B displays only frames every 60 s even if animals were imaged more frequently. Only animals which displayed a steady rate of growth cone advance indicating healthy animals were used for analysis.

For time lapse imaging to look at the details of GIP-2, RAB-11.1 and DHC-1 dynamics at the growth cone region, only one z section is acquired, and images were taken every 200 ms for up to 200–300 frames, and images were taken with a 100 ms exposure time and 70% laser power for both 488 nm and 561 nm channels. To image TBA-1 together with RAB-11.1, the 488 nm laser power was reduced to 50% while other parameters were the same.

Images were processed and analyzed using MetaMorph (Molecular Devices) and ImageJ to create kymographs or psuedocolored merged maximum intensity micrographs and assembled into figures using Adobe Photoshop and Illustrator. Statistical calculations and graphing were done in Prism 7 and Prism8 (GraphPad).

To display the dynamics of endogenously-GFP-tagged GIP-1 and GIP-2 localization to the PVD dendritic growth cone over time in Figure 2B,a 20 pixel-wide line segment was drawn over the growth cone outgrowth region and processed using the ImageJ function ‘Straighten’ to slightly straighten the region. The Make Montage function was then used to create a montage displaying that region every 60 s. Due to minor movements of the worm, uncropped merged time-lapse images were registered to each other using the ImageJ StackReg function before making a montage for GIP-1.

Kymographs were made in ImageJ by drawing a straight or segmented line (width of 6–10 pixels) along a process and using the KymographBuilder plugin or Stacks-Reslice function on the stacked time-lapse file.

To quantify EBP-2::GFP comet direction and frequency at interval distances across the outgrowing dendrite (line graphs in Figure 1C, Figure 4—figure supplement 1C and D), kymographs were created from recordings of numerous animals and the number of EBP-2::GFP minus-end-out and plus-end-out tracks was manually counted at each specified distance. For Figure 1C, the MTOC region was first defined as the region in which the majority of both directions of comets originated, and the center of that region was designated as zero.

To quantify overall MT polarity in the outgrowing PVD dendrite, for outgrowth stage, the imaging region was the whole process while the polarity quantification was done in the proximal dendrite region which is within 50 μm from cell body and excluding the MTOC region; for the mature dendrite, the quantification is also done in the proximal dendrite.

To quantify the co-localization between GIP-2 and RAB-11.1in different worms (Figure 3E), a 30 × 20 pixel ROI was drawn in the growth cone region first and then was duplicated to independent images for both GIP-2 and RAB-11.1 channels, then a random GIP-2 or RAB-11.1 localization image was generated by ImageJ JACop plugin, and then the thresholded Mander’s split colocalization coefficients(tM) values were measured by the Colocalization Threshold plugin which can generate the threshold automatically between GIP-2 and RAB-11.1, GIP-2 and randomized RAB-11.1, randomized RAB-11.1 and GIP-2.

To quantify the co-localization between GIP-2 and RAB-11.1 over time in the same worm, the thresholded Mander’s split colocalization coefficients, Pearson’s correlation coefficient (Rcoloc) and the scatterplot files were all generated by the Colocalization Threshold plugin for 100 frames in both the growth cone and cell body region.

To measure the GIP-2 cluster diameter and distance from t = 0 (Figure 4H–J, Figure 6G and H, Figure 4—figure supplement 2A–C), as shown in Figure 4H,a segmented line was draw along the GIP-2 moving track, and then the plot file for all time frames were generated in ImageJ and exported to an excel file to get the intensity value of different positions along the line and also the intensity values at different time points. Then the same segmented line was moved to a nearby region to get the background intensity value and also exported to an excel file. The final GIP-2 intensity value was the original intensity value minus the background intensity value. The pixel value of the cluster diameter at one given time point D_t was the distance between the first (P_first) and last (P_last) position in which the intensity value is above the 50% maximum intensity value in the same time point. The pixel position of the cluster P_t was the center of P_first and P_last (Figure 3F). And then:Diameter(μm)=0.109∗DtDistancefromt=0(μm)=−0.109∗(Pt−P0)(0.109μmpixelforthe100Xobjectivecalibration)

The RAB-11.1 distance from t = 0 (Figure 5E and F, Figure 5—figure supplement 1A) was calculated the same way as GIP-2, but the threshold was set to above 70% maximum RAB-11.1 intensity, as the mCherry::RAB-11.1 showed a higher background than GIP-2.

To measure the GIP-2 cluster maximum intensity dynamics during movement (Figure 6J), the background of the movie was subtracted using the ImageJ subtract background function, then the maximum intensity of both the green GIP-2 channel and red PVD morphology marker channel at different time points were measured using the ImageJ ROI multiple measure function. The green GIP-2 intensity was divided by the red PVD morphology channel intensity to correct for focus change or photo-bleaching, and the normalized GIP-2 maximum intensity at different time points we calculated by setting the start value as 1.

The area under the different maximum intensity curves were calculated using Prism.

Any movie that showed a slight movement was aligned by ImageJ StackReg plugin before analysis.

The previously generated strain Rluv3-Gal4, UAS-EB1::GFP (Yalgin et al., 2015) was used to investigate the organization of MT minus ends in D. melanogaster class I da sensory neurons. Preparation of embryos was carried out as previously described (Yalgin et al., 2015). Briefly, stage-15 embryos were quickly de-chorionated, washed in water, and mounted with halocarbon oil with one layer of double-sided tape as a spacer. Imaging of D. melanogaster was performed with a FV1200 Laser Scanning Microscope (Olympus) equipped with a 473 nm laser, and a PLAPON 60XO NA1.42 (Olympus) objective. Time-lapse images were acquired at 5% laser power at 5x zoom, imaging three z-sections spaced 0.6 µms apart, with 2.4 s between frames, for four minutes. Kymographs were prepared using ImageJ using an average projection of imaged z-sections.