Morphological characterization of “swellings” and filopodia during PVD dendrite outgrowth
To identify distinct morphological events during the development of PVD dendritic arbors, we performed in vivo live imaging of the PVD dendrite at the third larval stage (L3) of wild-type worms. At this stage, the dendrite elaborates its tertiary and quaternary branches (see Additional files 1 and 4). Using a myristoylated-GFP (myr-GFP) marker expressed specifically in PVD, we observed many dynamic growth and retraction events that ultimately led to the formation of menorah-like branches, consistent with previous findings [18, 26]. In addition, we noticed that the growth of the quaternary branches often occurred through a series of rapid filopodia-like growth events, which were interspersed by several pauses (Fig. 1A-C). When we plotted the growth of quaternary dendrites over time (Fig. 1B, three independent examples) and examined dendrite morphology during growth events, we found that existing dendrites often exhibited a local increase in width (blue bars) right before the initiation of new branches or further extension of existing dendrites. Quantification of dendrite width showed a significant increase 1–2 min before new branch initiation (Fig. 1C), and filopodia emerged rapidly from these enlarged locations (Fig. 1B-C). We herein refer to these local increases in width as “swellings”. Using a LifeAct:GFP marker, we confirmed that F-actin is enriched in these swellings prior to initiation or elongation of the dendritic branches (Fig. 1D). These data suggest that the swellings are actin-based, which is consistent with previous reports that PVD dendrite growth is primarily driven by actin polymerization [23, 24]. Finally, we noticed that filopodia grew from the swellings in various directions, but only a subset of the filopodia were stabilized as quaternary dendrites (see Additional files 1 and 4). This behavior is consistent with the notion that filopodia are used to probe extracellular environments, perhaps in search of external guidance cues such as the patterned SAX-7 stripes along which PVD grows [27].
Filopodia formation during PVD dendrite outgrowth depends on unc-34 (Ena/VASP) and unc-115 (abLIM)
To identify the molecular mechanisms underlying the formation of the actin-rich swellings and filopodia observed above, we next examined PVD outgrowth in mutants of various actin regulators. We first focused on unc-34, the sole C. elegans Ena/VASP protein homolog, and unc-115, an actin-binding LIM protein (abLIM) previously implicated in neurite outgrowth. Ena/VASP proteins play a critical role in polymerizing linear F-actin to initiate and extend filopodia [28]. UNC-115 is thought to be a Rac effector that can act downstream of Netrin [9, 29] and has been shown to regulate axon guidance by affecting the dynamics of filopodia and lamellipodia in axonal growth cones [9, 30, 31]. Specifically, overexpression of membrane-targeted UNC-115 caused the formation of ectopic neurites, lamellipodia, and filopodia in the C. elegans PDE neuron, and deletion of the actin-binding VHD domain greatly reduced these effects [30]. We found that null mutants of unc-34 and null mutants of unc-115 both exhibited a strong loss of quaternary branches (Fig. 2A-B), suggesting that these actin regulators are also required for reorganizing the actin cytoskeleton during PVD dendrite morphogenesis.
We next used live animal imaging to examine if the unc-34 or unc-115 mutants affected swelling or filopodia formation during PVD dendrite outgrowth (Fig. 2C-I; Additional files 2 and 3). We found that in both unc-34 and unc-115 mutants, filopodia formation was drastically reduced, whereas dendrite swellings remained unaffected (Fig. 2E–F). Furthermore, lack of filopodia in unc-34 or unc-115 mutants switched the wild-type “rapid-growth-with-pauses” mode of dendrite extension to a slow and steady outgrowth mode characterized by reduced speed of remaining outgrowth events (Fig. 2D, I, Fig. S1), and less variability in the speed of such outgrowth (Fig. 2H). In particular, unc-34 mutants exhibited large, persistent growth cones at the tip of growing dendrites, which occasionally split to create new dendritic branches (Fig. 2C, Fig. S1, Additional file 2). These growth cones were wider than wild type growth cones (Fig. 2G) and extended steadily forward (Fig. 2H), reminiscent of the increased persistence of lamellipodia observed in migrating fibroblasts depleted of Ena/VASP proteins [32]. Taken together, our results suggest that both UNC-34 and UNC-115 play a key role in filopodia formation during PVD dendrite branching. The fast-growing filopodia ensure efficient growth of dendritic arbors, especially in the initiation of new quaternary branches. Because unc-34 and unc-115 mutants have also been reported to disrupt filopodia initiation in axonal growth cones [33], similar cytoskeletal mechanisms may also be involved in axon development.
Swelling formation during PVD dendrite outgrowth depends on the WAVE Regulatory Complex
Because neither unc-34 nor unc-115 mutants reduced swellings, we next asked what molecular mechanisms are required to form these structures. We previously showed that the WAVE Regulatory Complex (WRC) plays a critical role in PVD dendrite development through its direct interaction with the guidance receptor complex that mediates PVD outgrowth [23]. Consistent with our previous results, we found that mutants in the WRC components Sra1, Nap1/Hem-2, or WAVE/Scar (named gex-2, gex-3, and wve-1 in C. elegans, respectively) all showed severely truncated dendritic arbors at the L4 stage (Fig. 3A-B). All three WRC mutants exhibited a maternal effect lethal phenotype, where the offspring of heterozygous mothers are viable but sterile, suggesting that the alleles are indeed strong loss-of-function [34]. However, because the maternal WRC is clearly sufficient for the animals to develop to adulthood, it is possible that variation in the amount of residual WRC activity may account for slight differences in phenotype across the three mutants, such as the slightly weaker phenotype of the wve-1 mutant (Fig. 3B).
To test if the WRC is required to form swellings, we performed live animal imaging on gex-3 mutants and found that tertiary dendrites failed to form both swellings and filopodia (Fig. 3E–F; Additional file 5). As a result, we observed very few growth events overall (Fig. 3G-H; compare Additional files 4 and 5). We note that the gex-3 phenotype is distinct from the phenotypes of unc-34 and unc-115 mutants in that both swellings and filopodia were absent in gex-3 mutants, whereas only filopodia, but not swellings, were disrupted in unc-34 and unc-115 mutants. Since filopodia arise from swellings in wild-type animals, it is plausible that the WRC promotes the formation of swellings, which then determines the locations for filopodia growth driven by UNC-34 and UNC-115. A similar mechanism, termed the “convergent elongation” model, has previously been proposed to describe filopodia formation from branched actin networks [35]. In this model, actin filaments from a branched network that is bundled together to form nascent filopodia. This mechanism has been proposed to explain filopodia formation in ex vivo chick sensory axons and cultured primary neuron growth cones [36, 37].
The WAVE Regulatory Complex can directly bind to UNC-34 through PPR-EVH1 interactions
What mechanisms might coordinate the local WRC activity in swellings with the UNC-34 or UNC-115 activity in filopodia formation? One possibility is that the WRC recruits Ena/VASP proteins through a direct interaction between the poly-proline region (PPR) of Abi and WVE, and the EVH1 domain of Ena/VASP. Such an interaction was previously observed between human and Drosophila proteins, but the exact PPR sequences in Abi and WVE are divergent between various animal species [12].
To verify whether the UNC-34 EVH1 domain also directly binds to the C. elegans WRC (ceWRC), we recombinantly purified MBP (Maltose binding protein)-tagged full-length (FL) Abi and WVE, MBP-tagged fragments of Abi and WVE, and a GST-tagged EVH1 domain of UNC-34 for pull-down assays (Fig. 4A). Consistent with previous reports [12, 28], GST-UNC34 EVH1 retained both Abi FL and WVE FL (Fig. 4B), but not truncated proteins that lack the PPR sequences (WVE 178 and Abi 159, Fig. 4B). By mutating each of the three PPR sequences in WVE and Abi, which we predicted to bind to EVH1 using the “LPPPP” motif [12], we determined that multiple PPR sequences in both WVE and Abi contributed to the binding to EVH1, albeit at different levels. Particularly, deletion of WVE PPR#2, Abi PPR#3, and to a lesser extent Abi PPR#1 or PPR#2, clearly decreased the binding of UNC-34 EVH1 (WVE ΔPPR#2, Abi ΔPPR#1, 2 & 3; Fig. 4A, C). Note that although deleting WVE PPR#1 or #3 did not appear to reduce the binding, deletion of WVE PPR#2 still showed a residual, but clear interaction, suggesting PPR#1 and #3 remaining in the full-length WVE had a weak, but specific contribution to the binding. This result was further confirmed by direct binding between GST-UNC-34 EVH1 and isolated fragments containing individual WVE PPR#2 or Abi PPR#3 (Fig. 4A, D). In addition, we observed weak, but specific interaction of the isolated WVE PPR#1 (Fig. 4D). We could not detect binding of the isolated WVE PPR#3, Abi PPR#1 or #2, likely due to the limit of pull-down assays in detecting weak interactions of affinity in, empirically, tens of micromolar range (Fig. 4D). The slight discrepancy between full-length proteins, which contain multiple PPRs, and isolated PPRs likely reflects an avidity effect of multivalent interactions in sustaining a binding through otherwise weak individual ones.
Finally, we tested if UNC-34 EVH1 bound to the trimeric subcomplex of the WRC containing Abi, WVE, and HSPC300 (ce3mer; Fig. 4A). We used the trimeric subcomplex as a surrogate for the whole ceWRC pentamer because the subcomplex contained all PPRs found in the ceWRC pentamer, behaved well in biochemical reconstitution, and was less difficult to purify than the pentameric complex [23]. We found that in the context of the ceWRC subcomplex, entirely removing the PPR regions from both WVE and Abi abolished the binding to UNC34 EVH1, whereas mutating either WVE PPR#2, or Abi PPR#3, or both, partially reduced binding (Fig. 4E). This result further supports the notion that multiple WVE and Abi PPR sequences interact with UNC-34 EVH1. Although individual PPRs have different, sometimes weak, affinity (Fig. 4B-C), the availability of many PPRs in the trimer provides avidity (and possibly some cooperativity) to sustain a robust binding to EVH1, which could overcome deletion of the two PPRs with strongest affinity (Fig. 4E). Taken together, we conclude that the ceWRC can directly bind to UNC-34 through multiple PPR-EVH1 interactions.
The EVH1 domain of UNC-34 is necessary for UNC-34 localization and function in vivo
Having established that the ceWRC can bind to UNC-34 in vitro, we next asked whether the WRC-UNC-34 interaction is necessary for recruiting UNC-34 in PVD dendrites in vivo. To do so, we examined UNC-34 localization during PVD outgrowth by expressing an UNC-34:GFP fusion protein with a PVD-specific promoter in unc-34 null animals (Fig. 5A). We found that, consistent with the importance of Ena/VASP proteins in elongating actin bundles, and with a recent report describing UNC-34 localization in PVD [24], UNC-34:GFP was enriched as puncta at the tips of growing dendrites (Fig. 5B, F; Additional file 6). Remarkably, UNC-34:GFP became enriched at the swelling sites before new branches started to emerge (Fig. 5D-F; Additional file 6). In contrast, when we expressed an UNC-34 lacking the EVH1 domain (UNC-34ΔEVH:GFP, Fig. 5A), the localization at both the tips of growing dendrites and the swelling sites before new branch initiation was significantly reduced (Fig. 5C-F; Additional file 7). Both wildtype UNC-34:GFP and UNC-34ΔEVH:GFP exhibit a decrease in signal after branch initiation (Fig. 5D) as a result of photobleaching during the imaging session. Note that removing the EVH1 domain from UNC-34 did not completely diminish its enrichment at the dendrite tips or swelling sites, which could be due other sequences in UNC-34, such as the EVH2 domain, binding the enriched F-actin [38, 39]. This interaction could account for the slight increase in UNC-34ΔEVH:GFP signal at 1 min before branch initiation (Fig. 5D). Nevertheless, the reduced enrichment of UNC-34ΔEVH:GFP at the swelling sites suggests the WRC-UNC-34 interaction plays an important role in recruiting UNC-34 to initiate filopodia formation. Finally, using the same constructs expressed in unc-34 null animals with a PVD morphology marker, we asked whether the WRC-UNC-34 interaction was required for the ability of UNC-34 to promote PVD dendrite growth. We found that while the wild-type UNC-34:GFP construct significantly rescued quaternary branch formation, UNC-34ΔEVH:GFP showed no rescuing activity (Fig. 6A-B). Using live animal imaging, we found that the wild-type UNC-34:GFP rescue increased the occurrence of filopodia per dendrite, while UNC-34ΔEVH:GFP failed to do so (Fig. 6C-D; Additional files 8 and 9). In comparison, formation of swellings remained unchanged in all above unc-34 null and rescued animals (Fig. 6E). Taken together, the above data support a model in which the WRC recruits UNC-34 to the swellings through the PPR-EVH1 interactions, and UNC-34 in turn promotes filopodia formation to initiate dendrite branching.