Here, we established a genetic system to dissect the mechanisms of branch generation and plasticity of arborized neurons in C. elegans. To determine branching patterns, we imaged transgenic animals expressing cytoplasmic ser-2p::GFP (green fluorescent protein) or plasma membrane DES-2::GFP in the PVDs (table S1). The PVDs contained repetitive structural units reminiscent of multibranched candelabras or menorahs (Fig. 1A). Although the number of menorah branches varied, the menorahs appeared to develop in a stepwise manner from the L2 larva to the adult (fig. S1). The stereotypical menorah structure is likely to form a functional unit necessary for the PVDs mechanosensory activities.

Mutations in the cell fusion gene eff-1 (9, 10) affected the pattern of PVDs arborization, resulting in disorganized and hyperbranched phenotypes (Fig. 1B). Moreover, eff-1(ok1021) mutant animals showed reduced sensitivity to strong mechanical stimuli (53%, n = 106) (11). To characterize menorah disorganization in eff-1 mutants, we quantified the number of processes at different degrees of the branching order (primary to senary branches; Fig. 1C and fig. S2). The frequency of secondary and tertiary branching was doubled in the eff-1(hy21) mutant compared with wild type. The eff-1(hy21) mutant had a strong branching phenotype, whereas the oj55 mutant displayed a weaker effect, correlating with their respective epithelial fusion-defective phenotype (figs. S2 and S5). In wild-type menorahs, most sprouting of branches and bending of tertiary processes occurred at right angles to the branches of origin (Fig. 1A). In contrast, eff-1(hy21) mutant menorahs showed varying branching angles (Fig. 1B). We observed in the mutant a 10-fold increase in the number of branches sprouting from the secondary branch (Fig. 1D) and a 20-fold increase in the number of branches that erred and turned back 180° (Fig. 1E). These phenotypes suggest that EFF-1 sculpts and maintains right-angle nonoverlapping branches.

Cell-specific expression of eff-1 in the PVDs (des-2p::eff-1) partially rescued neuronal eff-1 phenotypes (Fig. 2C). In contrast, PVD patterning defects were not rescued by expression of eff-1 in the neighboring epidermal tissue (fig. S3; dpy-7p::eff-1). In eff-1(hy21) mutants expressing eff-1 in both neural and epidermal tissues (des-2p::eff-1+dpy-7p::eff-1), the rescue was not significantly stronger than the rescue observed with only the PVD-specific des-2p::eff-1 (fig. S3). These rescue experiments, together with expression of EFF-1::GFP in the PVDs (fig. S4), provide evidence that eff-1 controls branching cell autonomously. Moreover, eff-1 overexpression in the PVDs of wild-type animals reduced branching (Fig. 2D). The remaining branches were organized in a gradient starting from the cell body toward the head and tail, where no branches could be observed. Thus, eff-1 may play a role in mechanosensory neurons restricting branching in a dosage-dependent manner to produce dendrite simplification (Fig. 2G).

The absence of excess branching in wild-type animals may reflect a situation where only the appropriate branches initiate outgrowth. Alternatively, an excess of branches may be generated and extended but at some later point undergo retraction, pruning, or fusion to repair branching errors. To determine how EFF-1 restricts branching, we followed the behavior of branches by using three-dimensional (3D) live-imaging confocal microscopy in wild-type animals overexpressing des-2p::eff-1. The tips of tertiary to senary menorah branches first contacted the muscle cells, and then the dendrites detached from the menorahs by fission events (Fig. 3A; arrowheads) causing spontaneous dendrite break-off (Fig. 3B; arrowheads). Thus, following dynamic outgrowth, fission events eliminate extra branches.

To further analyze eff-1–dependent remodeling of menorahs, we grew the eff-1(hy21ts) worms at the restrictive temperature and shifted them to the permissive temperature in the early L4 stage. The defective arborization pattern was static at the nonpermissive temperature. Four hours after down-shifting to the permissive temperature, we observed branches that met, touched, and retracted. Some branches appeared to be stably connected, forming loops of different sizes and shapes (43 stable loops, n = 18; fig. S6). Because EFF-1 is a fusogen, we hypothesized that transient meeting and attachment of branches expressing EFF-1 may have resulted in loop formation because of interbranch fusion. To distinguish between interbranch fasciculation and fusion, we analyzed the 3D structure of the loops by generating confocal z stacks, projections, and rotations. The loops were stably connected over time (55 loops, n = 18 animals; movie S1). Next, we imaged DES-2::GFP in loops and found that fluorescent particles moved freely through them. To test whether the loops resulted from eff-1 activity, we incubated temperature-sensitive mutants at the restrictive temperature to generate hyperbranching, then we shifted them to the permissive temperature to induce expression of active EFF-1. After generating loops for 3.5 hours, we upshifted the worms back to the restrictive temperature to stabilize the loops (fig. S8). We obtained a two- to threefold increase in the number of loops (119 stable loops in n = 16 animals; Fig. 3C). We also observed symmetric fluorescence recovery of photobleached areas within the loops (fig. S7 and movie S2), indicating that the loops are continuous. Thus, eff-1 induces loop formation by neurite autofusion and/or fasciculation.

To further demonstrate that neurites can fuse with each other, we turned to higher resolution images of menorah branches and membranes. By using archival serial section transmission electron micrographs [TEMs (6)], we identified arborizations derived from the PVDs. In transverse sections of an adult hermaphrodite (N2U), 30- to 80-nm-diameter branches sandwiched between the body wall muscles and the hypodermis were observed (Fig. 3D, arrowheads). We reconstructed parallel neurites corresponding to the quaternary branches of a menorah from serial sections (Fig. 3E and figs. S9 to S11). In 10 examples of PVDs and FLPL/R (highly arborized neurons anterior to the PVDs), we found 2 to 12 branches fusing at the midline and forming fused longitudinal processes (Fig. 3, D and E, arrows). In no cases did we observe distal dendrites fasciculate; instead these fused if they reached one another. Thus, neurite-neurite auto-fusion plays a role in PVD and FLP arborization.

In addition to loop formation, dendrites were highly dynamic, erring, growing, and retracting in wild-type worms (Fig. 4A and movie S3). PVDs overexpressing eff-1 showed similar retractions but, unlike in the wild type, showed limited branch growth (Fig. 4, B and C). Similarly, when eff-1(hy21ts) animals were downshifted from the restrictive to the permissive temperature, we observed a 20-fold increase in the number of retracting branches and a 5-fold reduction in branch growth compared with eff-1 mutants at 25°C (Fig. 4, D to G, and movie S1). In contrast, upshifting eff-1(hy21ts) worms resulted in excess neurite growth (Fig. 4F). Thus, we propose a model in which EFF-1 autonomously induces retraction of branches to simplify menorahs.

EFF-1 is both a sculptor of epithelial organs by cell fusion (10) and a menorah sculptor by controlling dendrite bending, retraction, and fusion. The activities of this fusogen may be due to its ability to induce membrane curvature, a process that is thought to constitute a major driving force in membrane fusion and fission (12–14). Proteins capable of bending membranes, such as atlastins (15, 16) and dynamins (17), can induce tubulation, fusion, and fission (12, 13). Three mechanistic principles may form and maintain branched tubes in the cytoplasm and in extracellular branched filopodia or neuronal arbors such as menorahs: first, assembly of specialized proteins on membranes; second, membranous tube formation involving growth and bifurcation of tubes; and third, membrane bending followed by membrane fusion and fission restricts excessive branching. How can EFF-1 control mechanistically different processes such as dendrite fusion and retraction? Different isoforms and interactions may account for diverse activities. For example, trans interactions between EFF-1 on dendrites will cause autofusion, whereas assembly of large EFF-1 complexes on dendrites may induce actin-mediated retraction.

Supplementary Data