Regulation of dendrite growth and maintenance by exocytosis

Axons and dendrites lengthen by several orders of magnitude during neuronal development, and this lengthening of neurites is essential for establishing the morphological features of a neuron that ultimately determine its connectivity and function. The addition of membrane proteins and lipids through exocytosis plays a fundamental role in this process, but growth in axons and dendrites appears to be regulated by distinct pathways. In de novo axon outgrowth, addition of new membrane occurs primarily through exocytosis of plasmalemmal precursor vesicles (PPVs) at the growth cone (Bray, 1970; Craig et al., 1995; Lockerbie et al., 1991), which is mediated, in part, by soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins (Pfenninger, 2009). PPVs that are synthesized in the cell body are transported to growing axons (Pfenninger and Johnson, 1983), and although the source for material to support dendrite growth is unclear, the polarized localization of satellite secretory pathways in dendrites suggests that PPVs might be locally generated in some dendrites (Horton et al., 2005; Pfenninger, 2009; Ye et al., 2007). Although it has been demonstrated that disrupting post-Golgi trafficking can impede dendrite growth (Horton et al., 2005; Ye et al., 2007), little is known about the role that exocytosis specifically plays in dendritic morphogenesis.

Most intracellular membrane fusion events require both SNARE and Sec1/Munc18 (SM) proteins. SM-family proteins are essential components of the molecular machinery that regulates exocytosis and cooperate closely with SNARE proteins to mediate vesicle fusion (Sudhof and Rothman, 2009). Munc18-1 (also known as syntaxin binding protein 1, STXBP1), the neuron-specific isoform of the SM protein family, is a key regulator of neurosecretion and mediates synaptic vesicle release through its interactions with syntaxin-1 and the SNARE complex (Hata et al., 1993; Verhage et al., 2000). However, Munc18-1 protein is not restricted to axon terminals but is distributed throughout neurons, suggesting that Munc18-1 might have other roles in the neuronal membrane that are separable from its function in neurotransmitter release (Garcia et al., 1995).

In yeast, the SM protein Sec1 physically interacts with subunits of the exocyst to facilitate SNARE-complex assembly and membrane fusion upon vesicle arrival at the plasma membrane (Morgera et al., 2012; Wiederkehr et al., 2004). The exocyst is an evolutionarily conserved octameric protein complex that was originally identified in yeast based on its role in the spatial targeting and tethering of post-Golgi vesicles to specific sites of membrane expansion (Novick et al., 1980; TerBush et al., 1996). In the absence of a functional exocyst, secretory vesicles are delivered to the plasma membrane, but SNARE complexes do not form and exocytosis is blocked (Grote et al., 2000). Although the exocyst plays well-established roles in neuronal polarization and the regulation of neurite outgrowth (Hazuka et al., 1999; Lalli and Hall, 2005; Murthy et al., 2003; Vega and Hsu, 2001), a role in dendrite development has not been specifically attributed to the exocyst. Likewise, whether Sec1–exocyst complexes form in higher eukaryotes is unknown.

Drosophila class IV dendritic arborization (C4da) neurons elaborate highly complex dendritic arbors that undergo a period of extensive growth during larval development, during which time the linear dimensions of dendrites increase more than tenfold (Grueber et al., 2002; Parrish et al., 2009), necessitating enormous amounts of material to drive the expansion of the plasma membrane. C4da neurons therefore provide an excellent system to investigate how new material is supplied in dendrites to support their growth and maturation. Here, we identify Rop, the Drosophila homolog of vertebrate Munc18-1, as a key regulator of dendrite development. In Rop mutants, C4da neurons show an initial defect in outgrowth of terminal dendritic branches, followed by progressive loss of terminal dendrites and degeneration of major dendritic branches. We found that Rop mediates dendrite growth, in part, through interaction with the exocyst subunit Sec6. These Rop–Sec6 complexes are enriched in the primary dendrites of C4da neurons, suggesting that this dendritic compartment is the predominant site for dendritic exocytosis. Indeed, using a pH-sensitive in vivo reporter, we found that exocytosis predominates along major branches during dendrite arbor expansion and that the lateral diffusion of material from primary dendrites to terminal dendritic branches is likely to support terminal dendrite growth. Finally, we demonstrate that the balance between exocytosis and endocytosis is not only important for dendrite growth, but is also required for maintenance of the dendritic arbor.

To identify regulators of dendrite growth, we performed an RNA interference (RNAi)-based screen of phylogenetically conserved genes in class I da (C1da) neurons, which elaborate simple, stereotyped dendrites and are amenable to RNAi-based analysis of dendrite morphogenesis (Foley and O'Farrell, 2004; Parrish et al., 2006). From this screen, we found that double-stranded RNA (ds)RNA-mediated knockdown of Rop caused severe reductions in dendritic outgrowth of embryonic C1da neurons (data not shown). To validate the observed RNAi phenotype, we examined C1da neurons in embryos homozygous for an amorphic allele of Rop (Rop^G27) and found that C1da neurons in Rop^G27 homozygous mutant embryos exhibited growth defects comparable to those upon Rop knockdown by using RNAi, including severe reduction in outgrowth of both the primary dendrite and secondary lateral branches (Fig. 1A,B), demonstrating that Rop is required for overall dendritic arbor growth. To explore the generality of Rop function in dendrite growth, we assayed for requirements of Rop function in the development of the more complex dendritic arbors of C4da neurons. Similar to C1da neurons, C4da neurons exhibited severe deficits in the outgrowth of primary dendrites and terminal dendritic branches in Rop^G27 homozygous mutants, resulting in dramatic reductions in receptive field size (Fig. 1C,D). Thus, Rop appears to be generally required to support dendrite growth from early stages of dendrite development in the Drosophila embryonic peripheral nervous system (PNS). Consistent with Rop functioning in a cell-autonomous manner in dendrite growth, we found that Rop is highly expressed in multidendritic (md) sensory neurons, including all da neurons, during embryogenesis. However, Rop is also highly expressed in other cell types in the body wall, including the epidermis, suggesting that the function of Rop is not exclusive to neurons (Fig. 1E–G).

Rop^G27 mutants fail to develop normally beyond mid-embryogenesis, failing to produce cuticle and exhibiting abnormalities in the underlying epidermis (Harrison et al., 1994). Given the importance of non-autonomous epithelial-derived cues in influencing dendrite growth (Jiang et al., 2014; Parrish et al., 2009) and the broad distribution of Rop expression (Fig. 1E–G), we sought to assay for cell-autonomous functions of Rop in dendrite development. To this end, we performed mosaic analysis with a repressible cell marker (MARCM) to generate single-cell C4da neuron clones homozygous for the Rop^G27 allele in an otherwise heterozygous background and assayed for effects on dendrite growth (Lee and Luo, 1999). Using sensory organ precursor (SOP)-FLP lines to facilitate the generation of C4da MARCM clones and reduce the impact of heat shock on developmental timing (Shimono et al., 2014), we found that dendrite defects were first manifest at 72 h after egg laying (AEL) in Rop^G27 mutant C4da neuron clones (Fig. 2A). Although there was no statistical difference in the number of terminal dendritic branches or total dendrite length between Rop mutant and wild-type control neurons (Fig. 2B), a noticeable thinning of terminal dendrites in Rop mutant C4da neurons was apparent at this time point. By 96 h AEL, Rop^G27 C4da MARCM clones exhibited severely reduced dendritic complexity, as indicated by significant reductions in total dendrite length and in the total number of terminal dendritic branches (Fig. 2A,B). This reduction in complexity in the Rop mutant was not due to an overall loss of dendrites, as the numbers of terminal dendrites in Rop mutants did not differ significantly at 72 h and 96 h AEL, and is more likely to reflect a failure in dendritic branch formation. Varicosities along some of the major dendritic branches also become apparent by this time (Fig. 2A, arrows). By 144 h AEL, Rop^G27 C4da MARCM clones showed a profound loss of dendrites and extensive beading along the major dendritic branches. In extreme cases, Rop mutant neurons were completely devoid of terminal dendritic branches, and primary dendrites appeared highly fragmented, features that are characteristic of dendrite degeneration (Fig. 2C). We further monitored the effects of Rop mutations on axon growth by visualizing axon projection patterns of individual da neuron MARCM clones. Although Rop^G27 C4da MARCM clones exhibited severe defects in dendritic morphology at 96 h AEL, there was no obvious abnormality in the axon projections of Rop mutant da neuron clones at this time point (data not shown). However, axons of Rop mutant neurons appeared thin with varicosities along the terminals by 144 h AEL (Fig. S1), suggesting that Rop might also be required for some aspects of axon development.

The dendrite phenotypes we observed could reflect a role for Rop in dendrite growth or could be a secondary effect of cell death caused by loss of Rop function. To differentiate between these two possibilities, we stained Rop^G27 C4da MARCM clones with markers of cell death and found that Rop mutant clones showed no detectable signal for the apoptotic effector Death caspase-1 (Dcp-1), and excluded the vital dye Acridine Orange (Fig. S1; data not shown). To further establish the requirement for Rop function in dendrite patterning, we monitored the effects of a second Rop allele (Rop^A19) on dendrite growth. Similar to what we observed with the Rop^G27 allele, C4da MARCM clones homozygous for the hypomorphic Rop^A19 allele showed significant reductions in total dendrite length and in the total number of terminal dendritic branches (Fig. 2D,E). Finally, we found that expressing an epitope-tagged version of Rop was sufficient to rescue the dendrite growth defects of Rop mutants (Fig. 2D,E), further demonstrating that Rop functions in a cell-autonomous manner in C4da neurons to promote dendrite growth.

Given the severity of the dendritic growth defect in Rop^G27 homozygous mutant embryos, we examined whether Rop protein perdurance contributed to the delayed phenotype of Rop^G27 mutant MARCM clones. Indeed, immunoreactivity of Rop was detectable in Rop^G27 mutant C4da MARCM clones at 72 h AEL, when dendrite defects first become apparent, whereas immunoreactivity of Rop was dramatically reduced in Rop^G27 mutant C4da MARCM clones at 120 h AEL (Fig. S2). Similarly, perdurance of maternal Rop function masks embryonic phenotypes in zygotic Rop mutants (Harrison et al., 1994). Thus, the late onset of dendrite defects is likely to reflect the timing of Rop activity exhaustion in Rop^G27 MARCM clones, with persistent Rop protein supporting dendrite growth up to 72 h AEL. Interestingly, the growth defects were not uniformly distributed in dendrite arbors of Rop^G27 mutant C4da clones – terminal dendrites exhibited heightened sensitivity to Rop function, suggesting that terminal dendrites have an increased requirement for exocytosis compared to other regions of the arbor, that exocytic membrane addition preferentially occurs at terminal dendrites, or both. To further explore the generality of the role of exocytosis in dendrite growth, we monitored the effects of RNAi-mediated knockdown of a subset of SNARE proteins that are important for vesicle trafficking and fusion, including Syntaxin1A, Snap24, Snap25, α-Snap and Synaptotagmin 1. Knockdown of these components led to significant reductions in terminal branching (Fig. S3), consistent with the notion that terminal dendrites have a heightened requirement for the exocytic machinery. Taken together, our results indicate a role for the exocytic machinery in dendritic morphogenesis and demonstrate a cell-autonomous requirement for Rop in the growth and subsequent maintenance of the dendritic arbor.

We next set out to identify genes that act with Rop to control dendritic morphogenesis, focusing our attention on the possible relationship between Rop and the exocyst, which regulates polarized trafficking of membranous material. To determine whether the exocyst is required in a cell-autonomous manner for dendrite morphogenesis, we used MARCM to generate C4da neuron clones that were homozygous mutant for loss-of-function alleles of two core components of the exocyst complex – Sec5 (Sec5^E10) and Sec6 (Sec6^Ex15) (Murthy et al., 2003, 2005). We found that both Sec5^E10 and Sec6^Ex15 C4da MARCM clones showed significant reductions in the total numbers of terminal dendritic branches as well as total dendrite length (Fig. 3A–E), demonstrating a cell-autonomous function of the exocyst in dendrite growth. Furthermore, consistent with Sec5 and Sec6 functioning together in dendrite development, Sholl analysis revealed a remarkable phenotypic similarity between C4da neurons deficient for Sec5 or Sec6 (Fig. 3F). Of note, a comparable reduction in the number of dendritic intersections was observed in Rop^G27 C4da MARCM clones (data not shown). Outgrowth of the major dendritic branches was only moderately affected in the exocyst mutants because the primary dendrites of Sec5^E10 and Sec6^Ex15 C4da MARCM clones extended to the dorsal midline and segment borders (Fig. 3B,C). However, varicosities along portions of major dendrites and thinning of terminal dendrites were apparent in Sec5 and Sec6 mutant clones (Fig. 3B,C), a phenotype reminiscent of Rop^G27 C4da MARCM clones.

Based on the phenotypic similarity of Rop and Sec5 or Sec6 mutants, we investigated whether Rop and the exocyst function together in dendrite development. First, we determined whether Rop interacts genetically with Sec5 and/or Sec6 in the control of dendrite morphogenesis. Dosage-dependent genetic interactions are often indicative of a shared function in C4da neuron dendrite development (Parrish et al., 2007), therefore we assayed for dendrite growth defects in larvae with trans-heterozygous combinations of Rop^G27 with the Sec5^E13 or Sec6^Ex15 allele (Fig. 3G–L). Indeed, trans-heterozygous combination of Rop^G27 with Sec5^E13 resulted in reduced numbers of terminal dendritic branches compared to C4da neurons in Rop^G27/+ and Sec5^E13/+ heterozygotes (P=0.06; Sec5^E13/+; Rop^G27/+ trans-heterozygotes versus Sec5^E13 heterozygotes). We also found that C4da neurons in Sec6^Ex15/+; Rop^G27/+ trans-heterozygotes exhibited significant reductions in terminal dendritic branching compared to C4da neurons in Rop^G27/+ and Sec6^Ex15/+ heterozygotes alone (P<0.05) (Fig. 3M). Thus, Rop genetically interacts with Sec5 and Sec6 to regulate dendrite development.

Because Rop genetically interacts with genes encoding exocyst subunits to regulate dendritic branching, we next tested whether Rop can physically associate with individual exocyst subunits in vivo. We performed co-immunoprecipitation experiments using a monoclonal antibody against Rop and found that Rop co-immunoprecipitated with endogenous Sec6 from adult Drosophila whole-brain extracts, as revealed by western blotting with an antibody against Sec6 (Fig. 3N). We further found that Rop co-immunoprecipitated with endogenous Sec15 from whole-brain extracts (Fig. 3N). Thus, Rop forms complexes with exocyst subunit proteins in vivo. Taken together, these data suggest that Rop can associate with multiple subunits of the exocyst complex, and that Rop and exocyst genes are likely to function in the same pathway to control dendritic morphogenesis.

During larval development, terminal dendrite growth and addition accounts for the majority of C4da dendrite growth (Fig. 4A,B). Our finding that mutations in Rop and exocyst subunits preferentially affected growth and maintenance of terminal dendrites during larval development suggests that growing terminal dendrites might be the primary target for exocytosis during neuronal growth. If this is the case, we reasoned that the exocytic machinery would be preferentially localized to terminal dendrites. To address this possibility, we monitored the endogenous distribution of Rop in larval C4da neurons using a monoclonal antibody to Rop (Harrison et al., 1994). Although we observed Rop immunoreactivity in major dendritic branches, the Rop signal was not detectable or much weaker at terminal dendrites of C4da neurons (Fig. 4C–E). However, Rop immunoreactivity was present throughout the body wall, precluding high-resolution analysis of Rop distribution in terminal dendrites. To circumvent this problem, we expressed an epitope-tagged version of Rop (Myc–Rop) in C4da neurons. As with endogenous Rop, epitope-tagged Rop was enriched in major dendrites; a weak signal was present in some higher order dendrites but rarely extended the entire length of terminal dendrites (Fig. 4F–H). We conclude that the distribution of Rop is not uniform in C4da dendrite arbors and appears to be predominantly localized to major dendrites.

It is possible that the exocyst spatially restricts the localization of assembled SNARE complexes (Carr et al., 1999; Grote et al., 2000; Wiederkehr et al., 2004), and in this manner, tethers Rop to specific sites of exocytosis. We reasoned that if the exocyst accumulates at terminal dendrites, then it might efficiently recruit Rop to terminal dendrites, even though the majority of Rop is restricted to major dendrites. To test this possibility, we monitored exocyst distribution in C4da neurons by co-expressing a GFP-tagged version of Sec15, which has been previously used to visualize the exocyst (Jafar-Nejad et al., 2005), in C4da neurons also expressing the CD4–tdTomato membrane marker. Interestingly, we observed that Sec15–GFP is enriched in the soma, where it accumulates in large punctate structures, and is also distributed along major dendrites with notable accumulations at dendritic branch points (Fig. 4I–K). As with Rop, Sec15–GFP was present at much lower levels in terminal dendrites. Therefore, our data suggest that the exocytic machinery is enriched in major dendrites.

The distribution of Rop and exocyst components in C4da dendrites suggests that dendritic exocytosis occurs primarily along the major dendritic branches rather than at terminal dendrites. To test this possibility, we used two imaging-based approaches. First, we reasoned that visualization of Rop–exocyst complex formation in vivo would allow us to identify sites of vesicle–plasma-membrane attachment. To this end, we used bimolecular fluorescence complementation (BiFC), in which two interacting proteins are fused to two non-fluorescent fragments of yellow fluorescent protein (YFP), the N-terminal (NYFP) and C-terminal (CYFP) fragments (Gohl et al., 2010; Hu et al., 2002). In this assay, binding of the interacting proteins facilitates YFP reconstitution, thus YFP fluorescence serves as a proxy for tracking protein–protein interactions in vivo. Reassembly of split YFP is irreversible in vitro (Hu et al., 2002), so BiFC should provide a cumulative record of Rop–exocyst complex formation and hence facilitate visualization of complexes throughout the cell.

Yeast Sec1 and Sec6 can directly interact in vitro (Morgera et al., 2012), and we found that Rop and Sec6 co-immunoprecipitate in Drosophila neurons (Fig. 3N), therefore we examined whether Rop and Sec6 directly interact in Drosophila S2 cells. First, as a negative control, we co-expressed the two halves of split YFP (CYFP and NYFP) in S2 cells and observed no detectable YFP fluorescence (Fig. S4), demonstrating that interacting proteins are required for efficient YFP reassembly. Next, we generated both N- and C-terminal fusions of Rop with NYFP, and N- and C-terminal fusions of Sec6 with CYFP, and tested for YFP reconstruction in S2 cells. A strong BiFC signal was apparent only when NYFP–Rop was transfected with Sec6-CYFP, demonstrating the specificity of the interaction (Fig. S4). Thus, we conclude that Rop and Sec6 can directly interact in Drosophila S2 cells.

Using NYFP–Rop and Sec6–CYFP, we next monitored complex formation between Rop and Sec6 in the Drosophila larval body wall. First, we assayed for in vivo function of NYFP–Rop and Sec6–CYFP. We expressed UAS-NYFP-Rop in Rop mutant C4da MARCM clones and UAS-Sec6-CYFP in Sec6 mutant C4da MARCM clones, and found that both transgenes rescued the respective mutant phenotypes (Fig. 2D,E; Fig. S4). These results indicate that NYFP–Rop and Sec6–CYFP are functional proteins, and suggest that the BiFC signal is likely to reflect the normal localization of Rop–Sec6 complexes in vivo. Sec6 complexes localize to basolateral junctional domains in polarized epithelial cells where they appear to mediate polarized delivery of cargo (Grindstaff et al., 1998), therefore we next examined whether the Rop–Sec6 BiFC signal accumulates at junctional domains in the larval epidermis. Indeed, co-expression of NYFP–Rop with Sec6–CYFP using the epidermal A58-Gal4 driver yielded robust YFP reconstitution in body wall epithelial cells, with Rop–Sec6 complexes enriched at junctional domains (Fig. S4). Finally, we assayed for Rop–Sec6 complex formation in C4da neurons. When we co-expressed NYFP–Rop and Sec6–CYFP in C4da neurons, we observed strong YFP fluorescence in major dendrites but little or no signal in terminal dendrites (Fig. 5A,B), suggesting that the Rop–exocyst interaction, and hence vesicle tethering to the plasma membrane, primarily occurs in major but not terminal dendrites. In contrast to the dendritic distribution of Rop–Sec6 complexes, co-expressing NYFP–Rop and Sec6–CYFP yielded robust YFP reconstitution in C4da axon terminals (Fig. 5C), consistent with previous reports that the exocyst is required for exocytosis of Golgi-derived vesicles in axon terminals of Drosophila motor neurons (Murthy et al., 2003, 2005). Thus, the Rop–Sec6 complexes are differentially distributed in different compartments within C4da neurons – they are enriched in axon terminals and major dendrites but present only at low levels in terminal dendrites. These results suggest that the capacity for exocytosis within C4da neurons is not uniform.

To more directly visualize sites of exocytosis and function, we utilized a pH-sensitive reporter (pHluorin-CD4-tdTomato; referred to as pH-Tom) comprising pH-sensitive GFP (ecliptic pHluorin) fused to the N-terminus of a red fluorescent membrane marker (CD4–tdTomato) (Kanamori et al., 2015) (Fig. 6A). The chromophore of ecliptic pHluorin is protonated at acidic pH (Miesenböck et al., 1998), hence pHluorin fluorescence is quenched within acidified intracellular compartments. As a result, the pH-Tom reporter fluoresces as both green and red at the cell surface but loses green fluorescence in acidified intracellular compartments, such as in exocytic vesicles. In cells expressing this reporter, regions with low levels of exocytosis should have reduced levels of green fluorescence compared to regions with high levels of exocytosis.

When we constitutively expressed pH-Tom in C4da neurons, beginning during late embryogenesis, we observed comparable ratios of green and red fluorescence throughout dendrite arbors in third instar larvae (Fig. 6B–D). However, over extended timescales, endocytosis and lateral diffusion might obscure regional differences in exocytosis as reported by pH-Tom. We therefore used a Flp-recombinase-inducible ppk-Gal4 driver (Kanamori et al., 2013) to inducibly and selectively express pH-Tom in C4da neurons, reasoning that by tracking newly synthesized pH-Tom, we could more accurately monitor patterns of exocytic vesicle delivery. Using this approach, we found that the pH-Tom reporter accumulates to detectable levels approximately 24 h after induction of Flp expression. We therefore induced Flp expression at 48 h AEL and monitored pH-Tom fluorescence in C4da dendrites at 72 h AEL, the time at which dendrite defects were first apparent in Rop mutant C4da clones (Fig. 6E). Similar to what we observed following constitutive expression of the reporter, green (pHluorin) and red (Tomato) fluorescent signals were readily detectable in major dendrite branches (Fig. 6F–H). However, compared to major dendrites, the ratio of green:red fluorescence was significantly reduced in terminal dendrites, suggesting that exocytosis preferentially occurs along major dendrites (Fig. 6H). Alternatively, these results could reflect heightened rates of endocytosis in terminal dendrites. However, we observed a similar distribution of pH-Tom fluorescence when we simultaneously induced pH-Tom expression and blocked endocytosis using a temperature-sensitive dominant-negative allele of shibire (shi^ts1) (Kitamoto, 2001) (Fig. 6H), suggesting that the absence of pHluorin fluorescence in terminal dendrites is not a consequence of endocytosis. Thus, although the pH-Tom reporter readily accumulates in terminal dendrites, pH-Tom is more efficiently delivered to the plasma membrane in major dendrites.

Our results thus far indicate that exocytosis occurs primarily along major dendrites, whereas the majority of dendrite growth occurs in terminal dendrites. We therefore investigated the possibility that material delivered to major dendrites might be supplied to terminal dendrites through diffusion to facilitate growth. If terminal dendrites rely on diffusion for materials to support membrane expansion, then diffusion of materials into terminals needs to occur at a faster rate than diffusion out of terminals. To examine this possibility, we monitored the dynamics of a membrane-associated protein in dendrites by measuring the lateral diffusion of a myristoylated photoactivatable fluorescent protein (myr-Eos) after photoconversion. Specifically, we photoconverted myr-Eos in major dendrites and monitored diffusion of the photoconverted protein in major dendrites and into terminal dendrites (Fig. 7A,B), or photoconverted myr-Eos in terminal dendrites and monitored diffusion within the terminals and into major dendrites (Fig. 7C). Using this approach, we observed robust lateral diffusion of myr-Eos within major dendrites or within terminal dendrites, modest diffusion from major dendrites into terminals, and very little diffusion from terminals into main dendrites (Fig. 7B,C). Although the photoconverted myr-Eos is likely to comprise both cytosolic and membrane-bound forms of the fluorescent protein, a cytosolic version of Eos that is not membrane-bound freely diffused between the two compartments (Fig. 7B,C), suggesting that the directional bias in myr-Eos diffusion is primarily due to its association with the membrane. We next tested the possibility that mutations in Rop affect the dynamics of membrane proteins by monitoring diffusion of a membrane-targeted GFP (mCD8–GFP) in both wild-type and Rop^G27 C4da MARCM clones using fluorescence recovery after photobleaching (FRAP). Notably, when we photobleached mCD8–GFP in terminal dendrites and assayed recovery from mCD8–GFP in major dendrites, we observed no difference between control and Rop mutant clones (Fig. 7D). Likewise, the recovery half-time and the percentage of fluorescence recovery in major dendrites were unchanged in Rop mutant neurons compared to that of wild-type controls (Fig. 7D), suggesting that lateral diffusion of membrane-associated proteins is not impaired when exocytosis is attenuated. Taken together, these results are consistent with the hypothesis that material delivered to primary dendrites by exocytosis diffuses into terminals to support terminal dendrite growth.

At first glance, our findings that terminal dendrites have a heightened requirement for exocytic machinery and that exocytic machinery and activity are concentrated in primary dendrites seem at odds with one another. However, dendrites in many systems exhibit high rates of endocytic activity, and the balance of exocytosis and endocytosis is likely to be a crucial factor in ensuring the proper development of the dendritic arbor (Blanpied et al., 2002; Parton et al., 1992). We therefore hypothesized that an imbalance in exocytosis and endocytosis contributes to the growth defect of terminal branches in Rop mutants. If this is the case, we reasoned that blocking endocytosis might partially limit the terminal dendrite growth defects caused by depleted Rop function. To test this possibility, we selectively blocked endocytosis in C4da neurons by expressing shi^ts1 in control and Rop^G27 MARCM clones. Larvae with C4da MARCM clones were shifted to the non-permissive temperature for shi^ts1 (30°C) at 48 h AEL, before the onset of dendrite growth defects in Rop mutants, and C4da neuron clones were visualized at 96 h AEL, when dendrite growth defects are broadly manifest in Rop mutant C4da neurons. Thus, shi was selectively inactivated in these clones from 48 h to 96 h AEL.

Under this temperature regime, we observed a modest decrease in dendrite branching of control neurons (compare Fig. 8E to Fig. 2B), reflecting a mild temperature sensitivity of the branching program, but little effect on overall dendrite length. Rop mutant clones still exhibited characteristic defects, including significant decreases in total dendrite length and an abundance of varicosities along major dendrites (Fig. 8B,E). Blocking shi function from 48 h to 96 h AEL reduced overall dendrite growth to a similar extent (Fig. 8C,E), consistent with a previous report that constitutively blocking shi function compromises dendrite growth (Yang et al., 2011). In addition, blocking shi resulted in a significant increase in dendrite–dendrite crossing events, suggesting that endocytic signaling is required for dendrite–dendrite avoidance (Fig. 8C,G). Interestingly, we found that blocking shi function from 48 h to 96 h AEL significantly ameliorated the dendrite defects of Rop^G27 C4da MARCM clones, leading to an overall increase in the number of terminal dendrites and dramatically reducing the appearance of varicosities along the major dendrites (Fig. 8D–F). Thus, endocytosis in the absence of exocytosis is likely to be a major contributor to the terminal dendrite growth defect and degeneration of primary dendrites in Rop mutant neurons. We also noted a synergistic effect on dendrite–dendrite crossing defects (Fig. 8D,G), demonstrating that endocytic–exocytic signaling plays crucial roles in dendrite–dendrite avoidance. Altogether, these results demonstrate that the balance of endocytosis and exocytosis is crucial in supporting dendrite growth and maintenance, and that these types of membrane fusion events are differentially localized during neuronal maturation to ensure the correct construction of neuronal architecture.

The plasma membrane of the developing neuron must continuously expand to keep pace with the rapid growth of the axon and dendrite. Although the addition of neuronal membrane is mediated by exocytosis, there is little in vivo evidence to support the requirement of SM proteins in this process. Our study not only reveals an autonomous function for the SM protein Rop and the exocyst in dendrite growth, but also reveals a differential requirement for the exocytosis machinery in the development of different compartments of the dendritic arbor. Depletion of Rop in C4da neurons causes an initial defect in terminal dendrites followed by profound degeneration of primary dendrites. Unexpectedly, we found that although Rop and the exocyst are required for terminal dendrite growth, Rop, the exocyst components Sec15 and Sec6, and Rop–Sec6 complexes are largely excluded from terminal branches. Likewise, we demonstrate that exocytosis predominantly occurs along major dendrites in these neurons, despite the fact that terminal dendrites account for the majority of dendrite growth in larval C4da neurons.

Our results show that terminal dendrites are likely to depend on the diffusion of material from primary dendrites to support their growth and, in the process, reveal key differences in how membrane is allocated in different compartments of the dendritic arbor. Primary and terminal dendrites of C4da neurons differ in their cytoskeletal composition and dynamics (Grueber et al., 2002; Jinushi-Nakao et al., 2007; Lee et al., 2015), which can create differences in membrane tension (Dai and Sheetz, 1999). As membrane tends to flow from a region of low tension to a region of higher tension (Dai and Sheetz, 1995), the resulting tension gradient could then explain how membrane proteins readily diffuse from major branches into terminals, but not in the opposite direction. Why then are terminal dendrites particularly sensitive to reduced Rop function? Growing dendrites exhibit high rates of endocytic activity compared to growing axons (Parton et al., 1992; Ye et al., 2007), and dynamic clathrin assembly and disassembly occurs throughout the dendritic arbor, consistent with a requirement for clathrin-dependent endocytosis in dendrite growth (Blanpied et al., 2002; Yang et al., 2011; Zheng et al., 2008). We found that blocking endocytosis significantly ameliorated the terminal dendrite loss observed in Rop mutant neurons, suggesting that unchecked endocytosis contributes to the rapid retraction of terminal dendrites in Rop mutants. Thus, even though terminal dendrites show low levels of exocytosis compared to primary dendrites, it is necessary to counterbalance the high levels of endocytic activity and replenish the plasma membrane, underscoring the delicate balance of exocytosis and endocytosis that is required for dendrite growth (Fig. 8H).

Although our data suggest that lateral diffusion of material contributes to terminal dendrite growth, we cannot exclude the possibility that satellite secretory pathways (or Golgi outposts) can act as a local source of membrane within some terminal branches (Horton et al., 2005; Ye et al., 2007). The presence of Golgi outposts within terminal branches of C4da neurons would seem to suggest that local exocytosis is an important mode of terminal dendrite growth, but recent studies support microtubule nucleation as a key role for Golgi outposts within these branches (Ori-McKenney et al., 2012). Therefore, we propose a model in which primary dendrites are the major sites of exocytosis and membrane addition for expansion of the dendritic arbor, and in which terminal dendrites primarily rely on the diffusion of material to support their growth (Fig. 8H).

In mature dendrites, exocytosis primarily occurs in dendritic spines that are enriched with clusters of the target (t)-SNARE syntaxin 4, which mediates activity-dependent exocytosis during long-term potentiation (Kennedy et al., 2010). By contrast, how sites of exocytosis are defined in a growing dendrite remains largely unresolved. The restricted localization of Rop in primary dendrites of C4da neurons suggests that the exocytic machinery is likewise targeted to specific sites of membrane expansion during development. Consistent with this notion, Rop interacts with exocyst subunits Sec5 and Sec6 to regulate dendrite growth, demonstrating that the exocyst–Sec1 interaction is functionally relevant in higher eukaryotes and could be an important mechanism by which SNARE complexes are recruited to specific sites of membrane expansion. In our BiFC analysis, we found that Rop directly interacts with Sec6 and that these complexes accumulate at sites with high exocytic activity, reflecting the importance of this interaction in biologically relevant contexts. In epithelial cells, we found that Rop–Sec6 complexes accumulate at junctional domains, consistent with the localized recruitment of the exocyst to regions of the epithelial membrane undergoing active exocytosis (Grindstaff et al., 1998), and suggesting that BiFC between Rop and Sec6 can be used as a proxy for exocytic activity in neurons. Indeed, we found that Rop–Sec6 complexes accumulate along major dendritic branches in C4da neurons, which correlate with the primary sites of exocytosis, as revealed by the in vivo pH-Tom reporter. Thus, we propose that the exocyst actively recruits Rop to specific sites of membrane expansion in dendrites in order to promote growth.

We also observed the accumulation of Rop–Sec6 complexes in C4da axon terminals. This result, together with the axon phenotype that we observed in Rop mutant neurons, points to a role for Rop in vesicular trafficking in axon growth and/or maintenance, in addition to its well-characterized role in synaptic vesicle release. The exocyst plays a key role in axon outgrowth and discriminates between vesicular trafficking mechanisms that regulate membrane insertion and cell growth, and those that regulate neurotransmitter secretion (Murthy et al., 2003). The exocyst itself is dispensable for neurotransmitter release (Mehta et al., 2005; Murthy et al., 2003), suggesting that binding of Rop to the exocyst can differentially affect Rop-mediated vesicular trafficking events in the presynaptic terminal.

Our finding that Rop^G27 C4da MARCM clones show a progressive loss of terminal dendrites and blebbing and/or fragmentation of primary dendrites during later stages of larval development further suggests that Rop is not only important for growth, but is also broadly required for dendrite maintenance. The neurodegeneration in Rop mutant neurons parallels observations in neurons from munc18-1 knockout mice, which show severe and widespread neurodegeneration (Heeroma et al., 2004; Verhage et al., 2000). Importantly, degeneration in munc18-1 null neurons cannot be fully explained by the loss of synaptic activity because mice deficient for both munc13-1 and munc13-2, which are synaptically silent, show no apparent signs of neurodegeneration (Varoqueaux et al., 2002). The function of Munc18-1 in neuronal maintenance is therefore likely to entail a function that is distinct from its role in neurotransmitter secretion. The ability of a dominant-negative allele of shibire to mitigate dendrite degeneration phenotypes of Rop mutant neurons suggests that a failure in exocytosis of vesicles derived from dynamin-dependent pathways is part of the mechanism by which Rop depletion causes degeneration of primary dendrites. Indeed, endocytic dysfunction is a common theme in neurodegeneration and is likely to underlie the pathology of many neurodegenerative disorders (Nixon, 2005). Given the strong association of Munc18-1 with syntaxin-1 (Sudhof and Rothman, 2009), it is not surprising then that a recent study also showed that expression of a dominant-negative form of dynamin could delay neurodegeneration induced by botulinum toxin-mediated cleavage of syntaxin-1 in cultured hippocampal neurons (Peng et al., 2013). Syntaxin-1 is therefore important for neuronal maintenance, but an early requirement for syntaxin-1 in neurite outgrowth (Igarashi et al., 1996) is consistent with the differential requirement for the exocytic machinery in the developing neuron.

Our findings are particularly interesting in light of the fact that mutations in MUNC18-1 and exocyst subunit genes are associated with a spectrum of neurodevelopmental disorders. De novo mutations in the gene encoding MUNC18-1 have been identified in individuals diagnosed with Ohtahara syndrome (Saitsu et al., 2008). In addition to intractable seizures, individuals with mutations in MUNC18-1 further exhibit profound intellectual disability, physical deficits and varying degrees of cerebral atrophy (Saitsu et al., 2008). Although disruption of MUNC18-1 is most closely associated with Ohtahara syndrome, de novo mutations in MUNC18-1 have also been identified in individuals with severe intellectual disability and non-syndromic epilepsy, individuals with non-syndromic intellectual disability, and individuals with Rett syndrome (Hamdan et al., 2011, 2009; Olson et al., 2015; Romaniello et al., 2015). Likewise, recent studies have revealed possible connections between mutations in exocyst genes and neurological disorders (Evers et al., 2014; Frühmesser et al., 2013; Wen et al., 2013). Given that dendrite defects are the strongest pathological correlate of intellectual disability (Kaufmann and Moser, 2000), we speculate that dendritic abnormalities might contribute to intellectual disability that is associated with MUNC18-1 and exocyst dysfunction.

The following lines were used: Gal4²²¹ (Grueber et al., 2003), ppk-Gal4 (Grueber et al., 2007), A58-Gal4 (Galko and Krasnow, 2004), ppk-Frt-stop-Frt-Gal4 (Kanamori et al., 2013), UAS-pHluorin-CD4-tdTomato (Kanamori et al., 2015), E7-2-36 (Brewster and Bodmer, 1995), SOP-FLP (Shimono et al., 2014), UAS-myrEos (Pfeiffer et al., 2010), ppk-CD4-tdTomato, UAS-CD4-tdGFP and UAS-CD4-tdTomato (Han et al., 2011), UAS-shi.K44A (Moline et al., 1999), and UAS-Dicer-2 (Dietzl et al., 2007). Rop^G27 and Rop^A19 (Harrison et al., 1994), Sec5^E10 and Sec5^E13 (Murthy et al., 2003), Sec6^Ex15 (Murthy et al., 2005), Sec6^Δ20 (Beronja et al., 2005), and UAS-GFP::Sec15 (Jafar-Nejad et al., 2005) have been described previously. RNAi lines were generated by the Transgenic RNAi Project at Harvard Medical School (Ni et al., 2009). UAS-NYFP-Rop and UAS-CYFP-Sec6 transgenic lines were generated by site-specific integration into attP40 and attP2 docker lines, respectively.

Rop^G27, FRT2A/TM6B, Rop^A19, FRT2A/TM6B or w; FRT2A were mated with 109(2)80-Gal4, UAS-mCD8::GFP, SOP-FLP; tub-Gal80, FRT2A females. Sec5^E10, FRT40A/CyO or w; FRT40A males were mated with elav-Gal4, UAS-mCD8::GFP, hsFLP; tub-Gal80, FRT40A females, and FRT42D, Sec6^Ex15/CyO or w; FRT42D males were mated with elav-Gal4, UAS-mCD8::GFP, hsFLP; FRT42D, tub-Gal80 females. For SOP-FLP-bearing crosses, embryos were collected for 2 h in standard cornmeal molasses agar and allowed to develop for 72, 96 or 144 h, at which time MARCM clones were imaged live using confocal microscopy. For hsFLP-bearing crosses, embryos were collected for 2 h on grape juice agar plates and allowed to develop for 3 to 5 h at 25°C before being heat-shocked for 1 h at 38°C. Embryos were then allowed to develop to late third instar larvae, at which point they were dissected and processed for immunocytochemistry.

Embryos and larvae were processed for immunocytochemistry, as previously described (Kim et al., 2006), using the following antibodies: mouse anti-Rop 4F8 at 1:200 (Developmental Studies Hybridoma Bank) (Harrison et al., 1994), rabbit anti-Myc at 1:1000 (A-14, Santa Cruz Biotechnology), rabbit anti-β-galactosidase at 1:5000 (Cortex Biochem), rat anti-mCD8 at 1:200 (Life Technologies), rabbit anti-Dcp-1 at 1:500 (Cell Signaling Technologies), Cy5-conjugated anti-horseradish-peroxidase at 1:250 (Jackson Laboratories), and appropriate fluorophore-conjugated secondary antibodies at 1:200 (Jackson Laboratories). Whole-mount embryos and image stacks of MARCM clones were acquired at 1024×1024-pixel resolution using either a Leica TCS SP2 or an Olympus FluoView FV1000 confocal microscope.

At the appropriate time, a single embryo or larva was mounted in 90% glycerol and imaged on a Leica SP5 confocal microscope with a 40×1.25NA lens. MARCM clones were imaged at 1024×1024-pixel resolution. For late time points, image stacks from multiple fields of view were assembled into a montage. For pHluorin imaging, larvae of the genotype ppk-gal4/UAS-pHluorin::CD4-tdTomato were imaged at 96 h AEL, and larvae of the genotype UAS hsFLP/+; ppk-Frt-Stop-Frt-Gal4/UAS-pHluorin::CD4-tdTomato; UAS-pHluorin::CD4-tdTomato/+ were heat shocked (37°C for 1 h) at 48 h AEL and imaged at 72 h AEL at 1024×512-pixel resolution. For BiFC imaging, larvae of the genotype ppk-Gal4/UAS-NYFP::Rop; ppk-CD4tdTomato/UAS-Sec6::CYFP were imaged at 96 h AEL at 1024×512-pixel resolution.

Photoconversion and FRAP assays were conducted on a Leica SP5 with a 40×1.25NA oil objective at 1024×256-pixel resolution. For photoconversion assays, 10 µm of either a terminal or a major dendrite of a ppk-gal4/+; UAS-myr-tdEosFP/+ or ppk-gal4/+; UAS-tdEosFP/+ larva at 72 h AEL was selected as a region of interest (ROI). The ROI was illuminated for 5 s with a 50-mW 405-nm Diode laser (Leica) at 40% output to photoconvert tdEOS and imaged at 10 s and 300 s post-conversion with a 500–520-nm bandwidth. The fluorescence intensity of the converted signal was measured for the central 8 µm of the converted region and the adjacent 8 µm in both directions. Intensity was analyzed using line plots (ImageJ) with eight individual samples. For FRAP assays, ROIs in control and Rop^G27 mutant C4da MARCM clones were photobleached for 3 s with a 405-nm laser at 60% power such that 50% of the initial fluorescence intensity was bleached. Baseline fluorescence was measured from five images acquired before bleaching, and fluorescence recovery was measured with 20 scans at 2-s intervals after bleaching. Background fluorescence values were subtracted from measurements of a nearby unbleached dendritic region at each time point. To control for photobleaching and sample movement out of focus, fluorescence intensity was monitored in multiple regions of the arbor outside of the ROI; only samples for which intensities remained steady in these control regions were included in data analysis. Recovery half-time and the recoverable pool were measured using Leica FRAP Wizard software.

Adult fly heads (50 per experiment) were isolated and homogenized in 500 ml of lysis buffer (15 mM Tris-HCl at pH 7.6, 0.25 mM sucrose, 15 mM NaCl, 15 mM MgCl₂, 25 mM EDTA, 1 mM EGTA and protease inhibitor cocktail). After pre-clearing with Protein-A agarose (Life Technologies), extracts were incubated with 15 ml of anti-Rop antibody and 50 ml of Protein-A agarose beads. The beads and input extract were eluted 1:1 in Laemmli buffer with 2-mercaptoethanol (Bio-Rad Laboratories) and boiled for 5 min. Samples were resolved on a 4–20% polyacrylamide gel and transferred to a nitrocellulose membrane. Membranes were blocked for 1 h in Blocking Buffer (Rockland Immunochemicals) and incubated overnight at 4°C with primary antibodies against Sec6 at 1:250 (Beronja et al., 2005) or against Sec15 at 1:250 (Mehta et al., 2005). Membranes were then incubated with secondary antibodies (anti-guinea pig IRDye700 at 1:10,000; Rockland Immunochemicals), and blots were imaged and processed on an Odyssey Infrared Imaging System (LI-COR Biosciences).

Rop and Sec6 coding sequences were PCR amplified from genomic DNA and inserted into pUAST-BiFC vectors (Gohl et al., 2010) using Gateway cloning (Life Technologies) to generate expression vectors with Rop fused to coding sequences for the N-terminal portion of YFP and Sec6 fused to the C-terminal portion of YFP (pUAST-NYFP-Rop, pUAST-Rop-NYFP, pUAST-CYFP-Sec6, pUAST-Sec6-CYFP). As negative controls, we generated pUAST-NYFP and pUAST-CYFP expression vectors that contained coding sequences for the N-terminal and C-terminal portion of YFP, respectively. To assay for YFP reconstitution, we used Effectene (Qiagen) to transfect S2 cells with Actin-Gal4, UAS-RFP and all possible combinations of the above NYFP and CYFP expression constructs; only pUAST-NYFP-Rop+pUAST-Sec6-CYFP yielded a signal upon BiFC analysis, therefore we used those expression constructs to monitor Rop–Sec6 complex formation in vivo.

Two-dimensional projections of image z-stacks were used for computer-assisted dendrite tracing and analysis with either the NeuronJ plugin in ImageJ (Meijering et al., 2004) or Neurolucida (Figs 2, 4 and 8; MBF Bioscience). For Sholl analysis (Sholl, 1953), neuron traces were generated in NeuronJ, and analysis was performed using the Sholl Analysis plugin in ImageJ. Statistical tests are indicated in the figure legends.

We thank Hugo Bellen, Thomas Schwarz, Ulrich Tepass, Kazuo Emoto, the Bloomington Stock Center, and the Developmental Studies Hybridoma Bank for fly stocks and antibodies; Sven Bogdan for BiFC vectors; and Kazuo Emoto, Peter Soba, Daichi Kamiyama and Mark Eddison for critical reading of the manuscript.