Prd1 associates with the clathrin adaptor α-Adaptin and the kinesin-3 Imac/Unc-104 to govern dendrite pruning in Drosophila

Refinement of the nervous system depends on selective removal of excessive axons/dendrites, a process known as pruning. Drosophila ddaC sensory neurons prune their larval dendrites via endo-lysosomal degradation of the L1-type cell adhesion molecule (L1-CAM), Neuroglian (Nrg). Here, we have identified a novel gene, pruning defect 1 (prd1), which governs dendrite pruning of ddaC neurons. We show that Prd1 colocalizes with the clathrin adaptor protein α-Adaptin (α-Ada) and the kinesin-3 immaculate connections (Imac)/Uncoordinated-104 (Unc-104) in dendrites. Moreover, Prd1 physically associates with α-Ada and Imac, which are both critical for dendrite pruning. Prd1, α-Ada, and Imac promote dendrite pruning via the regulation of endo-lysosomal degradation of Nrg. Importantly, genetic interactions among prd1, α-adaptin, and imac indicate that they act in the same pathway to promote dendrite pruning. Our findings indicate that Prd1, α-Ada, and Imac act together to regulate discrete distribution of α-Ada/clathrin puncta, facilitate endo-lysosomal degradation, and thereby promote dendrite pruning in sensory neurons.


Author summary
During the maturation of the nervous system, some neurons can selectively eliminate their unnecessary connections, including dendrites and axons, to retain specific connections. In Drosophila, a class of sensory neurons lose all their larval dendrites during metamorphosis, when they transition from larvae to adults. We previously showed that these neurons prune their dendrites via lysosome-mediated degradation of a cell-adhesion protein, Neuroglian. In this paper, we identified a previously uncharacterized gene, pruning defect 1 (prd1), which plays an important role in dendrite pruning. We show that Prd1 is localized and complexed with α-Adaptin and Imac, two other proteins that are also essential for dendrite pruning. Moreover, Prd1, α-Adaptin, and Imac act in a common pathway to promote dendrite pruning by down-regulating Neuroglian protein. Thus, our study highlights a mechanism whereby Prd1, α-Adaptin, and Imac act together to regulate a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

Introduction
Neuronal remodeling is a pivotal step in the formation of mature nervous systems during animal development. Developing neurons often outgrow superfluous axonal or dendritic branches at early developmental stages. Selective elimination of the unneeded branches without the death of parent neurons, referred to as pruning, is crucial for the refinement of neuronal circuits at late stages [1][2][3]. Neuronal pruning is a naturally occurring process in mammals and insects. In the central and peripheral nervous systems of mammals, many neurons often prune their unwanted or inappropriate neurites in order to establish proper and functional neuronal connections [4][5][6]. In insects, such as Drosophila, the nervous system is drastically remodeled during metamorphosis [7][8][9]. In the central nervous system, mushroom body (MB) γ neurons prune their larval axonal/dendritic branches and extend their adult-specific processes to be integrated into the adult brains prior to eclosion [10]. In the peripheral nervous system, dendritic arborization (da) neurons undergo either apoptosis or pruning during early metamorphosis. For example, in the dorsal cluster, class IV da neurons (ddaC) and class I da neurons (ddaD and ddaE) selectively eliminate their larval dendrites whereas their axons remain intact [11,12], while class III da neurons (ddaF) are apoptotic [12]. The dendrite-specific pruning event involves the formation of swellings and retracting bulbs [12], morphologically resembling the axon/dendrite degenerative process associated with brain injury and neurodegenerative diseases. Thus, understanding the mechanisms of developmental pruning would provide insight into neurodegeneration in pathological conditions. Dendrite pruning of Drosophila ddaC sensory neurons has emerged as an attractive paradigm to elucidate the molecular and cellular mechanisms of neuronal pruning. In response to a late larval pulse of the steroid-molting hormone 20-hydroxyecdysone (ecdysone), the dendrites of ddaC neurons are severed at their proximal region at 5-8 h after puparium formation (APF), and subsequently the detached dendrites are rapidly fragmented and undergo phagocytosis-mediated degradation by [16][17][18] h APF ( Fig 1A) [11,12]. It has been well documented that ecdysone and its nuclear receptors are required to induce the expression of several major downstream targets to initiate dendrite pruning [13,14].
Clathrin-mediated endocytosis (CME) is a major entry route that regulates the surface expression of transmembrane proteins and turnover of lipid membrane in eukaryotic cells [15]. The process is mediated by the highly conserved heterotetrameric Adaptor protein-2 (AP-2) protein complex composed of α, β, μ, and σ subunits [16]. AP-2 is a CME-specific clathrin-associated adaptor that recruits clathrin to the plasma membrane-specific lipid phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P 2 ), leading to the formation of clathrin-coated pits. The invaginated pits are cleaved by the small GTPase dynamin to form clathrin-coated endocytic vesicles. Newly formed endocytic vesicles can fuse with early endosomes, a process mediated by the key GTPase Rabaptin-5 (Rab5) [17]. Among several downstream routes, early endosomes can mature into multivesicular bodies in an endosomal sorting complexes required for transport (ESCRT)-dependent manner and subsequently fuse with lysosomes to degrade their protein and membrane components, a process known as endo-lysosomal maturation and degradation pathway [18]. It has been reported that CME regulates axon growth/guidance, dendrite extension/branching, and synaptic vesicle trafficking in vertebrate and invertebrate neurons [19,20].
pruning [23]. A parallel study also showed that Rab5/dynamin-dependent endocytosis appears to predominantly occur at the proximal regions of dendrites, leading to dendritic thinning and compartmentalized Ca 2+ transients in ddaC neurons [22]. In MB γ neurons, PI3K-cIII/dynamin-dependent endo-lysosomal degradation pathway down-regulates the Hedgehog receptor Patched to promote axon pruning [21]. These studies highlight a general requirement of endocytosis and endo-lysosomal degradation for regulating distinct modes of neuronal pruning. However, the regulatory mechanism that promotes endocytosis and endo-lysosomal degradation pathway during neuronal pruning remains poorly understood.
Here, we identified the critical role of a novel Drosophila SKIP-related gene, pruning defect 1 (prd1), in regulating dendrite pruning of ddaC sensory neurons. Mammalian Salmonella induced filament A (SifA) and Kinesin-interacting protein (SKIP, also known as PLEKHM2) was originally identified as a target of the Salmonella effector protein SifA [24]. In uninfected mammalian cells, SKIP regulates the distribution of late endosomes and lysosomes [25,26]. However, the in vivo roles of SKIP and its related homologues during animal development are unknown. We show that Prd1 colocalizes with the endocytic components α-Adaptin (α-Ada) and clathrin in the dendrites of ddaC neurons. It forms a protein complex with α-Ada, the α subunit of the AP-2 complex (also known as AP-2α), but not with the endosomal GTPase Rab5. Similar to Prd1, α-Ada and other subunits of AP-2 complex are all required for dendrite pruning in sensory neurons. Moreover, we show that Prd1 colocalizes and associates with the kinesin-3 immaculate connections (Imac)/Uncoordinated-104 (Unc-104). Importantly, both imac mutants and dominant-negative constructs exhibited severe dendrite pruning defects in ddaC sensory neurons. Prd1, α-Ada, and Imac facilitate endo-lysosomal degradation of Nrg prior to dendrite pruning. Furthermore, genetic interactions among prd1, α-ada, and imac suggest that they participate in the same pathway to promote dendrite pruning. Thus, our data demonstrate that the Prd1/α-Ada/Imac pathway promotes dendrite pruning via regulating α-Ada/clathrin distribution and endo-lysosomal degradation of Nrg.

prd1 is required for dendrite pruning in sensory neurons
To identify novel players in dendrite pruning, we expressed a collection of RNA interference (RNAi) lines using a class IV da neuron driver pickpocket-Gal4 (ppk-Gal4) to knock down gene function in ddaC neurons. Two independent RNAi lines, v108557 (#1) and v40070 (#2), were isolated with dendrite pruning defects. Both of RNAi lines target against a novel gene, CG17360, which we therefore named pruning defect 1 (prd1). RNAi knockdown of prd1 via ppk-Gal4 caused prominent dendrite pruning defects in ddaC neurons at 16 h APF (v108557, n = 15, Fig 1C, 1I and 1J; v40070, n = 21, Fig 1I and 1J). In contrast, at the same time point, larval dendrites were completely removed in the control neurons (n = 25; Fig 1B, 1I and 1J). prd1 encodes a previously uncharacterized protein with 1,354 amino acids, which contains a pleckstrin homology (PH) domain at its C-terminal portion (S1A Fig). Database searches revealed that in the Drosophila genome, Prd1 is most closely related to mammalian SKIP/PLEKHM2 (S1A Fig). They share amino acid sequence identity in their C-terminal portions, including their PH domains and the flanking regions (S1A Fig). SKIP was reported to regulate endosomal/lysosomal distribution in mammalian cells [25,26]. However, the function of Drosophila Prd1 was completely unknown.
To further verify the requirement of prd1 for dendrite pruning, we took advantage of a prd1 M56 mutant allele that was previously generated via flippase (FLP)-mediated recombination between two flippase recognition target (FRT)-containing P-element insertions (S1B Fig). It deletes the majority of the prd1 coding region (aa329-1,354) (S1B Fig) and hence is a strong hypomorphic allele. Mutants hemizygous for prd1 M56 and a small deletion Df(3R)Exel7310 (deleting the entire prd1 gene and its neighboring genes) died at the pharate adult stage and exhibited prominent dendrite pruning defects in ddaC neurons at 16 h APF. A total of 96% of those hemizygous mutant neurons exhibited dendrite severing defects and retained their larval dendrites with the attachment to their cell bodies by 16 h APF (n = 24; Fig 1D, 1I and 1J). These dendrite pruning defects in the mutant pupae hemizygous for prd1 M56 and Df(3R) Exel7310 were fully rescued by ectopic expression of full-length Prd1 (n = 22; Fig 1E, 1I and 1J) or Venus-tagged Prd1 (n = 19; Fig 1F, 1I and 1J). By mosaic analysis with a repressible cell marker (MARCM) analyses in ddaC neurons, another allele, prd1 PS1 (S1B Fig), phenocopied prd1 M56 /Df(3R)Exel7310 mutants. All prd1 PS1 mutant ddaC neurons failed to prune away their larval dendrites (n = 11; Fig 1G, S1D Fig) and exhibited 91% of severing defects and 9% of fragmentation defects (Fig 1I). The length of unpruned dendrites in prd1 PS1 mutants is comparable to that in hemizygous mutants of prd1 M56 and Df(3R)Exel7310 ( Fig 1J). Importantly, ectopic expression of full-length Prd1 also fully rescued their dendrite pruning defects in prd1 PS1 mutant ddaC clones (n = 12; Fig 1J), confirming that the dendrite pruning defects in prd1 PS1 mutant neurons are caused by loss of prd1 function.
We next examined whether Prd1 regulates α-Ada localization in the dendrites. When prd1 was knocked down via RNAi, GFP-α-Ada formed prominent aggregates in the dendrites of mutant neurons (n = 11, 45%; Fig 2F), compared with the control neurons (n = 11). Likewise, Venus-Prd1 often accumulated on several enlarged cellular aggregates in the dendrites of αada RNAi mutant neurons (n = 11; Fig 2G, arrowheads), compared with the control neurons (n = 11; Fig 2G). These results suggest that Prd1 and α-Ada are mutually required for their distributions in the dendrites. We previously reported that the expression of Rab5 DN or Vacuolar protein sorting-associated protein 4 (Vps4) DN led to robust accumulation of ubiquitinated protein on enlarged endosomes in ddaC neurons [23]. Interestingly, Venus-Prd1 signals were also enriched on enlarged ubiquitin-positive endosomes in Rab5 DN -or Vps4 DN  Collectively, Prd1 predominantly colocalizes with α-Ada/clathrin-positive puncta in the dendrites and its distribution requires the endocytic regulators α-Ada, Rab5, and Vps4.

Prd1 physically associates with α-Ada but not Rab5 in S2 cells
Given their close localization patterns, we next attempted to examine potential protein-protein interactions between Prd1 and α-Ada/Rab5. To this end, we co-transfected S2 cells and conducted co-immunoprecipitation (co-IP) experiments. α-Ada was present specifically in the immune complex when Prd1 was immunoprecipitated using an anti-Myc antibody ( Fig 3A). Reciprocally, Prd1 was also co-immunoprecipitated in the α-Ada complex using an anti-Flag antibody ( Fig 3B). The β subunit of AP-2 is Bap (β-Adaptin, also known as AP-2β or AP-1-2β), which is probably shared between Adaptor protein-1 (AP-1) and AP-2 complexes in Drosophila [28]. Similar to the association with α-Ada, Prd1 also formed a complex with Bap in S2 cells co-transfected with Myc-Prd1 and Flag-Bap (S6A and S6B Fig). In contrast, Prd1 did not present in the same immune complex with Rab5 in both directions of co-IP experiments (S7A and S7B Fig). Thus, Prd1 forms a protein complex with α-Ada and Bap, rather than with Rab5.
Collectively, α-Ada and other AP-2 subunits are critical for dendrite pruning in sensory neurons, whereas α-Ada is dispensable for neuronal apoptosis. Therefore, α-Ada and Prd1 play similar roles in dendrite pruning but not in neuronal apoptosis during metamorphosis.

Prd1 colocalizes with a neuronal kinesin Imac and links α-Ada to Imac
Prd1-related mammalian SKIP/PLEKHM2 binds to kinesin-1 and activates it to regulate the distribution of endosomes/lysosomes [25,26]. To investigate whether Prd1 regulates the distribution of α-Ada puncta via a motor protein, we examined potential colocalization between Prd1 and various motor proteins, including kinesin-1, 2, and 3 and dynein. Venus-Prd1 was distributed in a punctate pattern, which is distinct from the GFP-tagged Kinesin heavy chain (Khc-GFP) puncta (n = 11; S11A Fig), suggesting that unlike SKIP, Prd1 may function independently of kinesin-1 in Drosophila. Neither did Prd1 colocalize with Kinesin associated protein 3 (Kap3), a subunit of the heterotrimeric kinesin-2 motor (n = 24; S11B Fig). Remarkably, when Venus-Prd1 was co-expressed with red fluorescent protein (RFP)-tagged Imac (Imac-RFP), a key neuronal kinesin-3, Prd1 fully colocalized with Imac-RFP in the dendrites (arrowheads) and soma of ddaC neurons (94%, n = 462 puncta; Fig 5A). Moreover, we did not observe a similar distribution pattern between Prd1 and Dynein light intermediate chain (Dlic) or the dynein regulator Lis1 (n = 5 and 11, respectively; S11C and S11D Fig). Thus, Prd1 appears to specifically colocalize with kinesin-3, but not with kinesin-1 and 2 or dynein, in the dendrites of ddaC sensory neurons.

Imac plays a crucial role in dendrite pruning
Because the role of Imac in dendrite pruning is unknown, we next explored whether knockdown or loss of imac function caused dendrite pruning defects. Two independent imac RNAi lines, v47171 (#1) and v23465 (#2), when expressed in ddaC neurons via ppk-Gal4, led to severe dendrite pruning defects in all ddaC neurons at 16 h APF (n = 13 and 5, respectively; Fig 6B,  S14 Fig). In contrast, the control neurons showed normal dendrite pruning at 16 h APF (n = 17; Fig 6A). Moreover, using the previously reported null allele imac 170 [33], we generated their ddaC mutant clones and analyzed their defects in dendrite pruning at 16 h APF. ddaC clones homozygous for imac 170 exhibited strong dendrite pruning defects by 16 h APF, with full penetrance (n = 6, respectively; Fig 6C, 6G and 6H). Similar to that reported in a previous RNAi study [34], we also observed severe dendrite arborization defects in imac RNAi neurons and mutant clones. Only major dendrites were still present in the vicinity of mutant ddaC soma at the WP stage (n = 11 and 6, respectively; Fig 6B and 6C). Both dendrite morphology and pruning defects in imac 170 mutant neurons were fully rescued by the expression of Imac-RFP (n = 10; Fig 6D, 6G and 6H), suggesting that Imac-RFP functionally substitutes for endogenous Imac. Thus, Imac is required for both pruning and growth of dendrites in ddaC neurons.
Like nematode Unc-104, Imac contains a motor domain, three coiled coil domains, a forkhead-associated domain, and a PH domain (S13 Fig). We identified a conserved ATP-binding sequence (GQTGAGKS) within the motor domain of the Imac protein and generated two mutant forms, GQTGAEKS (Imac G102E ) and GQTGAAAA (Imac AAA ) (S13 Fig), which were reported to disrupt the motor activity of kinesins and myosins in other organisms [35]. We found that both Drosophila Imac G102E and Imac AAA behaved as dominant-negative forms, because their overexpression phenocopied imac loss-of-function mutants in terms of both dendrite arborization and pruning defects (n = 13 and 7, respectively; S14 Fig compared to Fig  6B and 6C). These data support the conclusion that the kinesin motor activity of Imac plays a crucial role in dendrite arborization and pruning.
To rule out the possibility that the imac-associated dendrite pruning defects are secondary to the initial dendrite arborization defect, we conducted the Gene-Switch experiments by inducing the expression of these two dominant-negative forms at the early third instar larval stage (72 h after egg laying [AEL]). The Gene-Switch manipulations enabled mutant ddaC neurons to arborize mature and complex larval dendrites, as shown at the WP stage (n = 12 and n = 8, respectively; Fig  Taken together, multiple lines of genetic evidence demonstrate that the Drosophila kinesin-3 Imac is a crucial motor protein regulating dendrite pruning in ddaC sensory neurons.

Imac regulates α-Ada/clathrin distribution and endo-lysosomal degradation
We next examined whether imac is important for the distributions of Prd1, α-Ada, and clathrin in ddaC neurons. First, we examined the distribution of the endogenous Prd1 protein with an anti-Prd1 antibody in imac knockdown or mutant neurons. In contrast to weak punctate signals of Prd1 in the control soma (n = 11; Fig 7A), Prd1 accumulated on several bright aggregates in either imac RNAi or imac 170 mutant neurons (n = 11 and 7, respectively; Fig 7A), suggesting that Imac promotes discrete distribution of Prd1 puncta in ddaC neurons. To better visualize its distribution in the dendrites, we expressed Venus-Prd1 and compared its distributions in control and imac RNAi ddaC neurons. In imac RNAi ddaC neurons, Venus-Prd1 strongly accumulated on many aggregates in the dendrites (#1, n = 21; control, n = 10; Fig 7B). Similar to the endogenous Prd1 protein (Fig 7A), Venus-Prd1 was also observed to form  Fig 7B, S16 Fig). Likewise, using a previously reported anti-α-Ada antibody [29], we observed that endogenous α-Ada also aggregated on several large puncta in imac RNAi or imac 170 mutant soma (n = 9 and 9, respectively; control: n = 12; Fig 7C). When overexpressed in imac RNAi ddaC neurons (#1 and #2), GFPα-Ada, like the endogenous protein, strongly accumulated on the aggregates (n = 8 and 12, respectively; S16 Fig). Moreover, mRFP-Chc also accumulated as more aggregates in the dendrites and soma of imac RNAi ddaC neurons (n = 13, Fig 7D), compared to its distribution with small discrete puncta in control RNAi neurons. In contrast, we did not observe any obvious aggregates of Venus-Prd1 and α-Ada in khc RNAi ddaC neurons (n = 6 and 19, respectively; S17 Fig). These findings indicate that the kinesin-3 Imac, rather than kinesin-1, plays an important role in distributing Prd1, α-Ada, and clathrin in the dendrites of ddaC neurons.
Given that Imac regulates the distributions of α-Ada/clathrin puncta, we further examined whether loss of imac function impairs endosomal distribution and lysosomal degradation. In imac RNAi ddaC neurons, the Venus-Prd1 aggregates were enriched with the endosomal markers anti-hepatocyte growth factor-regulated tyrosine kinase substrate (Hrs) (n = 6; S18A and S18D Fig) and Rab5-GFP (n = 8; S18B and S18D Fig) in the dendrites and soma, suggesting that these Venus-Prd1 structures are aberrant endosomes. We next examined whether these aberrant endosomes can fuse with lysosomes to undergo protein degradation. To this end, we utilized the LysoTracker dye to label highly acidified lysosomal compartments that normally fuse with endosomes to degrade the ubiquitinated proteins. The LysoTracker dye labeled discrete lysosomes in control ddaC neurons (n = 16; Fig 7E). However, the Lysotracker dye was not enriched on aberrant endosomes in imac RNAi mutant neurons (n = 10; Fig 7E), suggesting a compromise in endosomal acidification and maturation. Consistently, ubiquitinated proteins, which normally exhibited weak punctate structures in control neurons (n = 12, Fig 7F), were strongly enriched on Venus-Prd1 aggregates in imac RNAi neurons (n = 8; Fig  7F), suggesting impaired endo-lysosomal degradation. As controls, in imac RNAi ddaC neurons, the secretory vesicle marker Sec15 did not accumulate as aggregates (n = 4; S18C and S18D Fig) and the Golgi marker β1,4-galactosyltransferase (GalT)-GFP appeared to be normal in size and distribution in the soma and dendrites (n = 6; S18E Fig).
The L1-CAM Nrg is endocytosed and down-regulated via the endo-lysosomal degradation pathway, leading to dendrite pruning in ddaC neurons [23]. We next examined whether Prd1, α-Ada, and Imac regulate endo-lysosomal degradation of Nrg before the onset of dendrite pruning. In wild-type ddaC neurons, Nrg protein levels were significantly decreased in the somas, dendrites, and axons (n = 23; S19A and S19E Fig) at 6 h APF. Importantly, the Nrg protein accumulated dramatically in the somas, dendrites, and axons of prd1 M56 /Df(3R)Exel7310 (n = 23; S19B and S19E Fig), α-ada 3 MARCM (n = 13, S19C and S19E Fig), or Imac G102E (n = 9; S19D and S19E Fig) ddaC neurons, similar to those in Rab5 DN mutant neurons. These data indicate that Prd1, α-Ada, and Imac are required to promote Nrg endo-lysosomal degradation prior to dendrite pruning. Moreover, the expression of an nrg RNAi line, which has been shown to efficiently knock down its protein [23], significantly suppressed the pruning defects of prd M56  Thus, Prd1, α-Ada, and Imac act to promote dendrite pruning at least partly through global endo-lysosomal degradation of Nrg. Moreover, a previous study has reported that local endocytosis leads to the thinning of proximal dendrites and compartmentalized calcium transients [22]. We next investigated whether Prd1/Imac/ Unc-104 regulates local calcium transients at 6.5 h APF. While compartmentalized Ca 2+ transients were present in the vast majority of control neurons (n = 21; S21 Fig), the percentage of ddaC neurons with Ca 2+ transients at 6 h APF was drastically reduced in prd1 M56 /Df(3R) Exel7310 (n = 20) and Imac G102E (n = 13) ddaC neurons (S21 Fig). These data suggest that Prd1 and Imac are also required to regulate compartmentalized calcium transients before the onset of dendrite pruning.
Taken together, Imac is required for proper distributions of Prd1, α-Ada, and clathrin in the dendrites and facilitates lysosomal degradation of Nrg in ddaC sensory neurons.

Prd1 forms a protein complex with Imac but not Khc in S2 cells
To further explore the mechanisms whereby Prd1 and Imac regulate dendrite pruning, we conducted co-IP experiments to assess their potential physical association. In S2 cells co-transfected with Myc-Prd1 and Imac-HA, Prd1 was pulled down when Imac was immunoprecipitated using an anti-HA antibody (Fig 3C). Reciprocally, Imac was also co-immunoprecipitated in the Prd1 immune complex (Fig 3D). These co-IP data, together with the colocalization results (Fig 5A), strongly support a functional link between prd1 and imac during dendrite pruning. Moreover, in S2 cells co-transfected with Flag-α-Ada and Imac-HA, the association between Imac and α-Ada was not detectable (Fig 3E). Importantly, when Myc-Prd1 was cotransfected with Flag-α-Ada and Imac-HA, Imac was co-immunoprecipated by α-Ada ( Fig  3F). Thus, α-Ada and Imac form a protein complex in the presence of Prd1 in S2 cells. As a control, consistent with distinct localizations of Prd1 and the kinesin-1 Khc, we did not observe their physical association in the reciprocal co-IP experiments (S22A and S22B Fig). These data suggest the selectivity of the interaction between Prd1 and the kinesin-3 Imac. Furthermore, Imac was not co-immunoprecipitated with Rab5 in reciprocal co-IP experiments (S22C and S22D Fig). Taken together, our biochemical and cell biological data imply that Prd1 might interact with Imac and recruit α-Ada to Imac in vivo to facilitate endo-lysosomal degradation in ddaC sensory neurons.
In summary, multiple lines of genetic, biochemical, and cell biological evidence support the model that Prd1, α-Ada, and Imac act in the same pathway to regulate discrete distribution of α-Ada/clathrin puncta, facilitate endo-lysosomal degradation of Nrg, and thereby promote dendrite pruning in sensory neurons (in S24 Fig).

Discussion
Previous studies have demonstrated that endocytosis plays important roles in governing dendrite and axon pruning in ddaC and MB γ neurons, respectively [21][22][23]. However, the regulatory mechanism of endocytosis required for neuronal pruning is poorly understood. Moreover, kinesin motor proteins that regulate endocytosis during neuronal pruning are completely unknown. Here, we identified Prd1 and its associated proteins, the clathrin adaptor α-Ada and the kinesin-3 Imac, which all play important roles in dendrite pruning of ddaC neurons. Prd1 physically associates with α-Ada and Imac and colocalized with them in the dendrites. Similar to that in prd1 mutant, loss of either α-ada or imac resulted in prominent dendrite pruning defects in ddaC neurons. Moreover, loss of imac function caused the formation of aberrant Prd1/α-Ada/clathrin aggregates in the dendrites as well as impaired endolysosomal degradation. Finally, genetic interaction results further support that prd1, α-ada, and imac act in the same pathway to promote dendrite pruning. Our study demonstrates that a clathrin adaptor-associated protein, Prd1, acts together with α-Ada and a kinesin-3 to regulate proper distribution of α-Ada/clathrin puncta and facilitate endo-lysosomal degradation during neuronal pruning.

Prd1 regulates dendrite pruning
Growing evidence indicates that mammalian SKIP is involved in proper distribution of endosomes and lysosomes in mammalian cells. In infected cells, SKIP interacts with kinesin-1 to recruit the motor on the bacteria's replicative vacuole, leading to the formation and distribution of late endosomes or lysosomes [37]. In uninfected cells, SKIP binds to the small GTPase Arf-like GTPase 8 (Arl8) through its RUN domain or to the late endosomal GTPase Rab9 via its PH domain to regulate lysosomal distribution in a kinesin-1-dependent fashion [25,26].

(E) Dendrites of α-ada 3 /+, imac RNAi/+, and α-ada 3 /+; imac RNAi/+ ddaC neurons at 16 h APF. Quantification of total length of unpruned dendrites at 16 h APF. The number of samples (n) in each group is shown on the bars. Error bars represent SEM. Scale bar (A)
represents 50 μm. ÃÃ p < 0.01; ÃÃÃ p < 0.001. Dorsal is up in all images. The individual numerical values for panels A, B, C, D, and E can be found in S1 Data. The genotypes can be found in S1 Text. α-ada, α-adaptin; APF, after puparium formation; Df, deficiency; imac, immaculate connections; prd1, pruning defect 1; RNAi, RNA interference. Moreover, a mutation in human PLEKHM2 caused aberrant accumulation of both early and late endosomes in patients' fibroblast cells [38]. However, the physiological function of SKIP/ PLEKHM2 remained unknown. In this study, we demonstrate that a previously uncharacterized Drosophila SKIP-related gene, prd1, plays a critical role in dendrite pruning.
Several lines of evidence support the notion that Prd1 regulates dendrite pruning via AP-2 complex. First, Prd1 colocalized with the AP-2 α subunit α-Ada and clathrin in the dendrites. Second, Prd1 physically associated with α-Ada but not Rab5. Third, Prd1 and α-Ada are mutually required for their proper distributions in the dendrites of ddaC neurons. Fourth, Prd1 distribution also requires the endocytic regulators Rab5 or Vps4. Prd1, similar to clathrin and the clathrin adaptor AP-2, was enriched on enlarged endosomes in Rab5 DN and Vps4 DN mutant neurons, in contrast to its discrete distribution in wild type. Moreover, both prd1 and α-ada are important for endo-lysosomal degradation of Nrg. Finally, they genetically interacted during dendrite pruning. Reduction of prd1 gene dose (prd1 M56 /+) dominantly enhanced the pruning phenotypes of α-ada RNAi mutants. The functional link between Prd1 and α-Ada suggests that Prd1 and α-Ada act together to regulate endo-lysosomal degradation. Similar to Prd1, in a previous elegant study, Numb-associated kinase (Nak), a Drosophila Actin-related kinase (Ark) family member, was identified as another binding partner of the clathrin adaptor protein AP-2 [20]. Nak distributes clathrin puncta in higher-order dendrites to promote dendritic growth [20]. However, we did not observe prominent dendrite pruning defects using a null nak 2 mutant (n = 29), suggesting that AP-2-associated Nak is dispensable for ddaC dendrite pruning. Thus, it is conceivable that Prd1 acts as a novel binding protein of AP-2 to facilitate endo-lysosomal degradation of Nrg in the dendrites and promote dendrite pruning in sensory neurons.

Imac regulates α-Ada/clathrin distribution and endo-lysosomal degradation of the L1-CAM Nrg
Kinesin motor proteins transport intracellular cargos along microtubule tracks [39]. Imac/ Unc-104 belongs to the evolutionarily conserved kinesin-3 family. Imac/Unc-104 was reported to regulate synapse formation and synaptic vesicle transport in axons in fly and worm [33,[40][41][42][43]. In mammals, two Imac homologues, namely kinesin (KIF)1A and KIF1Bβ, transport synaptic vesicle precursors in axons [44,45]. Here, we provided multiple lines of genetic evidence using RNAi knockdown, loss-of-function mutants, and dominant negative approaches, unambiguously demonstrating that Imac is required for dendrite pruning, consistent with a recent report on imac RNAi knockdown phenotypes in ddaC neurons [46]. More importantly, for the first time we provide mechanistic insight into how the Drosophila kinesin-3 Imac regulates the distribution of α-Ada/clathrin puncta and promotes dendrite pruning of sensory neurons. imac mutant ddaC neurons showed severe defects in dendrite pruning and initial dendrite arborization, resembling α-ada mutants. imac displayed significant genetic interaction with αada. Moreover, Imac appeared to colocalize with the clathrin adaptor α-Ada and its interacting protein Prd1 in the dendrites. Interestingly, in imac mutant neurons, clathrin, α-Ada, and Rab5/Hrs accumulated into numerous aberrant aggregates, suggesting aberrant endosomal formation/distribution. The aberrant aggregates were also rich in ubiquitin but lacked Lyso-Tracker signals, indicative of defective endo-lysosomal degradation. Furthermore, the L1-CAM Nrg protein accumulated dramatically in the somas, dendrites, and axons of Imac G102E ddaC neurons (this study), whereas Nrg was drastically degraded via endo-lysosomal degradation pathway in wild-type neurons [23]. Therefore, our study demonstrates a novel role of Imac in promoting endo-lysosomal maturation/degradation in sensory neurons. Similarly, Caenorhabditis elegans Unc-104 regulates autophagosome formation and maturation in neurons [47]. Therefore, it is possible that the primary role of imac is to facilitate the formation or fusion of early endosomes and thereby lysosome-mediated degradation of the L1-CAM Nrg during dendrite pruning.

Prd1 recruits α-Ada-positive puncta to the kinesin-3 Imac in sensory neurons
Our cell biological, biochemical, and genetic data reveal a functional link between Prd1 and the kinesin-3 Imac in dendrite pruning. First, we show that Prd1 colocalizes with Imac but not with kinesin-1/2 or dynein. Second, Imac is required for the normal distribution of Prd1 and endocytic components α-Ada/clathrin, presumably endocytic vesicles, in dendrites. Third, Prd1 forms a protein complex with α-Ada and Imac but not with Rab5 or kinesin-1. Finally, prd1 and imac are required for dendrite pruning in the same genetic pathway.
How does imac regulate endo-lysosomal degradation of Nrg to promote dendrite pruning? Our previous findings demonstrated that Rab5-dependent endocytosis regulates dendrite pruning via endo-lysosomal degradation of the L1-CAM Nrg [23]. We show here that Imac colocalized with the clathrin adaptor α-Ada via Prd1 in the dendrites but not with Rab5. Moreover, co-IP experiments suggested that Prd1 formed a protein complex with α-Ada but not with Rab5. One possibility is that the Prd1/α-Ada/Imac pathway might function at the internalization step of endocytosis, preceding the function of Rab5, to facilitate the formation/ fusion of early endosomes and thereby endo-lysosomal maturation/degradation during dendrite pruning. We envision that upon the cleavage of clathrin-coated vesicles, Prd1 associates with the clathrin adaptor protein AP-2 to recruit the newly formed vesicles to the kinesin-3 Imac, which in turn delivers them for the formation and fusion of early endosomes; early endosomes undergo endo-lysosome maturation and degradation, leading to dendrite pruning. However, in the established model of CME, AP-2/clathrin-coated endocytic vesicles are uncoated after budding; clathrin and AP-2 are released back into the cytoplasm to participate in another round of CME. The possible speculation is that Imac might deliver the α-Ada/clathrin-coated vesicles before their uncoating. Alternatively, α-Ada/AP-2 might play a noncanonical role in regulating endo-lysosomal degradation after the formation of early endosomes. Growing studies have shown noncanonical functions of the adaptor proteins AP-1/AP-2 and clathrin in vitro. In cultured cells, clathrin or AP-1/AP-2-associated vesicles can be directly distributed by kinesin or dynein motors [48,49]. Moreover, noncanonical function of AP-2 has been also reported in a recent study, in which AP-2 acts as an adaptor that links autophagosomes to a dynein motor for proper vesicular distribution in cultured neurons [50]. Further investigations are required to distinguish these two possible mechanisms whereby the Prd1/α-Ada/Imac pathway regulates endo-lysosomal degradation of Nrg to promote dendrite pruning.
In summary, we identified a novel protein Prd1 and two associated proteins, including the clathrin adaptor α-Ada and the kinesin-3 Imac, which all play crucial roles in regulating dendrite pruning of sensory neurons. Mechanistically, Prd1, α-Ada, and Imac act in the same pathway to regulate discrete distribution of α-Ada/clathrin puncta, facilitate endo-lysosomal degradation of the L1-CAM Nrg, and thereby promote dendrite pruning in sensory neurons.

Generation of prd1, Bap, and imac transgenes
prd1 and Bap full-length cDNAs were PCR from EST LP07755 and LP17054 (DGRC) into Topo Entry vector (Life Tech, Carlsbad, CA). The GATEWAY pTW or pTVW vectors containing the respective fragment of the cDNAs were constructed by LR reaction (Life Tech, Carlsbad, CA).
The variants of Imac were generated by G102E and GKS102-104AAA site mutagenesis (Agilent Tech) using pUAST-imac-HA as a template, respectively. The respective cDNA fragments were amplified by PCR and subcloned into pTWH vector (DGRC). The transgenic lines were established by the Bestgene.

Generation of Bap mutant
P{EPgy2}EY01200 P-element insertion flies were crossed with the fly strain carrying the Δ2-3 transposase to induce imprecise excision events. Nearly 600 independent lines were established based on a loss of the w+ marker. One of lethal lines was found to be rescued by the duplication line Dp(1;Y)BSC140. Subsequent genomic PCR and DNA sequencing analysis indicated that this mutant Bap Δ1 harbors a 1,381-bp deletion.

Generation of Prd1 antibody
cDNA fragments corresponding to aa151-350 (antigen 1), aa300-575 (antigen 2), and aa 591-724 (antigen 3) of prd1 isoform A were amplified from EST LP07755 by Expand High Fidelity PCR System (Roche) and verified by DNA sequencing. The resultant products were expressed by the GST expression vector (pGEX4T-1, Pharmacia). After protein purification, the purified protein was used to immunize various guinea pigs and rats to generate polyclonal antibodies against Prd1. The specificity of the antibody was verified in ddaC neurons expressing both prd1 RNAi and imac RNAi lines.

Live imaging analysis
To image Drosophila da neurons at the wandering third instar (wL3) or WP stage, larvae or pupae were first washed in PBS buffer briefly, followed by immersion with 90% glycerol. For imaging da neurons at 16 h APF, pupal cases were carefully removed before they were mounted with 90% glycerol. Dendrite images were acquired on Leica TSC SP2. Subcellular localization images were acquired on Leica TCS SP8 STED 3× super-resolution microscope.

MARCM analysis of da sensory neurons
MARCM analysis, dendrite imaging, and quantification were carried out as previously described [13]. ddaC or other da clones were selected and imaged at the WP stage according to their location and morphology. The ddaC or other da neurons were examined for dendrite pruning defects at 16 h or 20 h APF.

RU486/mifepristone treatment for the Gene-Switch system
Embryos of appropriate genotype were collected at 6-h intervals and reared on standard food to the early third instar larva stage. The larvae were transferred to the standard culture medium, which contains 240 μg/mL mifepristone (Sigma Aldrich M8046). Neither puparium formation onset nor adult eclosion was affected by RU486 treatment. White prepupae with appropriate genotype were picked up, subject to dissection and phenotypic analysis at 16 h APF.

Immunohistochemistry and antibodies
The following primary and secondary antibodies were used for immuno-histochemistry at the indicated dilution: rat anti-Prd1 For immune-staining assays, wild-type and mutant pupae or larvae were dissected in cold PBS and fixed in 4% formaldehyde for 12 min. The samples within the same group of experiments were stained in the same tube and mounted in VectaShield mounting medium, and the samples were directly visualized by Leica TCS SP8 STED 3 × super-resolution microscope and processed in parallel. Data analysis and statistics were performed via Excel (Microsoft), and Imaris software. Based on the intensity profiles, the localization patterns were divided into three categories: colocalization, adjacent localization, and non-colocalization, as shown in the representative images (S25 Fig).

Quantification of ddaC dendrites
Live confocal images of da neurons expressing mCD8-GFP under the control of ppk-Gal4 or elav-Gal4 were shown at WP, 16 h APF and 20 h APF. For wild-type or mutant ddaC neurons, the percentages of fragmentation defect and severing defect were quantified in a 275 μm × 275 μm region of the dorsal dendritic field, originating from the abdominal segments 2-5. The severing defect was defined by the presence of dendrites that remain attached to the soma at 16 h APF, whereas dendrite fragmentation defect is referred to as the presence of dendrite branches near the ddaC territory that have been severed from their proximal parts at 16 h APF [11][12][13]. Total length of unpruned dendrites was measured in a 275 μm × 275 μm region of the dorsal dendritic field using ImageJ. The number of samples (n) in each group is shown on the bars. Statistical significance was determined using either two-tailed Student t test (two samples) or one-way ANOVA and Bonferroni test (multiple samples) ( Ã p < 0.05, ÃÃ p < 0.01, ÃÃÃ p < 0.001, n.s., not significant). Error bars represent SEM. Dorsal is up in all images.

Co-IP assay
S2 cell culture and western blotting were carried out as described below. Myc-Prd1, Flag-α-Ada, Flag-Rab5, Khc-Flag, and Imac-HA expression vectors were generated by Gateway cloning. S2 cells were cultured at 25˚C in Express Five SFM (Gibco) medium supplemented with 1% L-glutamine. For transfection, S2 cells were plated at a density of 2 × 10 6 cells per 35-mm dish 1 d before transfection. After 24 h culture, around 0.1-0.5 μg of each expression plasmid was transfected into the S2 cells using Effectene Transfection Reagent (Qiagen); cells were harvested 48 h post-transfection. Transfected S2 cells were homogenized with lysis buffer (25 mM Tris pH 8/27.5 mM NaCl/20 mM KCl/25 mM sucrose/10 mM EDTA/10 mM EGTA/1 mM DTT/10% [v/v] glycerol/0.5% Nonidet P40) with protease inhibitors (Complete, Boehringer; PMSF 10 mg/mL, sodium orthovanadate 10 mg/mL). The supernatants were used for IP with anti-Myc, anti-Flag, or anti-HA overnight at 4˚C followed by incubation with protein A/G beads (Pierce Chemical Co.) for 2 h. Protein A/ G beads were washed four times using cold PBS. Bound proteins were separated by SDS-PAGE and analyzed by western blotting with anti-Myc, anti-Flag, and anti-HA HRP-conjugated antibody. Co-IP experiments were repeated three times.

Calcium imaging
Calcium imaging was performed with Olympus FV3000 using 60× Oil lens. Calcium images from 6.5 h APF ddaC neurons were acquired for 400-500 frames at 1 frame per 1.8-2.25 s and analyzed using Metamorph (Molecular Devices) and ImageJ software. Red arrowheads indicate ddaC somas. All dendrites of control ddaC neurons are pruned at 16 h APF; however, dendrites failed to be pruned in α-ada RNAi ddaC neurons. The number of samples (n) in each group is shown on the bars. Scale bar represents 50 μm. Dorsal is up in all images. The individual numerical values for panels can be found in S1 Data. The genotypes can be found in S1 Text. α-ada, α-adaptin; APF, after puparium formation; mCD8-GFP, membrane-associated green fluorescent protein; RNAi, RNA interference; WP, white prepupal. Ada and Imac-RFP in ddaC axons. GFP-α-Ada puncta did not overlap with Imac-RFP in the axons. (D) Distribution of Venus-Prd1, GFP-α-Ada, and Imac-RFP in ddaC axons. Venus-Prd1, GFP-α-Ada, and Imac-RFP colocalized in ddaC axons (arrowhead). Scale bars represent 10 μm. The genotypes can be found in S1 Text. α-Ada, α-Adaptin; GFP, green fluorescent protein; Imac-RFP, immaculate connections fused with red fluorescent protein; Prd1, Pruning defect 1. Quantification of total length of unpruned ddaC dendrites at 16 h APF. The number of samples (n) in each group is shown on the bars. Error bars represent SEM. ÃÃÃ p < 0.001 as assessed by one-way ANOVA test. Scale bar represents 50 μm. The individual numerical values for panels A, B, and C can be found in S1 Data. The genotypes can be found in S1 Text. α-ada, α-adaptin; APF, after puparium formation; imac, immaculate connections; MARCM, mosaic analysis with a repressible cell marker; nrg, neuroglian; prd1, pruning defect 1; RNAi, RNA interference. The percentage of neurons with dendritic calcium transients was reduced at 6.5 h APF in prd1 M56 /Df(3R)Exel7310 and Imac G102E mutant neurons, compared to the control neurons. The number of samples (n) in each group is shown on the bars. The individual numerical values for panels can be found in S1 Data. The genotypes can be found in S1 Text. APF, after puparium formation; imac, immaculate connections; prd1, pruning defect 1. (DOCX) S1 Data. Data for Figs 1, 2, 4, 5, 6 and 8 and S1, S5, S8, S10, S14, S15, S18, S19, S20, S21, S23 and S25 Figs.