Rab2 drives axonal transport of dense core vesicles and lysosomal organelles

Long range fast axonal transport of neuropeptide-containing dense core vesicles (DCVs), endolysosomal organelles and presynaptic components is critical for maintaining the functionality of neurons. How the transport of DCVs is orchestrated remains an important unresolved question. The small GTPase Rab2 has previously been shown to mediate DCV biogenesis and endosome-lysosome fusion. Here we use the Drosophila model system to demonstrate that Rab2 also plays a critical role in bidirectional axonal transport of DCVs, endosomes and lysosomal organelles, most likely by controlling molecular motors. We further show that the lysosomal motility factor Arl8 is required as well for axonal transport of DCVs, but unlike Rab2 is also critical for DCV exit from cell bodies into axons. Our results uncover the mechanisms responsible for axonal transport of DCVs and reveal surprising parallels between the regulation of DCVs and lysosomal motility.


INTRODUCTION
Peptide neurotransmitters and hormones are critical for controlling behavior, metabolism, and growth. In neurons, neuropeptides released from both synaptic and extrasynaptic sites mediate direct neurotransmission, neuromodulation and endocrine functions. Neurons and secretory cells use secretory granules or dense core vesicles (hereafter referred to as DCVs) to store and transport neuropeptides and peptide hormones prior to regulated release. After synthesis in the endoplasmic reticulum (ER) and transit through the Golgi apparatus, neuropeptides and accessory proteins are packaged into immature DCVs that bud from the trans-Golgi network (TGN). This is followed by a maturation process involving exchange of material between immature DCVs and the endosomal system, whereby inappropriate proteins are removed while genuine DCV cargoes are retained (Gondre-Lewis et al., 2012;Topalidou et al., 2016).
Small GTPases of the Rab and Arf/Arl families regulate organelle identity, fusion and motility in eukaryotic cells (Donaldson and Jackson, 2011;Stenmark, 2009). They cycle between an inactive, GDP-bound, cytosolic state and an active, GTP-bound, membrane-associated state, in which they recruit specific effector proteins such as molecular tethers and motors to the membranes of their cognate organelles. In nematodes and mammals, the highly conserved Rab family GTPase Rab2 has an important role in DCV maturation. Together with its effector CCCP-1 and the endosome-associated recycling protein (EARP) tethering complex, Rab2 promotes retrieval of regulated secretory cargo from the endosomal system to maturing DCVs (Ailion et al., 2014;Laurent et al., 2018;Topalidou et al., 2016), and impairment of Rab2 or its effectors leads to the loss of DCV cargo to the endosomal system (Edwards et al., 2009;Sumakovic et al., 2009).
Recently, we and others reported that Rab2 is critical for biogenesis of mature degradative endoand autolysosomal compartments by controlling the fusion of late endosomes (LEs) with lysosomes and transport vesicles carrying lysosomal membrane proteins, and fusion of autophagosomes with endolysosomal organelles (Lorincz et al., 2017;Lund et al., 2018) (Ding et al., 2019).
Unlike small synaptic vesicles, neuronal DCVs cannot regenerate locally, and must be transported from the soma through axons (or dendrites) to their release sites. Generally, axonal transport relies on movement of organelles and protein complexes by molecular motors along axonal microtubules (MTs). Anterograde transport is mediated by the diverse family of plus-enddirected kinesin motors, while retrograde transport depends on cytoplasmic dynein. However, a high degree of interdependence exists between anterograde and retrograde transport (Barkus et al., 2008;Lim et al., 2017;Martin et al., 1999;Twelvetrees et al., 2016). Motor recruitment and regulation is mediated by adaptor proteins and small GTPases and is organelle-specific, producing different movement patterns for different organelles (Guedes-Dias and Holzbaur, 2019). In both mammals and invertebrates, axonal transport of DCVs is mediated by the Kinesin-3 KIF1A/unc-104 and Kinesin-1 (Barkus et al., 2008;Gumy et al., 2017;Lim et al., 2017;Lo et al., 2011), but how these motors are recruited and regulated is not well understood.
Axonal transport of endosomes and lysosomal organelles is also vital for neuronal development and function, partially through delivery of building blocks for axonal outgrowth and presynaptic biogenesis (Farias et al., 2017;Vukoja et al., 2018). Axonal transport of LEs and lysosomes critically depends on the BORC-Arl8 complex, which recruits Kinesin-1 or Kinesin-3 motors (Farias et al., 2017;Rosa-Ferreira and Munro, 2011;Wu et al., 2013). Perturbation of axonal transport can have devastating consequences for human health and has been implicated in a number of nervous system disorders, most notably neurodegenerative and neurodevelopmental conditions (De Vos et al., 2008;Guedes-Dias and Holzbaur, 2019).
Here we investigate the role of Rab2 in DCV biogenesis and transport in the Drosophila model system. We show that Drosophila Rab2 promotes DCV biogenesis and neuropeptide storage. Surprisingly, we also find that active Rab2 localizes to moving DCVs in axons and is required for fast and processive bidirectional axonal transport of DCVs, but not for DCV axonal entry from neuronal cell bodies or for synaptic capture and regulated exocytosis. Loss of Rab2 causes accumulation of static and slowly moving DCVs in the axonal compartment, and reduces the amount of DCVs at synaptic terminals. In line with this, Rab2 exhibits activity-dependent nanometer-range proximity to the Kinesin-3 unc-104 in vivo, suggesting that active Rab2 is part of a motor protein-containing molecular complex located on the surface of DCVs. We further find that the LE/lysosomal GTPase Arl8, but not Rab7, is also involved in axonal transport of DCVs. Unlike Rab2, however, Arl8 is both needed for axonal entry of DCVs from neuronal cell bodies and for their transport through the axon. Rab2 is also required for effective axonal transport of early endosomes (EEs), LEs, lysosomes, and lysosome-related active-zone transport vesicles. Our results demonstrate that the small GTPases Rab2 and Arl8 are essential regulators of axonal DCV transport, and reveal a novel role for Rab2 in the axonal transport of endosomes and lysosomes.

Rab2 is enriched in peptidergic neurons and required for DCV biogenesis
Detailed examination of the expression pattern of hemagglutinin-tagged Rab2 (HA-Rab2) controlled by endogenous Rab2 regulatory elements  in third instar (L3) larval brains revealed that Rab2 is enriched in somata of long-axon high-capacity peptidergic neurons expressing the DIMM transcription factor (Hamanaka et al., 2010;Park et al., 2008) (Fig. 1A-C). This is consistent with Rab2 being a transcriptional target of DIMM (Hadzic et al., 2015). Among the Rab2-enriched neurons were insulin-producing cells (IPCs) that synthesize and release insulinlike peptides (Ilp) 2, 3, and 5 to control growth and metabolism (Rulifson et al., 2002). A strong HA-Rab2 signal was detected in both IPC somata and axonal projections to the aorta and ring gland ( Fig. 1D). Compared to wild type, the Ilp2 content in IPC somata in normally fed larvae was substantially reduced (to one third or less) in Rab2 Δ1 null mutants (Fig. 1E, F) or when Rab2 was depleted by RNA interference (RNAi) (Fig. S1A, B). The Rab2 Δ1 phenotype in IPCs was rescued by expressing an mCherry-Rab2 transgene (Fig. 1E, F). Strikingly, the starvation-induced accumulation of Ilp2 in IPC somata of wild type larvae (Fig. 1G) (Geminard et al., 2009) did not occur in Rab2 Δ1 animals (Fig. 1G).
The reduction in Ilp2 content in Rab2 Δ1 mutants was not due to post-transcriptional effects, as similar decreases in IPC cell body content were observed in Rab2 nulls for GFP-tagged preproAtrial Natriuretic Factor (ANF-GFP), a DCV luminal cargo reporter (Rao et al., 2001), and for a GFPtagged variant of the DCV transmembrane protein IA-2 (IA2-GFP) when expressed in peptidergic cells driven by 386Y-Gal4 (Fig. S2A, B, D, E).
Microscopy and western blotting revealed that Rab2 Δ1 larvae exhibited a substantial reduction in whole-central nervous system (CNS) content of ANF-GFP expressed generally in peptidergic cells (386Y>ANF-GFP) ( Fig. S3A-C). We also observed a marked increase in the amount of a lower molecular weight ANF-GFP cleavage product in Rab2 Δ1 CNS samples (Fig. S3C), indicating that loss of Rab2 may alter neuropeptide processing. No changes were observed upon Rab2 overexpression (Fig. S3C), suggesting that Rab2 is not rate limiting for neuropeptide storage or processing in wild type. Moreover, direct stochastic optical reconstruction microscopy (dSTORM) demonstrated a small increase in DCV size in Rab2 Δ1 specimens (Fig. S3E, F), consistent with DCV size alterations commonly associated with defects in DCV biogenesis (Asensio et al., 2013;Grabner et al., 2006;Sumakovic et al., 2009). In large peptidergic somata HA-Rab2 tightly enveloped TGN clusters (Fig. S3G) where DCV biogenesis occurs (Gondre-Lewis et al., 2012).
Consistent with an important role of Rab2 in neuropeptidergic function, RNAi depletion of Rab2 specifically in peptidergic neurons resulted in either death prior to eclosion, or wing expansion defects ( Fig. S4A-C) associated with dysfunction of the protein hormone Bursicon (Dewey et al., 2004), secreted from DIMM-positive (Park et al., 2008), Rab2-enriched neurons (Fig.   S4D).
These findings indicate that Drosophila Rab2 is upregulated in peptidergic neurons and is involved in DCV biogenesis.

Rab2 is required for DCV axonal transport
A striking feature of the Rab2 Δ1 phenotype in larval brain IPCs was that, despite the dramatic reduction in Ilp2 somata, the Ilp2 signal was not reduced in the projections to the dorsal aorta and ring gland (Fig. 1E, G). This was observed in both starved and fed larvae (Fig. 1G), and for ANF-GFP expressed in IPCs ( Fig. S2A-C). Furthermore, the IA2-GFP signal was substantially increased in the aortal projection of Rab2 Δ1 larvae (Fig. S2D, F), despite a strong IA2-GFP reduction in IPC somata (Fig. S2D, E). Also, we observed focal Ilp2 accumulations in axonal shafts of Rab2 Δ1 IPCs, but not in wild type or rescued specimens (Fig. 1E, lower right insets). Similar results were obtained in Bursicon-containing neurons in the larval ventral nerve cord (VNC) that project commissural axons across the VNC midline, where a collateral branches off in each segment to form a local arbor ( Fig. 2A, asterisks). In Rab2 Δ1 larvae, the Bursicon content was strongly elevated in both the midline arbors and in focal accumulations in the commissural axonal shafts ( Fig. 2A, B). These observations indicate that neuropeptides suffer an axonal transport defect in Rab2 Δ1 animals.
To investigate DCV transport parameters in more detail, we used GFP-tagged Ilp2 (ILP-GFP) (Wong et al., 2012) as a DCV cargo marker, controlling its expression with the motor neuron driver OK6-Gal4. The Ilp2-GFP signal is substantially brighter than ANF-GFP, and examining motor neurons allowed us to differentiate between anterograde and retrograde transport. In Rab2 Δ1 larvae, focal accumulations of Ilp2-GFP appeared in axons near the roots of A1-7 nerves and along much of the A8/9 nerves. We therefore imaged the A7 nerve 0.5-1.0 mm from the VNC to avoid potential steric (traffic jam type) effects on transport. We photobleached a 60 µm long nerve segment on each side of a 10 µm central region, so that DCVs initially located in the unbleached center could be monitored with limited interference from unbleached DCVs entering from the sides (Fig. 3A, Video 1). While in Rab2 Δ1 mutants no DCV cargo accumulated in A7 motor axons (Fig. S6B, C), the proportion of static DCVs approximately tripled (Fig. 3A, B, Video 1), and DCV speed during both anterograde and retrograde runs (i.e., bouts of uninterrupted movement at constant velocity) was reduced (Fig. 3D). Moreover, pause durations and the relative time spent pausing (duty cycle) were substantially increased for dynamic DCVs in Rab2 Δ1 (Fig. 3E, F). As reported earlier (Barkus et al., 2008), most moving vesicles exhibited a preferred direction, either anterograde or retrograde, punctuated by pauses or shorter runs in the opposite direction. To estimate the processivity (length of uninterrupted movement between pauses) of axonal DCV transport, we divided the total run length in a particular direction by the number of pauses preceded by a run in that direction. The mean distance of uninterrupted movement in wild type -73.7 ± 5.7 µm (anterograde), 44.4 ± 2.1 µm (retrograde) -was significantly reduced to 42.8 ± 3.1 and 24.0 ± 3.1 µm, respectively, in Rab2 Δ1 animals ( Fig. S6A). Furthermore, in Rab2 Δ1 nerves markedly fewer DCVs traveled in both anterograde and retrograde directions, although retrograde transport was more severely affected Comparable results were obtained when imaging transport of ANF-GFP-positive DCVs in A8/9 nerves of 386Y>ANF-GFP larvae, although we could not differentiate between anterograde and retrograde transport due to the mixed orientation of the involved axons ( Fig. S5F-J).
When expressing Ilp2-GFP in pharate adult stage Bursicon-secreting neurons, and depleting Rab2 using RNAi, the axonal transport defect observed in Rab2 Δ1 larval motor neurons was essentially reproduced (Fig. S7A-H). A subset of these neurons, located in the VNC, project axons densely adorned with neurohormone-filled en passant boutons into the body cavity (Fig. S7A). We quantified DCV transport in single axonal shafts at a point near the VNC (Fig. S7A, yellow asterisk). Rab2 knockdown strongly reduced bidirectional flux, preferentially affecting the retrograde component (Fig. S7E, F). A reduction of the anterograde speed of transiting DCVs (Fig.   S7G) and a large increase in the proportion of static DCVs in the axon (Fig. S7H) was also observed. Overall, the Ilp2-GFP signal in the bouton-containing distal axons was reduced in Rab2depleted neurons (Fig. S7A Staining for endogenous Rab7, and live imaging of mCherry-Rab7, showed either no or only a modest overlap between LEs and ANF-GFP-marked DCVs in mid-nerve axonal shafts of Rab2 null larvae (Fig. S6D, E and Video 2). Therefore, the apparent alterations in DCV axonal transport are not caused by rerouting of DCV cargo to LEs, which are expected to have different transport characteristics than bona fide DCVs. Importantly, Vps39 Δ1 hemizygous null mutant larvae, in which LE-lysosome fusion is blocked as in Rab2 nulls (Lorincz et al., 2017;Lund et al., 2018), did not display the defects in transport of Ilp2-GFP-positive DCVs in motor axons that characterized Rab2 Δ1 animals ( Fig. 3B-E). Likewise, mutating the genes encoding the Rab2 effectors Golgin104 and BicD did not affect DCV transport ( Fig. 3B-E). Similarly, loss of these three Rab2 effectors did not recapitulate the Rab2 Δ1 -associated DCV transport phenotype in peripheral nerve axons of 386Y>ANF-GFP larvae ( Fig. S5G-J).
Therefore, the perturbed axonal transport in Rab2 nulls is not due to a block in the endo-lysosomal and autophagic pathways and also cannot be explained by the dysfunction of the DCV maturation process controlled by Rab2 through the Golgin104/CCCP-1 pathway. Moreover, the Rab2-binding dynein adaptor BicD does not play a major role in axonal transport of DCVs.
These data show that Rab2 is required for normal axonal transport of DCVs, but not for DCV entry from the cell body into the axon.

Rab2 is not important for stimulated release and synaptic capture of DCVs
Since Rab2 is involved in sorting of DCV components during DCV biogenesis (Ailion et al., 2014;Edwards et al., 2009;Sumakovic et al., 2009;Topalidou et al., 2016), we tested if Rab2 is required for stimulated exocytosis of DCV cargo. We stimulated OK6>Ilp2-GFP larval fillets by K + -induced depolarization and monitored both DCV cargo release and subsequent replenishment in muscle 3 type Ib NMJs as changes in synaptic Ilp2-GFP fluorescence. As for ANF-GFP (Fig. S5A, B), the Ilp2-GFP content of muscle 3 type 1b terminals of Rab2 Δ1 mutants was substantially reduced (Fig.   4B, C), likely due to decreased anterograde DCV flux in the distal axon (Fig. 4A). Also, terminals often appeared morphologically aberrant ( Fig. 4B and Fig S5A) (Mallik et al., 2017). However, both the proportional DCV cargo release upon stimulation and the subsequent partial recovery were similar in mutant and wild type terminals, although the relative release amplitude tended to be somewhat lower in Rab2 Δ1 (Fig. 4D, E). Hence, Rab2 is neither critically required for stimulated exocytosis of DCVs, nor for the fast post-stimulus recovery of the synaptic DCV pool that relies on activity-induced capture of anterograde circulating DCVs (Cavolo et al., 2016;Shakiryanova et al., 2006).
Without stimulation, DCVs circulate between the axon initial segment and terminal boutons, with a low probability of being captured when passing through individual en passant boutons (Wong et al., 2012). In Rab2 Δ1 larvae, Ilp2-GFP-positive DCVs in both en passant and terminal boutons of muscle 3 type Ib terminals did not display increased synaptic capture ( Fig. S8A-C) or decreased retrograde movement from the terminal bouton ( Fig. S8D-F), suggesting that the increase in static DCVs in axonal shafts and the deficit in retrograde DCV flux in Rab2 nulls is not due to enhanced baseline DCV capture. Furthermore, the single-vesicle Ilp2-GFP intensity was preserved in Rab2 nulls, suggesting that the decreased DCV cargo content in Rab2 Δ1 mutant terminals is due to lower vesicle (quantal) content in boutons rather than reduced cargo loading in individual vesicles (Fig. S8G).
These findings show that Rab2 is neither critically required for regulated exocytosis of DCVs, nor does it play a major role in synaptic capture.

Rab2 controls axonal transport of endosomes and lysosomal organelles, but not mitochondria
To test if Rab2 exclusively regulates axonal transport of DCVs, we examined the distribution and motile characteristics of mitochondria and organelles belonging to the endolysosomal family. We detected no difference in axonal transport of mito-GFP-labeled mitochondria between Rab2 Δ1 and wild type controls (Fig. S9A, B). Also, the axonal tubulin cytoskeleton was unperturbed when visualized by immunolabeling the axonal microtubule-binding protein Futsch (Hummel et al., 2000), both in terms of staining intensity and microtubule continuity in the distal axon and presynaptic arbor of muscle 3 1b terminals (Fig. S9C).
Rab2 is closely associated with the function of the endolysosomal system and is present on the membranes of endosomes and lysosomes, albeit transiently or in very low abundance in most cell types (Gillingham et al., 2014;Lorincz et al., 2017;Lund et al., 2018). Accordingly, both EEs, marked by GFP-Rab5 and LEs, marked by GFP-Rab7, showed markedly reduced motility in peripheral motor axons in the absence of Rab2. Wild type axons exhibited both a brightly labeled static and a considerable, but more weakly labeled, dynamic population of EEs undergoing bidirectional transport (Fig. 5A, Video 3). In Rab2 Δ1 larvae, both the anterograde and retrograde dynamic EEs were almost entirely absent, while the proportion of static GFP-Rab5-positive organelles was strongly increased (Fig. 5A, Video 3). A large population of motile LEs was also present in wild type axons, in addition to a minor static component. In contrast to EEs, however, LEs in controls exhibited a significant directional bias towards retrograde transport (Fig. 5B, Video 4), as described in other systems (Cheng et al., 2015;Ferguson, 2018). Rab2 loss increased the proportion of static LEs, decreased the number and average velocity of motile LEs, and strongly reduced the retrograde transport bias (Fig. 5B, Video 4). However, the overall effect was less dramatic than for EEs.
Next, we examined the axonal motility of the lysosomal transmembrane marker Spinster-GFP Presynaptic components including active zone proteins are delivered to the synapse in nondegradative lysosome-related vesicles positive for transmembrane lysosomal markers such as Spinster . We tested if axonal transport of a GFP-tagged variant of the active zone scaffold Bruchpilot (brp-GFP) is perturbed in Rab2 Δ1 larvae. In wild type L3 motor axons, relatively few brp-positive organelles were present. Those that were motile (43.5 %) exhibited a pronounced anterograde bias, consistent with their role in presynaptic biogenesis (Fig. 5D). Loss of Rab2 substantially increased the proportion of static brp-organelles and markedly reduced the mobility of the remaining dynamic organelles, but did not affect the anterograde transport bias (Fig.   5D).
These results show that, while Rab2 is not universally required for axonal transport, it regulates axonal transport of both DCVs and organelles belonging to the endolysosomal family, including specialized derivatives responsible for delivery of presynaptic components.

Rab2 is present on moving DCVs
If Rab2 is directly involved in axonal transport of DCVs and endolysosomal organelles, either as a motor adaptor or as a regulator of other factors coupling to motors or microtubules, one would expect it to be present in the axono-synaptic domain and more specifically on the membranes of the transported organelles. HA-Rab2 was highly abundant in IPC axons and neurohemal release zones that harbor many Ilp2-containing DCVs (Fig. 1D) and was likewise present in axonal shafts located in larval peripheral nerves (data not shown). However, with standard confocal live imaging GFP-Rab2 appeared to be largely diffusely distributed in peripheral nerves, with occasional strongly labeled immobile organelles. We reasoned that if Rab2 is present on the membranes of moving DCVs it might be in very low abundance or very transiently, similar to its fleeting interaction with endosomal membranes (Lorincz et al., 2017;Lund et al., 2018). Moreover, its presence on the DCVs would be obscured by cytosolic Rab2 in the axoplasm and in the surrounding glial sheath. To visualize such low-level recruitment, we bleached a segment of peripheral nerve expressing fluorescently labeled Rab2 to eliminate signal from unbound cytosolic protein. This was followed by confocal imaging with a high pixel dwell time and a high degree of averaging to detect small amounts of Rab2 on motor-driven organelles entering the bleached region before diffusion of cytosolic material could "catch up". Due to the slow scanning speed, moving vesicular organelles appeared as streaks with slopes inversely proportional to the velocity of movement, similar to kymographic representations (Fig. 6A). We observed a high density of GFP-Rab2-positive vesicles moving in axons with speeds of up to ~3 µm/s (Fig. 6A, top). Similar motile vesicles were also brightly labeled with neuronally expressed gain-of-function GTP-locked pHluorin-Rab2 Q65L (Fig.   6A, bottom) or wild type pHluorin-Rab2 (Fig. 6B). In contrast, very few moving vesicles were labeled with inactive GDP-locked pHluorin-Rab2 S20N (Fig. 6A, middle), demonstrating that recruitment of Rab2 to organelles undergoing axonal transport depends on its activation state.

Rab2 exhibits activity-dependent nanometer-range proximity to unc-104
To detect if active Rab2 forms a molecular complex with motor proteins responsible for axonal transport of DCVs in vivo, we employed proximity-dependent biotinylation. This approach relies on a promiscuous biotin ligase conjugated to a bait protein to biotinylate closely neighboring prey proteins, which can then be purified and detected. With the original BirA ligase this technique has a documented labeling radius of ~10 nm (Kim et al., 2014) and is advantageous over classical coaffinity purification by allowing for detection of transient or weak molecular interactions. We found that a neuronally expressed fusion of the enhanced BirA variant TurboID (Branon et al., 2018) to gain-of-function Rab2 Q65L (TurboID-Rab2 Q65L ) biotinylates the main anterograde Drosophila DCV motor (Barkus et al., 2008;Lim et al., 2017), the Kinesin-3 unc-104 (Fig. 6C). In contrast, very little biotinylation of unc-104 was observed using wild type TurboID-Rab2, and essentially no biotinylation was observed using loss-of-function TurboID-Rab2 S20N or in control animals not carrying any transgenes (Fig. 6C). These results suggest that active Rab2 is part of a molecular complex that includes the anterograde DCV motor unc-104.

Arl8, but not Rab7, is required for axonal transport of DCVs
The finding that the small GTPase Rab2 regulates biogenesis and axonal transport of both endolysosomal vesicles and DCVs may imply a connection between these organelles. This prompted us to ask if other small GTPases involved in the function of late-stage compartments of the endolysosomal pathway also play a role in axonal transport of DCVs. Apart from Rab2, fusion and microtubule-based motility of LEs and lysosomes is controlled by the small GTPases Rab7 and Arl8 (Boda et al., 2019;Farias et al., 2017;Johansson et al., 2007;Jordens et al., 2001;Khatter et al., 2015;Marwaha et al., 2017;Pankiv et al., 2010;Raiborg et al., 2015;Rosa-Ferreira and Munro, 2011;Rosa-Ferreira et al., 2018). Strikingly, the number of Ilp2-GFP-marked DCVs in peripheral nerve motor axons and terminals of Arl8 e00336 null larvae was severely reduced, with the remaining axonal DCVs being predominantly static (Fig. 7A, 3B-E, Video 6). The loss of axonal and synaptic DCVs was most likely due to a failure in the ability to exit the cell bodies, as Ilp2-GFP was still abundant in the VNCs of Arl8 mutant animals (Fig. 7B). This mirrors the behavior of lysosomerelated presynaptic carriers in Arl8 nulls  and resembles the disruption of the DCV distribution in loss-of-function mutants for the Arl8-linking Wu et al., 2013) unc-104 motor (Barkus et al., 2008;Lim et al., 2017;Pack-Chung et al., 2007). The few axonal DCVs that were still mobile exhibited strongly reduced run speeds, dramatically extended pause durations, and severely reduced processivity ( Fig. 7A and Fig. 3D-E, Fig. S6A). Interestingly, despite the severity of the Arl8 e00336 phenotype, we did not observe the relative reduction in retrograde DCV flux which characterized Rab2 Δ1 mutants. In contrast to Arl8, axonal transport of DCVs in Rab7 ∆1 null larvae was similar to wild type ( Fig. 7A and 3B-E, Video 6). This further confirms that the observed DCV transport phenotypes in Rab2 and Arl8 mutants are not a secondary consequence of blocked fusion of LEs and autophagosomes with lysosomes.
In light of Arl8's critical role in axonal transport of DCVs, and the general assumption that Arl8 is primarily a lysosomal GTPase, we sought to establish its localization in relation to axonal DCVs. Imaging of peripheral nerve axons using the same method as that used to visualize fluorescently tagged Rab2 on moving organelles revealed neuronally expressed Arl8-GFP to be present on transiting ANF-mCherry-positive DCVs (Fig. 7C).
The fact that the late endosomal/lysosomal small GTPases Rab2 and Arl8 control axonal transport of both DCVs and lysosomes (Farias et al., 2017;Vukoja et al., 2018), as well as the close ultrastructural similarity of lysosome-related presynaptic carriers to classically described DCVs , raises the question whether these organelles are to some extent identical. Coimaging of RFP-Spinster with Ilp2-GFP did indeed reveal some instances of transiting axonal DCVs clearly labeling with this lysosomal marker (Fig. S10A), but these events were rare.
Likewise, brp-GFP and ANF-mCherry only coincided in a minor proportion of transport events involving active zone transport vesicles (Fig. S10B).
Collectively, these findings show that Arl8 is present on DCVs and is critically required for their axonal transport, similar to Arl8's role in the transport of lysosomes and lysosome-related presynaptic carriers. However, DCVs are for the most part distinct from lysosomal organelles including those responsible for presynaptic biogenesis.

DISCUSSION
DCVs are specialized organelles that underpin virtually all cellular signalling employing peptide hormones and neurotransmitters. This mode of signalling is especially prevalent in the nervous system, making highly critical the long-range motility of DCVs through axonal projections, from where most DCV exocytosis takes place (Persoon et al., 2018). However, the mechanisms controlling cytoskeletal transport of DCVs to their targets remain underexplored, especially regarding the upstream regulatory proteins that control molecular motors. Here we identify the small GTPases Rab2 and Arl8 as critical regulators of axonal DCV transport (Fig. 7d).
In addition to controlling DCV transport, we demonstrate that Rab2 is also required for efficient axonal transport of endo-lysosomal organelles. Likewise, Arl8 is a central regulator of lysosomal transport (Farias et al., 2017;Rosa-Ferreira and Munro, 2011;Vukoja et al., 2018). This illustrates an intriguing parallel between the biology of DCVs and lysosomes, reinforced by the fact that Rab2 is involved the biogenesis of both lysosomes (Lorincz et al., 2017;Lund et al., 2018) and DCVs (Edwards et al., 2009;Sumakovic et al., 2009;Topalidou et al., 2016). This and other similarities in the biogenesis of DCVs and lysosomes, such as the shared requirement for AP-3, Vps41 and BLOC-1 (Asensio et al., 2010;Asensio et al., 2013;Hao et al., 2015), may imply an evolutionary relationship between DCVs and lysosomal organelles. As Arl8 is also required for the biogenesis of lysosomes (Khatter et al., 2015;Lund et al., 2018), an interesting question for future studies would be if Arl8 plays a role in the formation or maturation of DCVs.
The picture emerging from our findings is that Rab2 exerts extensive control over DCV biology, sustaining both their formation and subsequent intracellular transport. The recruitment of Rab2 to maturing DCVs could conveniently position this GTPase for regulation of the transport motor complex. The fact that Rab2 and other transport-associated proteins such as unc-76 are targeted by the DIMM transcription factor (Hadzic et al., 2015) suggests that the cellular differentiation program initiated by DIMM to establish the high-capacity, far-projecting secretory phenotype in neurons (Hamanaka et al., 2010) includes upregulation of the axonal transport machinery to match the logistic challenge of supplying DCVs to distal release sites.
Importantly, our work also demonstrates that Arl8 and Rab2 have distinct functions in the orchestration of DCV transport. While Arl8 is critically required for driving DCV exit from somata and is a prerequisite for axonal motility, Rab2 supports bidirectional transport of circulating DCVs in the axonal shaft, especially in the retrograde direction. When driving lysosomal transport, Arl8 is known to recruit and activate Kinesin-3/unc-104 (Guardia et al., 2016;Vukoja et al., 2018;Wu et al., 2013), which is also the main anterograde DCV motor (Barkus et al., 2008;Lim et al., 2017;Lo et al., 2011). Consistent with this notion, the unc-104 loss-of-function DCV transport phenotype closely resembles that which we observe in Arl8 nulls (Barkus et al., 2008). Our results indicate that active Rab2 is likely involved in recruiting or activating unc-104 by participating in a molecular complex with this motor. However, the disproportionate effects of manipulating Rab2 on retrograde DCV transport also suggest that it may regulate DCV-related dynein function, consistent with previous work linking mammalian Rab2 to recruitment of dynein to ER-Golgi intermediates (Tisdale et al., 2009).
Dysfunction of human Rab2A has been linked to autism spectrum disorders (ASD) (Iossifov et al., 2012;Sanders et al., 2012;Takata et al., 2016), a set of high-heritability neurodevelopmental conditions associated with perturbations in the pro-reproductive and pro-social oxytocin (OT) and Arginine Vasopressin (AVP) neuropeptide systems (Meyer-Lindenberg et al., 2011). OT and AVPproducing neurons in the mammalian hypothalamus extend axonal projections to forebrain regions to regulate complex social behaviors. In our experiments, Rab2 loss in flies depleted presynaptic DCV stores due to accumulation of stalled DCVs in the axon and reduced transport of active zone components to synaptic terminals. It is tempting to speculate that similar transport defects contribute to the ASD phenotype in individuals carrying deleterious Rab2A alleles.
In summary, our results uncover the mechanisms controlling axonal transport of DCVs, revealing surprising parallels between DCV and lysosomal motility.

Fly husbandry, genetics, and stocks
Flies were reared on Nutri-Fly™ Bloomington Formulation medium (Genesee Scientific, San Diego, California) at 26°C, except for RNAi experiments, where they were reared at 29°C. In RNAi experiments, UAS-Dicer-2 was co-expressed with the RNAi transgene. In Rab2 ∆1 mutant experiments, larvae were reared on apple juice plates supplemented with yeast paste. For starvation experiments, L1 larvae aged 24-36 h after egg deposition were transferred to yeasted apple juice plates. After 48 h they were transferred to amino acid-poor starvation plates (agar, PBS, 1% (w/v) sucrose) (10-12 larvae/plate) and kept there for 24 h, before being dissected for immunohistochemistry. Fed control larvae were kept on yeasted apple juice plates. The full genotypes of animals used in experiments are listed in Table S1.
All DNA injections, selection of P-element and Phi31C transformants and selection of CRISPR/Cas9 knockouts (see below) was performed by BestGene Inc.. The 2xHA-TurboID-tagged transgenes encoding Rab2, Rab2 S20N and Rab2 Q65L were inserted into the M{3xP3-RFP.attP}ZH-86Fb attP site on the third chromosome using Phi31C transformation to ensure equal levels of expression.
The Golgin-45 ∆1 , CG3703 ∆1 and BicD ∆1 knockout lines were generated using CRISPR/Cas9catalyzed homology-directed repair according to the method outlined in (Gratz et al., 2014). In brief, two gRNA-expressing DNA constructs targeted near each end of the coding sequence of the genes were injected into yw; nos-Cas9(II-attP40) embryos expressing Cas9, together with a donor DNA construct based on the pHD-DsRed-attP plasmid. As a result, most of the coding sequence of the genes (X: 945355-942705 for CG3703, 3R: 9403771-9402693 for CG9356/Golgin-45, 2L: 17460549-17465046 for BicD) was replaced with a DsRed transgene. For Golgin104 we used one gRNA-construct together with a matching homology-directed repair construct to target a DsRed transgene for insertion into the second coding exon of Golgin104 between positions 3L:16350798-16350799, thus truncating the C-terminal 496 residues of the encoded protein, including the Cterminal Rab2-binding domain (Cattin-Ortola et al., 2017;Gillingham et al., 2014). Flies homozygous for the resulting Golgin104 ins1 allele had reduced fecundity and were perceptibly smaller than wild type flies. The BicD ∆1 allele, in combination with the overlapping deficiency Df(2L)Exel7068, was hemizygously semi-lethal at the pupal stage, with 53 out of 199 tested pupae failing to eclose. We obtained similar results with hemizygotes over the same deficiency carrying the point nonsense-mutation allele BicD r5 , previously described as amorphic (Ran et al., 1994).

Molecular biology
To generate the RNAi-resistant UAS-Rab2 RNAi-insens transgenic construct, a Rab2 coding region cassette was synthesized by GenScript (Nanjing, China) in which all degenerate codons were changed into synonymous codons by base pair substitution. This cassette was amplified and cloned into the pUAST vector. pUAST-pHluorin-Rab2 S20N was generated by GenScript by mutagenizing the pre-existing pUAST-pHluorin-Rab2 construct . To generate the UAS-IA2-GFP transgenic construct, we first PCR-amplified the EGFP coding region from the pEGFP-N2 plasmid and inserted it into the pUAST vector. We then amplified the IA2 coding region from cDNA clone GH05223 and inserted it into the resulting plasmid to obtain the final pUAST-IA2-GFP construct. The pUAST-ANF-mCherry construct was generated by GenScript by synthesizing the ANF-mCherry coding region and inserting it into the pUAST vector. Likewise, the pUASTattB-2xHA-TurboID-Rab2, pUASTattB-2xHA-TurboID-Rab2 S20N and pUASTattB-2xHA-TurboID-Rab2 Q65L constructs were generated by GenScript.

In vivo biotinylation and purification of biotinylated proteins
Flies expressing TurboID-Rab2, TurboID-Rab2 S20N or TurboID-Rab2 Q65L under control of the panneuronal elav-Gal4 driver and w 1118 control flies were reared on Nutri-Fly™ Bloomington Formulation medium supplemented with 100 µM Biotin. Adult flies were collected 0-3 days after eclosion, flash-frozen in liquid nitrogen and stored at -80°C.
For purification of biotinylated proteins, 1-2 ml of frozen flies were transferred to pre-cooled dounce homogenizers on ice, quickly dounced 5x, then dounced 15x in 3 ml RIPA lysis buffer (

Western Blotting
For western blotting of ANF-GFP, late L3 larval CNS with the ring gland and aorta attached were  Table 1. Localizations were fit using the program ThunderSTORM (Ovesny et al., 2014), where localizations were detected with a threshold of 1.5 * standard deviation (Wave.F1) using the Local Maximum method. Drift correction was then performed with cross correlation. Images were generated by rendering the normalized Gaussian image with a 30x magnification through ThunderSTORM.

Confocal microscopy
All confocal microscopy was carried out at the Core Facility for Integrated Microscopy To examine DCV transport in peripheral nerves of larvae expressing OK6>Ilp2-GFP we chose a proximal region on the A7 nerve ~0.5-1.0 mm from its exit from the VNC (or, in larvae expressing 386Y>ANF-GFP, on the A8/9 nerve ~0.5 mm from its exit). If required, the dissection dish was elevated on one side with Parafilm strips to orient the nerves horizontally. After fitting a 29 130 x 5.6 µm acquisition window (56 x 15 µm for 386Y>ANF-GFP) to the nerve, all fluorescent signal except a central 10 µm nerve segment was photobleached once and time series were recorded with a frame rate of ~5 frames per second (fps). A similar approach was used to examine the axonal transport of mito-GFP, GFP-Rab5, GFP-Rab7, spin-GFP and brp-GFP, though no bleaching was To examine DCV dynamics in NMJs, muscle 3 Ib terminals in segments A3-5 og filleted L3 larvae expressing OK6>Ilp2-GFP were imaged employing a wide optical slice (2.1 µm, corresponding to 1.85 Airy units) and a frame rate of 1.1 fps. To measure capture, the entire terminal was photobleached, followed by a pause of a duration manually adjusted such that 2-9 vesicles were allowed to enter the terminal from the distal axon. The terminal was then imaged (800 frames) while photobleaching the distal axon every 20th frame to bleach any new vesicles entering the terminal. The duration of each bleach cycle was 1-1.5 s depending on the shape of the bleach box. To measure the efflux of vesicles from the most distal bouton, the entire terminal and distal axon except for the distal bouton was photobleached, followed by time-lapse microscopy at settings identical to those described above, while photobleaching the distal axon every 20 frames.
For neuropeptide release experiments, L3 larvae expressing OK6>Ilp2-GFP were dissected in Ca 2+ -free saline (128 mM NaCl, 2 mM KCl, 4 mM MgCl 2 , 35.5 mM sucrose, 0.5 mM EGTA and 5 mM Na HEPES, pH 7.2) and Ilp2-GFP fluorescence was then recorded in Ib terminals of muscle 3 NMJs in segments A3-5 as fast confocal stacks. Release was evoked by depolarization obtained by replacing the bath solution with high-K + saline (130 mM KCl, 1.8 mM CaCl 2 , 4 mM MgCl 2 , 35.5 mM sucrose and 5 mM Na HEPES, pH 7.2) for 3 min, followed by replacement of bath solution with Ca 2+ -free saline to terminate the stimulus and allow for muscle relaxation. Confocal stacks were recorded immediately before stimulation and immediately after the washout of the high-K + saline (0 min), as well as at 5, 10 and 15 minutes after termination of stimulation.

Image processing and analysis
FRAP experiments. To quantify the FRAP of bleached peripheral nerves in larvae expressing 386Y>ANF-GFP we measured the fluorescence intensity in the bleached area and performed double normalization of the raw intensity (I raw ) to the pre-bleach intensity (I pre ) and to two non-bleached segments of nerve (I cont ) to control for bleaching during acquisition, while subtracting the baseline intensity (I base ) defined as the intensity of the bleached area immediately after bleaching (t = 0): A double exponential decay equation was fit to the data using the ImageJ curve fitting tool.
Mid-nerve axonal transport. Trajectories of trafficking fluorescent organelles entering the bleached areas were traced manually on kymographs by drawing them as segmented lines in Fiji/ImageJ (Schindelin et al., 2012). Since the complexity caused by the many trafficking organelles generally Visibility-enhancing image processing. Since the mCherry-Rab7 signal (Video 2) suffered heavily from bleaching during imaging of axonal transport, the contrast of each separate frame was adjusted by saturating 0.35% of the pixels before converting the series to a movie. The kymographs from time lapse imaging of OK6>Rab5-GFP, OK6>Rab7-GFP, and OK6>spin-GFP transport (Fig. 5) exhibited a steep bilateral intensity gradient, causing excessive saturation when contrasting the images to visualize the faint trajectories in the center of the images. Structures larger than 100 pixels were therefore filtered using the FFT bandpass Plugin in Fiji before producing the Figure. The red channel ( Table S1.   Quantification of (B). Student's t-test. (D) False-color coded GFP fluorescence intensity before, immediately after, and five minutes after high K + -induced Ilp2-GFP release. For each genotype, pixel values were normalized to the maximum intensity across images (100%, red). (E) GFP fluorescence relative to the pre-stimulus level. Two-factor ANOVA. Data from 14 control and 13 Rab2 ∆1 larvae. In (A) and (C), data are mean ± s.e.m., and the number of larvae analyzed is indicated inside bars. Scale bars: (B), 10 µm; (D), 5 µm.  Video 4. Rab2 loss disrupts late endosome axonal transport. Live confocal imaging illustrating the reduced transport of OK6>GFP-Rab7-positive LEs in Rab2 null motor neuron axons. The lateral regions of the imaged segment of the A7 peripheral nerve were photobleached after the first frame as in (Fig. 3A) to enhance the visibility of moving organelles. Same experiments as shown in (Fig.  5B, left). Ant, anterograde (towards synaptic terminals); Ret, retrograde (towards neuronal cell bodies).
Video 5. Rab2 loss disrupts axonal transport of lysosomes and lysosome-related vesicles. Live confocal imaging illustrating the reduced transport of OK6>Spinster-GFP-positive lysosomes and lysosome-related vesicles in Rab2 null motor neuron axons. The lateral regions of the imaged segment of the A7 peripheral nerve were photobleached after the first frame as in (Fig. 3A) to enhance the visibility of moving organelles. Same experiments as shown in (Fig. 5C, left). Ant, anterograde (towards synaptic terminals); Ret, retrograde (towards neuronal cell bodies).
Video 6. Arl8, but not Rab7, is critical for axonal transport of DCVs. Live confocal imaging illustrating the disrupted transport of OK6>Ilp2-GFP-positive DCVs in motor neuron axons of Arl8 null, but not Rab7 null larvae. The lateral regions of the imaged segment of the A7 peripheral nerve were photobleached after the first frame as in (Fig. 3A) to enhance the visibility of moving organelles. Ant, anterograde (towards synaptic terminals); Ret, retrograde (towards neuronal cell bodies).