The Drosophila LC8 homolog cut up specifies the axonal transport of proteasomes

Proteasomes are large protein complexes responsible for degradation of normal short-lived ubiquitylated proteins as well as mutant, misfolded or damaged proteins (Rock et al., 1994). All cells require regulated protein degradation; however, nerve cells are of particular interest due to their complex compartmentalization and the requirement of protein degradation for normal neuronal function (Broadie et al., 1999; Ding and Shen, 2008; Yi and Ehlers, 2005). In addition, a large number of neurological disorders are characterized by accumulations of proteinaceous aggregates, suggesting that impaired protein degradation is an important disease etiology of many neurodegenerative diseases (Bingol and Sheng, 2011; Tai and Schuman, 2008). Because protein degradation in neurons occurs on short timescales and is highly compartment specific, neurons must possess molecular mechanisms that can precisely position proteasomes near to where they are uniquely required, but also maintain a physical separation between proteasomes and neuronal targets to preserve the efficacy of regulated protein turnover.

Numerous studies have demonstrated the important physiological requirement of proteasome activity throughout the many compartments of the neuron. Pre-synaptically, proteasome-dependent protein degradation is critical for synapse formation, synaptic efficacy and neurotransmitter release (Collins et al., 2006; DiAntonio et al., 2001; Speese et al., 2003; Wan et al., 2000; Willeumier et al., 2006). Post-synaptically, proteasomes have been implicated in regulating several forms of synaptic plasticity including long-term potentiation (LTP), long-term facilitation (LTF) and long-term depression (LTD) (Dong et al., 2008; Ehlers, 2003; Hegde, 2010, 2004; Hegde et al., 2014; Zhao et al., 2003). Furthermore, acute depolarization of neurons causes a global change in ubiquitylated active zone proteins at the synapse, supporting the role of proteasomes in the rapid turnover of proteins in response to neuronal activity (Aravamudan and Broadie, 2003; Broadie et al., 1999; Tai and Schuman, 2008). Collectively, these data suggest that proteasomes function locally at pre- and post-synaptic sites where they act as an important modulators of synaptic structure, function and plasticity (Collins et al., 2006; DiAntonio et al., 2001; Ding and Shen, 2008; Dong et al., 2008; Hegde, 2010, 2004; Speese et al., 2003; Tai and Schuman, 2008; Wan et al., 2000; Willeumier et al., 2006; Yi and Ehlers, 2005; Zhao et al., 2003).

In addition to the synaptic compartments, there is evidence that proteasome function also plays an important role in the growth, development and regeneration of axons. Recent work on neuronal development has shown that changes in retrograde axonal transport of proteasomes are critical during the specification and growth of the axon (Hsu et al., 2015). Studies in Drosophila provide evidence that the degeneration of axons that occurs during developmental pruning or in response to injury requires the ubiquitin–proteasome system (UPS) (Watts et al., 2003; Xiong et al., 2012). Consistent with these observations in flies, inhibition of the UPS in rodent models delays the axonal die-back observed during Wallerian axonal degeneration demonstrating a role for protein degradation during programmed axonal degeneration in mammals (Hoopfer et al., 2006; MacInnis and Campenot, 2005; Zhai et al., 2003). These data provide strong evidence for the evolutionarily conserved requirement of proteasome activity within the axon under both normal and pathological conditions.

Despite the critical compartment-specific requirements for proteasome function in neurons, little is known about the molecular mechanisms that govern proteasome transport and their targets within neurons. The trafficking of organelles and transport vesicles within all cells is predominantly mediated by microtubule (MT)-based transport mechanisms utilizing two distinct molecular motor proteins, kinesins and cytoplasmic dyneins. Kinesins mostly mediate MT plus-end-directed transport, including anterograde axonal transport in neurons (Maday et al., 2014). In the human genome, 45 genes code for the kinesin superfamily, supporting a genetic basis for the large diversity of cargo-specific kinesin-based transport events. Cytoplasmic dynein mediates MT minus-end-directed transport including retrograde axonal transport. However, unlike for the kinesin motor, cytoplasmic dynein is encoded by relatively few genes leading to the hypothesis that the cargo specificity of the dynein motor complex is accomplished by the heterogeneneity of dynein complex subunits and various dynein-associated accessory proteins. The LC8 dynein light chains (DYNLL1 and DYNLL2 in mammals), have been proposed as cargo-adaptors potentially providing specificity for the minus-end-directed MT transport of vesicles and organelles. This notion was supported by studies that found LC8 to simultaneously associate with the dynein motor and with a number of cargos that undergo MT-mediated transport (Cianfrocco et al., 2015; Fejtova et al., 2009; Lee et al., 2006; Navarro et al., 2004; Schnorrer et al., 2000).

In the current study, we use a combination of genetics, biochemistry and in vivo imaging to compare the MT-based transport of proteasomes and synaptic proteins in Drosophila motor neurons. These analyses found that proteasomes use MT-based axonal transport in axons and that the axonal transport is qualitatively similar to that of synaptic proteins. However, quantitative analysis of proteasome trafficking reveals significant differences in the retrograde transport of proteasomes compared to that of synaptic proteins. These data suggests potential molecular differences in the dynein motor complexes utilized by these two distinct cargo types. In support of this idea, a forward genetic screen identified the cut up (ctp) gene, a Drosophila homolog of LC8, as being required specifically for the axonal transport of proteasomes but not synaptic proteins. These results provide molecular evidence that proteasomes and their targets utilize specific dynein motor components during MT-based transport in neurons.

To characterize axonal transport of proteasomes in Drosophila neurons, we generated a transgenic fly line expressing the proteasome β5 subunit tagged with a red fluorescent protein under the control of Gal4 (prosβ5-RFP) (Fig. 1A). Neuronal expression of the prosβ5-RFP transgene using the Gal4:UAS binary expression system followed by live-imaging of Drosophila third-instar larvae revealed a large population of moving proteasomes within the Drosophila axonal tracts (Fig. 1B and C). To determine if these neuronal proteasomes are trafficking on MTs we co-expressed a green fluorescent protein-tagged tubulin (tubulin-GFP) and prosβ5-RFP in all neurons and performed continuous live-imaging in the green and red channels. The inspection of the acquired videos is consistent with the hypothesis that the proteasomes traffic along MTs (Fig. 1B). Mutations in the gene encoding the p150 dynactin 1 subunit (DCTN1-p150 in flies; DCTN1 or p150^glued in mammals) encoding an obligatory subunit of the dynactin complex have been previously shown to impair dynein-dependent axonal transport in mice and flies providing a diagnostic for MT-based transport (Chang et al., 2013; Duncan and Goldstein, 2006). To investigate the requirement for the dynactin complex for the axonal transport of proteasomes, we combined live imaging of prosβ5–RFP particles and genetic inhibition of dynactin complex function. For this analysis, we used a transgene:driver combination that in our hands results in the strongest inhibition dynactin complex function but still generates viable larvae [C155-Gal4; UAS-DCTN1-P150Δ (Eaton et al., 2002)]. Qualitative analysis of movies generated by the live imaging of prosβ5–RFP particles in axons revealed that 64% of the total population of particles traveled in one continuous direction and only 3% of the particles were stationary over the course of a recording (Fig. 1D, black bars). The remaining particles (33%) exhibited a frequent reversing behavior without considerable progression in either direction (Fig. 1C; arrowheads). We observed a highly significant increase in the number of stationary particles in animals harboring a dominant mutation in DCTN1-P150 (denoted DNGlued) compared to controls (Fig. 1D). Particle tracking of proteasome movement from the live imaging sessions was used to generate individual particle trajectories, which were broken into segments defined by uninterrupted periods of continuous movement for ten or more frames. It should be noted that the C155-Gal4 drives in both motor and sensory neurons making it impossible to assign directionality of the movements in these analyses. These studies revealed that proteasome transport is highly processive, consistent with MT-based transport. Furthermore, the values determined by this analysis for segmental velocity and run lengths are consistent with transport by MT motors (Fig. 1E,F, black bars) suggesting that neuronal proteasomes use fast axonal transport (Duncan and Goldstein, 2006; Maday et al., 2014). Inhibition of dynactin complex function significantly reduced both proteasome run length and velocity (Fig. 1D–F, gray bars). It should be noted that because the Gal4 driver used in these studies expresses in all neurons, we cannot be sure of the directionality of movement and therefore are likely underestimating the effect of DCTN1 mutations on velocity. These data support the idea that trafficking of proteasomes within larval motor neurons utilizes MT-based transport mechanisms requiring normal dynactin complex function.

We used standard biochemical approaches to verify that our tagged Prosβ5–RFP assembled into functional proteasomes. First, glycerol gradient fractionation of flies expressing prosβ5-RFP in the nervous system followed by immunoblot analysis of the eluted fractions revealed that Prosβ5–RFP co-fractioned with endogenous 26S proteasomes (Fig. S1A). Additionally, we performed a co-immunoprecipitation experiment using Drosophila larvae expressing prosβ5-RFP in motor neurons. We observed that our Prosβ5–RFP (20S subunit) co-immunoprecipitated with the endogenous 19S subunit when it was pulled down with the anti-5SA antibody against the 19S Rpn10 subunit (Fig. S1B,C). To determine whether the tagged proteasomes retained catalytic activity, we performed a proteasome activity assay to investigate the chymotrypsin-like activity in our immunoprecipitated samples. In this activity assay, we observed an increase in proteasome activity in samples from flies expressing prosβ5-RFP immunoprecipitated with anti-RFP antibodies compared to the IgG control samples (Fig. S1D, white squares and circles). Additionally, the chymotrypsin-like activity was completely abolished in our Prosβ5–RFP samples when a potent proteasome inhibitor MG-132 was added (Fig. S1D, black squares). These data support the idea that our tagged Prosβ5–RFP subunit assembled into a functional 26S proteasome complex in neurons.

Because we observed the transport of proteasomes in the axon, we predicted that we would also observe proteasomes trafficking within the neuromuscular junction (NMJ). Previous studies have demonstrated a critical role of proteasomes in the presynaptic nerve terminal (Aravamudan and Broadie, 2003; Broadie et al., 1999; Collins et al., 2006; DiAntonio et al., 2001; Speese et al., 2003; Tai and Schuman, 2008; Wan et al., 2000; Willeumier et al., 2006); however, very little is known about the trafficking and positioning of proteasomes within the presynaptic nerve terminal. To investigate proteasome trafficking and positioning at the Drosophila NMJ, we imaged NMJs in animals expressing prosβ5-RFP in motor neurons. Immunostaining of the NMJ from animals expressing fluorescent proteasomes in neurons revealed that Prosβ5–RFP localized to the boutons, suggesting, that proteasomes are trafficked to the NMJ (Fig. 2A). In addition, these proteasomes colocalized with the MT-associated protein Futsch supporting the idea that proteasomes within the NMJ are associated with MTs. It is interesting to note that we do not see proteasomes equally distributed across the NMJ. To determine the trafficking behavior of proteasomes within the nerve terminal, we performed live imaging of the NMJ in motor neurons expressing prosβ5-RFP and observed a population of proteasomes moving within the Drosophila NMJ (Fig. 2B; Movie 1). These analyses represent the first live imaging of proteasomes in the presynaptic nerve terminal in any system. The analysis of proteasome trafficking in the NMJ revealed some interesting differences between the axons and the synapse. First of all, we found a much greater population of stationary and reversing proteasomes in synapses (gray bars) compared to what was seen in the axons (black bars) (Fig. 2D). Additionally, tracking of individual Prosβ5–RFP particles revealed that segmental velocity is slower in the NMJ (gray bars) than in the axons (black bars) (Fig. 2C). Because of the potential for a more heterogeneous polarity of MTs in the NMJ compared to in axons (Pawson et al., 2008), we have pooled all velocities together for these analyses. These data provide the first quantification of proteasome transport in the presynaptic nerve terminal and demonstrate that compartment-specific differences in proteasome trafficking exist that favor the residence of proteasomes within the NMJ.

We wanted to directly compare the axonal transport of proteasomes to the synaptic vesicle protein synaptobrevin (Syb) to determine whether these cargoes share common transport mechanisms. Axonal tracts from third-instar larvae expressing UAS-based transgenes encoding either prosβ5-RFP (Fig. 3A, right diagram) or syb-GFP (Fig. 3A, left diagram) by means of a combination of motor neuron-specific Gal4 drivers were imaged live using video microscopy to generate videos of axonal transport (Fig. 3A, kymographs). Under these imaging conditions, the directionality of transport was able to be determined. We compared the trafficking of Prosβ5–RFP (dark gray) versus Syb–GFP (light gray) and observed no difference in the percentage of particles moving in the anterograde or retrograde direction, the percentage of particles that were stationary, or the percentage of particles that exhibit a reversing behavior (Fig. 3B). By using particle-tracking software, we then measured the run length and segmental velocity in both the anterograde and retrograde direction for both particles. We found no difference in either mean run length or segmental velocity between the Prosβ5–RFP (dark gray bars) and Syb–GFP (light gray bars) particles in the anterograde direction (Fig. 3C–E). We did, however, observe a significant difference in the mean run length and the segmental velocity in the retrograde direction between the two populations, with Prosβ5–RFP exhibiting ∼32% shorter run lengths and ∼36% slower velocities than Syb–GFP particles (Fig. 3C,D,F). Detailed analysis of the velocity distribution of individual particles revealed that the difference in retrograde velocity is due to a uniform average of slower moving Prosβ5–RFP particles rather than a bimodal distribution of fast and very slow moving particles (Fig. 3F). These differences between the anterograde and retrograde transport rates of proteasomes are similar to what was recently observed in developing mammalian neurons in culture (Hsu et al., 2015).

To further investigate this difference in retrograde transport between Prosβ5–RFP and Syb–GFP particles, we challenged the system by introducing a mutation in one of the components of the transport machinery. Since many of the key motor protein genes are required for viability in Drosophila larvae, we took an advantage of a hypomorphic mutation in the kinesin light chain gene (Klc^c02312) that is homozygous viable. Importantly, mutations in the mammalian KLC1 gene have previously been shown to affect axonal transport in both anterograde and retrograde direction, supporting the model that interactions between the dynein and kinesin motor systems are required for normal cargo transport (Encalada et al., 2011; Gross, 2004; Hancock, 2014; Reis et al., 2012). Klc mutant larvae expressing the prosβ5-RFP or syb-GFP transgenes in their motor neurons were live-imaged and assayed for axonal transport. As expected, the analysis of segmental velocity in anterograde direction revealed a significant decrease in Prosβ5–RFP and Syb–GFP velocities consistent with both of these particles requiring Klc for anterograde transport (Fig. 4A). In contrast, the analysis of retrograde velocity in the Klc mutant background revealed a significance reduction in the retrograde velocity of Syb–GFP particles but no reduction in the retrograde velocity of Prosβ5–RFP particles (Fig. 4C). These results support the model that proteasomes and synaptic vesicle proteins share anterograde transport mechanisms, but utilize distinct retrograde motor proteins during axonal transport. A prediction arising from this model is that Syb–GFP and Prosβ5–RFP particles are transported independently within the axon. To investigate this prediction, we drove the expression of the prosβ5-RFP and syb-GFP transgenes in same motor neurons and directly assayed whether the two populations are transported together (Fig. 4E). The analysis of the co-transport revealed that only 8% of all Syb–GFP and 2.6% of all Prosβ5–RFP particles are co-transported, with the large majority of these particles being transported independently (Fig. 4F). Taken together, these data support the novel model that proteasomes utilize a distinct retrograde transport machinery compared to synaptic vesicle proteins.

Our data revealed a difference in retrograde but not anterograde transport between the Prosβ5–RFP and Syb–GFP particles suggesting the existence of distinct retrograde transport molecular machinery for proteasomes. We hypothesized that this difference in retrograde transport could be due to the existence of a unique dynein motor protein for proteasome trafficking. To investigate this possibility, we performed a reverse genetic RNAi screen against all putative dynein light chains (DLCs) in Drosophila to determine whether DLCs specify proteasome transport in neurons (Fig. 5A). Our strategy was to screen larvae expressing transgenic RNAi encoding putative DLC genes to identify individual DLC genes that specifically altered Prosβ5–RFP trafficking when knocked down in motor neurons by using live imaging of third-instar larvae axonal tracts. For each RNAi condition, three larvae were dissected and imaged. From this live imaging, we determined transport velocity and looked for the presence of large accumulation of proteasomes (proteasome blockage) within the axonal tracts for each RNAi target. Post-imaging larval preparations were fixed and subjected to immunofluorescence microscopy to investigate the presence of axonal blockages of synaptic vesicle proteins using an antibody against the vesicular glutamate transporter (VGluT). The presence of axonal blockages of synaptic proteins has been extensively used in Drosophila to investigate dynein motor function during axonal transport (Chang et al., 2013; Martin et al., 1999). We predicted that DLCs that specifically regulate proteasome transport (but not synaptic vesicle protein transport) would have reduced proteasome transport but would lack VGluT-positive axonal blockages. By using this strategy, we found that RNAi knockdown of gene CG6998 resulted in the largest reduction in proteasome transport without any axonal blockages of synaptic proteins (Fig. 5C,D; blockage data for screen not shown). CG6998, also known as cut up (ctp), encodes the Drosophila ortholog of the mammalian LC8-type DLC proteins DYNLL1 and DYNLL2 (Fig. 5B). In support, mammalian LC8 has previously been implicated in the axonal transport of the proteins destined for the synapse (Fejtova et al., 2009). We next extensively characterized axonal transport of proteasomes and synaptic proteins in ctp mutants.

To confirm the results of our screen that identified ctp as a DLC specific for axonal transport of proteasomes, we utilized a fly line carrying a well-characterized mutation (P{lacW}ctp^G0371) in the ctp gene (Fig. 6A) (Batlevi et al., 2010). Although this mutation is a recessive lethal, this stock will produce hemizygous ctp/y males at a very low frequency allowing us to investigate the effects of the loss of this protein on proteasome transport in male larvae. To confirm the role of Ctp in the axonal transport of proteasomes, we live imaged the axonal tracts from control, ctp/+ and ctp/y mutant larvae expressing the prosβ5-RFP transgene in motor neurons (Fig. 6B; Movies 2 and 3). Consistent with our screen results, we observed that the axonal transport of proteasomes in ctp/y mutants was significantly impaired (Fig. 6B–I). Specifically, we found fewer Prosβ5-RFP particles moving in anterograde and retrograde direction in ctp/y (light gray bars) and ctp/+ larvae (dark gray bars) compared to in the control (ctrl, black bars) (Fig. 6C). Additionally, ctp/y mutant (light gray bars) and ctp/+ larvae (dark gray bars) showed a pronounced increase in the number of stationary and reversing particles when compared to what was observed in the wild-type control (ctrl, black bars) (Fig. 6C). Particle-tracking analysis revealed a large decrease in segmental velocity and in run lengths for Prosb5–RFP transport in the retrograde and anterograde directions in the ctp/y mutants (light gray bars) and ctp/+ larvae (light gray bars) (Fig. 6D–I). Taken together, these analyses demonstrate a requirement of ctp for an efficient axonal transport of proteasomes in Drosophila motor neurons.

Our genetic and live-imaging data suggest a physical interaction between Ctp and proteasomes. To investigate this possibility, we performed a co-immunoprecipitation experiment using Drosophila wild-type larvae. Although there are no antibodies against Ctp, the amino acid conservation between human DYNLL1 (hereafter denoted hDlc1) and Ctp suggested that the available anti-hDlc1 antibody would cross-react with Ctp. Larvae were lysed, and the Ctp was immunoprecipitated using anti-hDlc1 antibody-coated protein G beads. We found that both the 20S (left panel) and 19S (right panel) proteasome subunits were co-immunoprecipitated with the anti-hDlc1 antibody and not with IgG control antibodies (Fig. 6J, lanes 3 versus 4), supporting that idea that Ctp physically associates with the 26S proteasome. The interaction between proteasomes and Ctp is similar to previous data showing an interaction between mammalian LC8 DLCs and the presynaptic protein bassoon (Fejtova et al., 2009).

Our screen results predict that the axonal transport of synaptic proteins will be not be affected by mutations in ctp. To investigate this possibility, we first assayed the presence of synaptic blockages in ctp/y mutant axons by using immunofluorescence microscopy with antibody against VGluT. Digital images were analyzed to quantify the size of the VGluT-positive blockages and the number of blockages per unit length of axons in different genetic backgrounds. Consistent with our screen results, we found no difference in the size or the number of VGluT puncta between the control and the ctp/y mutant axons (Fig. S2A–D). We extended this analysis by investigating whether ctp genetically interacts with other motor protein genes, similar to what has been shown for other motor mutants (Martin et al., 1999). We analyzed synaptic blockages in lines harboring one copy of our ctp allele in trans to one copy of null Khc, Klc and Dhc64C alleles. Dominant trans-heterozygotic interactions between Khc, Klc and Dhc64C alleles have been previously demonstrated (Martin et al., 1999). The transheterozygote combinations of these alleles with ctp/+ did not show any synergistic effects on the size or the number of VGluT axonal blockages compared to what was seen in the appropriate controls (Fig. S2A–D). These data suggest that ctp is not required for the axonal transport of synaptic vesicle proteins.

Because blockage analysis does not directly measure axonal transport, we decided to directly analyze the axonal transport of the synaptic vesicle protein synaptotagmin 1 (Syt1) in ctp mutants by using live-imaging microscopy. Axonal tracts from control, heterozygous female (ctp/+) and hemizygous male (ctp/y) larvae expressing Syt–GFP in motor neurons were live-imaged to generate movies of axonal transport (Fig. 7A; Movies 4 and 5). By using particle-tracking, we determined the run lengths and velocities in our control, and ctp/+ and ctp/y mutants. We observe no difference in segmental velocity or in run length in the retrograde or anterograde directions of Syt–GFP transport in ctp/y (light gray bars) or ctp/+ (dark gray bars) mutant backgrounds compared to that seen in the wild-type control (ctrl, black bars) (Fig. 7B–G). These data suggest that Syt does not utilize Ctp for axonal transport. This observation also suggests that unlike proteasomes, Syt–GFP transport vesicles should lack Ctp. To investigate this possibility we performed a co-immunoprecipitation experiment using Drosophila larvae expressing Syt-GFP in motor neurons. Ctp was immunoprecipitated as previously described above. These analyses found that Syt–GFP does not co-immunoprecipiate with the endogenous Ctp (Fig. 7H, compare to Fig. 6J). Taken together, these data strongly support the model that ctp is not required for the axonal transport of synaptic vesicle proteins in Drosophila motor neurons.

Previous studies investigating the role of the UPS in synapse growth and remodeling suggest that reduced proteasome activity should result in increased presynaptic growth (Collins et al., 2006; DiAntonio et al., 2001; Ehlers, 2003; Wan et al., 2000). Because ctp mutations affect both anterograde and retrograde transport, we predicted that the mutation in ctp would result in fewer proteasomes at the synapse and consequently enhanced synapse growth. To investigate this possibility, we quantified synapse growth in wild-type and ctp/y NMJs by counting the number of 1b boutons at the muscle 4 NMJs. These analysis found that ctp/y mutants have significantly increased numbers of 1b boutons compared to that in controls (Fig. 8A–D). It is important to note that the increase in bouton number in the ctp/y mutant is mostly due to an increase in the number of supernumerary ‘satellite’ boutons that concentrate distally along the synaptic branch (Dickman et al., 2006) (Fig. 8Ai,ii,Bi,ii). This synaptic growth phenotype is also seen at NMJs in larvae expressing a dominant-negative temperature-sensitive proteasome mutation known to block proteasome function in the motor neuron (C155-Gal4; UAS-DTS7) (Fig. 8E) (Belote and Fortier, 2002; Smyth and Belote, 1999). Quantification of total bouton numbers supports that both ctp mutations and proteasome inhibition results in similar increases in synaptic growth compared to that seen in controls (Fig. 8D,E). Finally, live imaging of the NMJs of ctp mutants expressing the prosβ5-RFP transgene found only the rare stationary proteasome, consistent with ctp mutations impairing the delivery of proteasomes to the synapse (data not shown). Taken together, these data are consistent with mutations in ctp impairing transport of proteasomes to the NMJ. It is also possible that the effects of ctp mutation on synapse growth are the result of impaired trafficking of proteasomes in the nerve terminal.

In this study, we used fluorescence time-lapse imaging and single-particle tracking in Drosophila third-instar larvae to investigate trafficking of proteasomes in motor neurons. Our data demonstrate that proteasomes use fast MT-based axonal transport to traffic in Drosophila motor neurons, including within the presynaptic nerve terminal. The quantitative analyses of proteasome trafficking in axons of motor neurons revealed that the velocity of retrograde transport of proteasomes is significantly slower than the velocity of anterograde transport. This is similar to what has been observed in developing hippocampal cultures derived from mouse brains, supporting the idea of conserved transport mechanisms (Hsu et al., 2015). Furthermore, we found that the values for retrograde velocity and run length of proteasomes are significantly less than that of synaptic vesicle proteins, and that mutations in the Klc gene had a much stronger effect on the retrograde transport velocities of synaptobrevin than they did on proteasomes. Finally, our genetic analysis of proteasome transport revealed that the Drosophila homolog of the mammalian LC8 dynein light chain, cut up (ctp), is required for retrograde axonal transport of proteasomes but is dispensable for the retrograde axonal transport of synaptic vesicle proteins. These analyses strongly support the model that proteasomes utilize a different dynein motor complex for transport to that used by other synaptic cargo. It is interesting to note that synaptotagmin has been identified in both proteome studies of proteasome-dependent protein degradation and observed in polyQ-induced protein aggregates (Lee et al., 2004; Khan et al., 2006; Willeumier et al., 2006). These data support the idea that synaptotagmin is likely degraded by the proteasome, perhaps in the synapse. Given the physiological significance of changes in the abundance of key synaptic molecules such as synaptotagmin, perhaps it is not surprising that the proteasome and its synaptic targets utilize distinct transport mechanisms. We would predict that this arrangement would protect against inadvertent interactions between key substrates and the proteasomes, preserving the physiological efficacy of regulated changes in protein abundance.

Our model that proteasomes and synaptic proteins are trafficked independently is seemingly in conflict with data from a recent study of axonal transport of proteasomes from cultured hippocampal neurons that suggested that proteasomes are co-transported with various membrane-associated cargos, including the synaptic vesicle protein synaptophysin (Otero et al., 2014). We specifically addressed this possibility by simultaneously co-expressing proteasomes and synaptobrevin in the same motor neuron and analyzing transport. These analyses of co-transport revealed that only 8% of the synaptobrevin transport vesicles co-transported with proteasomes, despite the comparatively large number of proteasome particles. We also find that in our study most of the proteasomes (∼65%) are moving, whereas Otero et al. (2014) report a relatively small population of mobile proteasomes (∼20%) with majority of particles exhibiting a random-like, reversing motion (∼80%). Furthermore, the velocities of both retrograde and anterograde proteasome transport reported in Otero et al. (2014) are much slower than what we have observed in our study. It should be noted that the velocities that we report for both anterograde and retrograde transport of proteasomes are consistent with velocities of proteasomes observed in cultured hippocampal neurons during axonal differentiation (Hsu et al., 2015). These inconsistencies between studies may reflect differences due to different neuronal cell types or reporter expression, or differences between our in situ model and cultured primary neurons.

Cytoplasmic dynein is a multi-subunit motor protein responsible for the MT-based transport of a wide range of cargos. Our data suggests that the Drosophila DLC Ctp can specify the axonal transport of distinct cargoes. Consistent with this role, the mammalian LC8 (DYNLL1 and DYNLL2) family of DLCs have been shown to simultaneously associate with the dynein motor and a range of cargo proteins including active zone components and proteins involved in mRNA localization during embryogenesis (Fejtova et al., 2009; Lee et al., 2006; Navarro et al., 2004; Schnorrer et al., 2000). Importantly, mutations that disrupt the interaction of these proteins with LC8 have been shown to disrupt their dynein-mediated MT transport, providing a link between LC8-binding partners and MT-dependent trafficking. In addition to demonstrating that ctp was necessary for the normal axonal transport of proteasomes, we provide biochemical evidence that Ctp physically interacts with the 20S and 19S proteasome subunits. It is not clear from these co-immunoprecipitation studies if Ctp directly binds to these specific subunits or perhaps some other proteasome subunit. Structural studies have indicated that a large number of diverse cargoes bind to the same groove in the LC8/Ctp dimer and in certain cases can either compete or facilitate binding with other cargoes including the intermediate chain of dynein (Benison et al., 2007; Nyarko et al., 2011). Based on our data, we would predict that the binding of the 26S proteasome to Ctp would favor the association of Ctp with the intermediate chain of dynein and facilitate dynein-dependent transport. Neither synaptobrevin nor synaptotagmin would be predicted to have this activity.

In addition to providing cargo specificity, our results also suggest that Ctp can affect the processivity of the dynein motor. First, we observe that proteasomes have a slower retrograde velocity of transport than does synaptobrevin. In addition, the retrograde velocities and run lengths of proteasome transport are reduced, but not absent, in ctp mutants. These results suggest that Ctp association with the dynein motor alters the biochemical function of the resulting motor complex. Currently, little is known about how DLCs participate in motor processivity in any system. Previous studies have demonstrated that the processivity and activity of the dynein motor can be altered by interactions with various regulatory proteins. For example, the dynactin complex has been shown to significantly increase dynein processivity similar to what our data show for ctp (Kardon et al., 2009; Karki and Holzbaur, 1995; King and Schroer, 2000; Schroer, 2004; Vallee et al., 2012). Interestingly, several recent studies suggest that the normal interaction between dynein and the dynactin complex requires an LC8 dynein light chain (Jie et al., 2015; Stuchell-Brereton et al., 2011). In addition, mutations that disrupt Dyn2 function (a yeast homolog of LC8) also impaired the recruitment of the dynactin complex to the dynein motor complex (Stuchell-Brereton et al., 2011). Further studies will be required to determine whether Ctp alters the processivity of proteasomes due to the recruitment of the dynactin complex.

Studies utilizing pharmacological inhibition of proteasome activity have shown that inhibition of proteasomes results in a rapid increase in synapse function, including neurotransmitter release, suggesting that proteasomes are localized near the neurotransmitter release site (Speese et al., 2003; Willeumier et al., 2006; Yi and Ehlers, 2005). Despite the evidence supporting the presence of proteasomes within the presynaptic nerve terminal, direct evidence for this is absent. Our study is the first to visualize and study proteasome trafficking in the presynaptic nerve terminal. Furthermore, the consistency in transport dynamics between axon and synapse, the colocalization with Futsch and the lack of movement in ctp mutants support the idea that synaptic transport of proteasome is MT based. Furthermore, analysis of synapse morphology in ctp mutants demonstrates that proteasomes are required within the NMJ for normal synapse growth during larval development. Recent studies aimed at studying post-synaptic proteasomes have suggested that proteasomes undergo dynamic changes in their subcellular localization in response to depolarization suggesting a role for proteasome transport during synaptic plasticity (Bingol and Schuman, 2006; Glickman and Raveh, 2005; Shen et al., 2007). It will be important to investigate whether ctp mutations have effects on synaptic plasticity at the Drosophila NMJ.

Interestingly, we also found that the trafficking behavior of proteasomes in the synapse has some important differences from trafficking in the axons. We observe a substantial difference between these two neuronal compartments in terms of the number of stationary particles, with the number of stationary proteasomes in the synapse increasing by ∼400% (Fig. 2D). Additionally, we found that synaptic proteasomes move, on average, more slowly than proteasomes in the axons (Fig. 2C). A similar change in the trafficking between the axon and nerve terminal has also been observed for neurexin (Nrxn) proteins, which are also transported more slowly in the synapse versus axon (Neupert et al., 2015). The combination of these changes in trafficking behavior favor a longer residence of an individual proteasome in the synapse compared to in the axon. The distribution of stationary proteasomes throughout the NMJ could facilitate the local degradation of synaptic proteins within the bouton. This behavior is in contrast to studies of the trafficking of neuropeptide-containing dense core vesicles (DCVs) within the NMJ, which have few stationary vesicles and utilize a continuous ‘conveyor belt’ model transport at the synapses ensuring continuous source of DCVs (Wong et al., 2012). These differences suggest MT-based transport within the synapse is tailored to the specific cargo and their respective functions.

All fly stocks were maintained at 25°C on normal laboratory food. The following genotypes were used: UAS-DNGlued (Chang et al., 2013), C155-Gal4 [Elav; (Robinow and White, 1988)], OK6-Gal4 (Bloomington Drosophila Stock Center, Bloomington, IN), C380-Gal4 (a gift from the Bhat laboratory, The University of Texas Health Science Center at San Antonio, TX), UAS-syb-GFP, UAS-syt-GFP, PBac{Klc^c02312/TM6B,Tb¹} (Bloomington Drosophila Stock Center, Bloomington, IN), P{lacW}ctp^G0371 (a gift from the Baehrecke laboratory, University of Massachusetts Amherst, MA), Dhc64C 4-19 (Bloomington Drosophila Stock Center, Bloomington, IN), and UAS-DTS7 (Belote and Fortier, 2002; Bloomington Drosophila Stock Center, Bloomington, IN). All CG RNAi lines were ordered from the Vienna Drosophila RNAi Center (Vienna, Austria). C155-Gal4, OK6-Gal4 and C380-Gal4 lines were used for targeted expression of transgenes in the nervous system. Transgenic UAS-prosβ5-RFP lines were generated by microinjection into w¹¹¹⁸ (Rainbow Transgenics Flies, Camarillo, CA).

A full-length cDNA for prosβ5-RFP (LD0717) was used as a PCR template for directional TOPO cloning into pENTR (Invitrogen, Carlsbad, CA). All subsequent tagging of Prosβ5–RFP was to the C-terminus and achieved by using the Gateway cloning system (Invitrogen) and appropriate pUAS-based destination vectors for microinjection (Drosophila Genomics Resource Center, Indiana University, Bloomington, IN).

For analysis of axonal transport, larvae of indicated genotypes were dissected and imaged live in HL3 solution (Mahoney et al., 2014) supplemented with 7 mM L-glutamic acid. Live imaging was performed with the 100× oil Plan-Chromomat (NA 1.4) objective on an Axiovert 200 M microscope equipped with the Orca2 camera set to 2×2 binning. Images were captured using a 200-ms exposure at 1.8 Hz. Each recording consisted of ∼30 s of uninterrupted imaging. All imaging and post-image analysis was performed by using Slidebook software (3I, Denver, CO). For each genotype, individual particles were classified as stationary, anterograde, retrograde or repeatedly reversing. Single-particles were manually tracked and their trajectories were analyzed for particle velocity and run length. Segmental velocity is defined as mean velocity of a trajectory segment uninterrupted by a pause or particle reversal. A minimum of ten frames was used for the analyses of segmental velocity and run length; 10–20 animals per genotype were analyzed.

Third-instar larvae of indicated genotypes were dissected and fixed as previously described (Eaton et al., 2002). The larvae were stained with rabbit anti-VGluT (1:5000; Chang et al., 2013). Goat anti-rabbit-IgG conjugated to Alexa Fluor 555 (1:400; cat. no. A-21428; Invitrogen, Carlsbad, CA) was used to visualize VGluT staining. All fixed images were captured with a 63× oil Plan-Neofluar (NA 1.25) objective using an Orca-2 backcooled charge-coupled device camera (Hamamatsu, Hamamatsu, Japan) attached to a Zeiss Axiovert 200 M (Carl Zeiss, Jena, Germany). The Z-stack images were collected using the 63× objective and deconvolved using a constrained iterative algorithm (Slidebook software). Individual V-GluT-positive puncta were masked and subjected to quantitative analysis of the voxel size; 12–28 larvae were dissected and analyzed for each genotype.

A total of 300 larvae expressing prosβ5-RFP in all neurons (C155-Gal4;UAS-prosβ5-RFP; UAS-prosβ5-RFP) were homogenized in an extraction buffer [40 mM Tris-HCl, 50 mM NaCl, 2 mM β-mercaptoethanol, 5 mM MgCl₂, 2 mM ATP, 10% glycerol and protease inhibitor (Roche, Indianapolis, IN)]. The lysates where centrifuged for 5 min at 1000 g at 4°C. The supernatant (S1) was centrifuged again for 10 min at 6000 g at 4°C. The resulting supernatant (S2) was incubated at 4°C with protein G Dynabeads previously incubated with anti-RFP antibody (10 μg/ml; Abcam, Cambridge, MA) or rabbit serum IgG as described in the protocol (Life Technologies, Carlsbad, CA). The Proteasome Activity Fluorometric Assay Kit II was used to assay proteasome chymotrypsin-like activity of purified neuronal Prosβ5-RFP as described in the manufacturer's protocol (UBPBio, Aurora CO). A final concentration of 50 μm of MG-132 (UBPBio, Aurora, CO) was used to inhibit the activity of Prosβ5–RFP in vitro.

A total of 300 larvae of indicated genotypes were homogenized in 1.5 ml of lysing buffer [50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM Na₃VO₄, 20 mM NaF and protease inhibitors (Roche, Indianapolis, IN)]. The lysates were centrifuged for 5 min at 1000 g at 4°C. The resulting supernatant was assayed for protein concentration by using the DC Protein Assay (BioRad, Hercules, CA) and diluted (1:1) in a modified RIPA buffer [50 mM Tris-HCl pH 7.4, 5.0 mM NaF, 1.0 mM Na₃VO₄, 1% Triton X-100 and protease inhibitors (Roche, Indianapolis, IN)]. The diluted lysates were incubated at 4°C overnight with protein G Dynabeads previously incubated with anti-DYNLL1 antibody (1:100; cat. no. ab51603; Abcam, Cambridge, MA), anti-19S 5SA (1:100; cat. no. ab20239; Abcam, Cambridge, MA) or rabbit serum (1:100; GenScript, Piscataway, NJ) as described in a protocol (Life Technologies, Carlsbad, CA). The immune complexes were eluted from the beads by adding 2× sample buffer containing 5% β-mercaptanol followed by boiling for 5 min. Proteins were separated on SDS-PAGE gels (10% Tris-HCl gels) and transferred onto nitrocellulose membranes for 1.5 h at 350 mA. Membranes were blocked with 5% milk in 1% Tween 20 in Tris-buffered saline (TBS) pH 7.4, at room temperature for 30 min. Membranes were probed with the rabbit anti-19S 5SA (1:1000; Abcam, Cambridge, MA) and rabbit anti-Drosophila Prosβ5 (dβ5) (1:1000; a gift from the Figueiredo-Pereira laboratory; Vernace et al., 2007), mouse anti-RFP (1:1000; cat. no. ab125244; Abcam, Cambridge, MA) or mouse anti-GFP (1:1000; cat. no. ab1218; Abcam, Cambridge, MA) antibodies followed by secondary horseradish peroxidase (HRP)-labeled goat anti-rabbit-IgG antibody (1:2500; cat. no. sc-2004; Santa Cruz Biotechnology, Dallas, TX) or goat anti-mouse IgG antibody (1:2500; cat. no.:sc-2005; Santa Cruz Biotechnology). Immunostained proteins were visualized by using the ECL detection method (Sigma-Aldrich, St. Louis, MO).

Fractionation of proteasomes from adult Drosophila was performed exactly as previously described (Vernace et al., 2007). Briefly, mixed populations of flies were chilled and then ground using a dounce homogenizer in cold lysis buffer (25 mM Tris-HCl pH 7.5, 2 mM ATP and 1 mM DTT). Homogenates were cleared by centrifuging for 10 min at 19,000 g and the resulting supernatants subjected to a 10–40% glycerol gradient. Fractions were acetone precipitated and subjected to standard SDS-PAGE and processed for immunoblotting as described above.

All multiple comparisons were performed using one-way ANOVA with a Tukey correction for multiple comparisons. Student's t-test was used for all two-way comparisons. The nonparametric Kolmogorov–Smirnov test was used for the analysis of distribution. *P<0.05, **P<0.01.

The authors would like to thank Leo Chang for technical assistance. Stocks obtained from the Bloomington Drosophila Stock Center (NIH grant P40OD018537) were used in this study.