|Home | About | Journals | Submit | Contact Us | Français|
The location, size and number of synapses critically influence the specificity and strength of neural connections. In axons, synaptic vesicle (SV) and active zone (AZ) proteins are transported by molecular motors and accumulate at discrete presynaptic loci. Little is known about the mechanisms coordinating presynaptic protein transport and deposition to achieve proper distribution of synaptic material. Here we show that SV and AZ proteins exhibit extensive co-transport and undergo frequent pauses. At the axonal and synaptic pause sites, the balance between the capture and dissociation of mobile transport packets determines the extent of presynaptic assembly. The small G protein ARL-8 inhibits assembly by promoting dissociation, while a JNK kinase pathway and AZ assembly proteins inhibit dissociation. Furthermore, ARL-8 directly binds to the UNC-104/KIF1A motor to limit the capture efficiency. Together, molecular regulation of the dichotomy between axonal trafficking and local assembly controls vital aspects of synapse formation and maintenance.
Neurotransmission relies on the fusion of synaptic vesicles (SVs) with the plasma membrane at the presynaptic terminals, where SVs are clustered near the active zones (AZs). AZs are specialized regions of the plasma membrane defined by a protein meshwork which contain the molecular machinery necessary for SV recruitment and recycling (Jin and Garner, 2008; Owald and Sigrist, 2009; Südhof, 2012). The number, size and location of synapses vary among different types of neurons and critically impact the efficacy of neurotransmission (Atwood and Karunanithi, 2002; Holderith et al., 2012). For example, in both vertebrate and invertebrate nervous systems, some neurons specify a single synaptic connection at the axon terminal, while others elaborate sequential release sites called en passant synapses. Although many extrinsic cues and cell surface molecules have been shown to shape synaptic connectivity (Shen and Scheiffele, 2010), our understanding of the intracellular mechanisms involved in synaptic patterning remains incomplete.
The targeting of SVs and AZ proteins to specific sites depends on their directed axonal delivery by molecular motors (Goldstein et al., 2008; Hirokawa et al., 2010). Electron and light micrographic studies have demonstrated that many SV components are trafficked in SV protein transport vesicles (STVs) (Matteoli et al., 1992; Ahmari et al., 2000; Tao-Cheng, 2007). Live imaging has revealed that STV packets travel along axons bi-directionally and intermittently, occasionally splitting into smaller packets or merging into larger clusters (Kraszewski et al., 1995; Dai and Peng, 1996; Ahmari et al., 2000; Sabo et al., 2006). In addition, they can rapidly accumulate at new axodendritic contact sites and become capable of stimulation-evoked SV recycling (Ahmari et al., 2000; Washbourne et al., 2002; Sabo et al., 2006). On the other hand, the 80-nanometer dense core Piccolo-Bassoon transport vesicles (PTVs) are proposed to represent modular packets that assemble the AZ cytomatrix in vertebrate neurons (Zhai et al., 2001; Shapira et al., 2003; Maas et al., 2012). Interestingly, recent electron micrographic (EM) and live imaging studies reported that AZ and SV proteins may be pre-assembled into multi-vesicle transport complexes and co-trafficked in cultured neurons (Tao-Cheng, 2007; Bury and Sabo, 2011).
We previously identified a conserved Arf-like small G protein, ARL-8, as a critical regulator of presynaptic patterning and axonal transport in C. elegans (Klassen et al., 2010). Loss-of-function mutations in arl-8 caused ectopic accumulation of presynaptic specializations in the proximal axon and a loss of presynapses in distal segments, leading to deficits in neurotransmission. Time-lapse imaging revealed that arl-8 mutant STVs prematurely associate into immotile clusters en route, suggesting that ARL-8 facilitates the trafficking of presynaptic cargo complexes by repressing excessive self-assembly during axonal transport.
To further understand the molecular mechanisms coordinating presynaptic protein transport with assembly, we performed forward genetic screens to identify molecules that functionally interact with arl-8. Here we report that loss-of-function mutations in a JNK MAP kinase pathway partially and strongly suppress the abnormal distribution of presynaptic proteins in arl-8 mutants. We show that the JNK pathway is required for excessive STV aggregation during transport in arl-8 mutants and promotes the clustering of SVs and AZ proteins at the presynaptic terminals. Time-lapse imaging further reveals that transiting AZ proteins are in extensive association with STVs and promote STV aggregation during transport, with ARL-8 and the JNK pathway antagonistically controlling STV/AZ association en route. In addition, the anterograde motor UNC-104/KIF1A functions as an effector of ARL-8 and acts in parallel to the JNK pathway to control STV capture at the presynaptic terminals and during transport. Collectively, these findings uncover novel mechanisms that modulate the balance between presynaptic protein transport and self-assembly and highlight the close link between transport regulation and the spatial patterning of synapses.
The C. elegans cholinergic motoneuron DA9 provides an in vivo model to investigate the molecular mechanisms regulating presynaptic patterning. DA9 is born embryonically. During development, its axon elaborates a series of en passant synapses with the dorsal body wall muscles within a discrete and stereotyped domain, as visualized with YFP-tagged SV protein synaptobrevin (SNB-1::YFP) (Figures 1A–1B; White et al., 1976; Klassen and Shen, 2007). This synaptic pattern is already present at hatching, but the synapses continue to grow in size and number during post-embryonic development. Loss of function in arl-8 results in ectopic accumulation of SNB-1::YFP in the proximal axon and the appearance of abnormally large clusters in this region, accompanied by a loss of distal puncta (Figure 1C; Klassen et al., 2010). To identify additional molecules regulating presynaptic patterning, we performed forward genetic screens for suppressors of the arl-8 phenotype and isolated two recessive mutations, wy733 and wy735, which strongly and partially suppressed the abnormal distribution of SV proteins in arl-8(wy271) loss-of-function mutants (Figure S1A–S1D).
Using SNP-SNIP mapping, whole-genome sequencing and transgenic rescue, we identified these two suppressors as missense mutations at conserved amino acid residues in JKK-1, a homolog of the mammalian MAP kinase kinase MKK7 (Figure S1E). A null allele jkk-1(km2) (Kawasaki et al., 1999) suppressed the arl-8 phenotype to the same degree as wy733 and wy735 (Figure 1D), indicating that loss of JKK-1 activity suppresses the arl-8 phenotype. We used the jkk-1(km2) allele for our subsequent analyses.
We quantified the suppression of arl-8 by jkk-1. Compared to the arl-8 single mutants, SNB-1::YFP is redistributed into more distal axonal regions in arl-8; jkk-1 double mutants (Figures 1I and 1J). The size of the proximal SNB-1::YFP puncta (0–25 μm from commissure) is also significantly reduced in the double mutants (Figure 1J). Furthermore, while the arl-8 mutants exhibited reduced number of presynaptic SNB-1::YFP puncta, this defect is abolished in the arl-8; jkk-1 double mutants (Figure 1K). Similar results were obtained with two additional SV proteins, RAB-3 (Figures S2A–S2D) and SNG-1/synaptogyrin (data not shown). Therefore, jkk-1 mutations partially and strongly suppressed multiple aspects of the arl-8 mutant phenotype in DA9.
MAP kinases (MAPKs) act in cascades in which each MAPK is activated via phosphorylation by a MAPK kinase (MAPKK), which is in turn activated by a MAPKK kinase (MAPKKK) (Davis, 2000). JKK-1 was shown to be a specific upstream activator of JNK (c-Jun N-terminal kinase) −1, a homolog of mammalian JNK3 (Kawasaki et al., 1999). A null mutation in jnk-1, gk7, caused the same degree of suppression of the arl-8 phenotype as jkk-1(km2) (Figures 1E and 1I–1K). Moreover, the degree of suppression in arl-8, jnk-1; jkk-1 triple mutants is indistinguishable from that in either double mutants (Figures 1F and 1I–1K), indicating that jkk-1 and jnk-1 function in the same pathway. jkk-1 and jnk-1 mutants were previously shown to partially mislocalize SNB-1::GFP to the dendrite in the DD motoneurons (Byrd et al., 2001). We found that the jkk-1 and jnk-1 single mutants appeared grossly normal in SV protein localization in DA9 (Figures 1G and 1H) and did not show mislocalization of SV proteins to the DA9 dendrite (data not shown). However, these mutants did exhibit subtle but significant decreases in SNB-1::YFP puncta size (Figure 1J) and increases in puncta number (Figure 1K), suggesting that JNK also promotes SV clustering in wild-type animals.
To determine whether JNK functionally interacts with arl-8 broadly in the C. elegans nervous system, we examined several other neuron types, including the cholinergic motoneuron DB7, the GABAergic DD motoneurons and the thermosensory neuron AFD, all of which have a proximal axonal region devoid of presynapses and form en passant presynapses in the distal axon (Hallam and Jin, 1998; Klassen and Shen, 2007; Hellman and Shen, 2011). As in DA9, these neurons accumulate abnormally large presynaptic puncta within their proximal axon at the expense of distal puncta in arl-8 mutants (Klassen et al., 2010; data not shown). jkk-1(km2) strongly suppressed these phenotypes in all three types of neurons (Figures S2E–2H and data not shown), indicating that arl-8 genetically interacts with the JNK pathway to regulate presynaptic protein distribution in many neuron types.
We previously showed that the proximally mislocalized STV accumulations in arl-8 mutants also contain AZ proteins and ultrastructurally resemble bona fide presynapses (Klassen et al., 2010). To address whether JNK also regulates AZ localization, we used an integrated UNC-10::GFP transgene to characterize changes in AZ protein distribution. UNC-10 encodes the C. elegans homolog of mammalian RIM1, an integral AZ protein which interacts with several presynaptic proteins to regulate SV docking, priming and synaptic plasticity (Südhof, 2012). In wild-type DA9, UNC-10::GFP forms discrete puncta juxtaposed with mCherry::RAB-3 puncta at presynaptic terminals (Figures 2A and S3A). In arl-8 mutants, similar to SV proteins, UNC-10::GFP forms large puncta ectopically in the proximal axon and is reduced in the distal axon (Figures 2B and S3B). jkk-1(km2) partially and robustly suppressed this phenotype, reducing proximal puncta size and increasing distal puncta (Figures 2C, 2F, 2G and S3C). Other AZ proteins, including the scaffold protein SYD-2/Liprin-α and the calcium channel β-subunit CCB-1, showed similar mislocalization in arl-8 mutants and these phenotypes were also strongly and partially suppressed by jkk-1(km2) (Figures S3D–S3I), suggesting that ARL-8 and JKK-1 systematically regulate presynaptic differentiation rather than select markers. arl-8 mutants also displayed proximal shift and distal loss of AZ proteins in other types of neurons, including UNC-10::tdTomato in DDs (Figures 2I and 2J) and SAD-1::YFP in AFD (data not shown). Again, these defects were largely suppressed in arl-8; jkk-1 double mutants (Figure 2K and data not shown). jkk-1(km2) single mutants displayed a reduction in presynaptic UNC-10::GFP puncta size in DA9 (Figures 2D, 2F and 2G), indicating that jkk-1 also supports AZ assembly in wild-type animals. These results suggest an antagonistic relationship between ARL-8 and the JNK pathway in regulating the clustering of presynaptic components. While the JNK pathway promotes presynaptic assembly, ARL-8 limits the extent of clustering. Consistent with this model, we found that overexpression of arl-8 in DA9 led to decreases in presynaptic RAB-3 and UNC-10 puncta size (Klassen et al., 2010; Figures 2E and 2H). The observation that JNK pathway inactivation did not completely suppress the arl-8 mutant phenotype suggests that other pathways function in parallel to JNK to antagonize arl-8 in regulating presynaptic protein clustering.
To address whether the suppression of the presynaptic morphological defects in arl-8 by jkk-1 also ameliorates synaptic function, we first visualized the distribution of the postsynaptic GABA receptor UNC-49 in the dorsal body wall muscles, which receive presynaptic inputs from the DD neurons. In wild-type animals, UNC-49::YFP forms evenly distributed clusters apposed to presynapses in DDs (Gally and Bessereau, 2003; Figure 2L). In arl-8 mutants, DDs accumulate large UNC-10::tdTomato puncta in the proximal axon with a loss of distal puncta (Figures 2I and 2J). Interestingly, the distribution of UNC-49::YFP is similarly shifted (Figures 2L and 2M) and this phenotype can be suppressed by expressing arl-8 solely in the presynaptic neurons (wyEx3666; strong rescue in 47/50 animals), suggesting that trans-synaptic communication is preserved in the arl-8 mutants. In arl-8; jkk-1 double mutants, the uniform distribution patterns of both UNC-10 and UNC-49 were largely restored (Figures 2K and 2N), indicating that the jkk-1 mutation suppressed both the pre- and postsynaptic defects of the arl-8 mutants. Secondly, we assessed the efficacy of cholinergic neurotransmission using the aldicarb sensitivity assay (Mahoney et al., 2006). The arl-8 mutants exhibited resistance to the acetylcholinesterase inhibitor, aldicarb, indicating impaired cholinergic transmission (Klassen et al., 2010; Figure 2O). This phenotype was robustly suppressed by jkk-1(km2) (Figure 2O), reflecting improvements in cholinergic synaptic function in the arl-8; jkk-1 double mutants. The jkk-1 single mutants also displayed some degree of aldicarb resistance (Figure 2O), consistent with reduced AZ and SV assembly in these mutants. We conclude that loss of the JNK pathway affects not only synapse morphology but also synapse function.
Both JKK-1 and JNK-1 are expressed in the C. elegans nervous system throughout development (Kawasaki et al., 1999). To determine whether they function cell-autonomously in neurons to suppress the arl-8 phenotype, we expressed jkk-1 or jnk-1 cDNA under the Pitr-1 pB or Pmig-13 promoter, which we use to label DA9, in arl-8; jkk-1 or arl-8, jnk-1 double mutants. These manipulations robustly rescued the suppression of arl-8 by the kinase mutations, whereas expression in the postsynaptic muscles or of a mutant JNK-1 lacking kinase activity (Hanks et al., 1988) failed to rescue (Figure S4A and data not shown). Together, these data suggest that jkk-1 and jnk-1 interact with arl-8 cell-autonomously in the presynaptic neuron to shape synaptic organization in a kinase-dependent manner.
The arl-8 mutant phenotypes and the jkk-1/jnk-1 suppression are already present at hatching. To test whether JNK also functions in the maintenance of synapse distribution, we induced jkk-1 expression driven by a heat-shock promoter (Stringham et al., 1992) at the L4 larval stage in the arl-8; jkk-1 double mutants and examined SV protein distribution at the young adult stage. This manipulation strongly rescued the suppression of the arl-8 phenotype by the jkk-1 mutation, whereas transgenic animals not receiving the heat-shock treatment or heat-shocked animals without the transgene showed no rescue (Figure S4A and data not shown). These results indicate that JNK continues to impact presynaptic pattern after its initial establishment.
To understand where JNK might function within the neuron, we further examined the subcellular localization of JKK-1 and JNK-1 in DA9 using functional GFP fusion proteins. Consistent with a presynaptic function, both JKK-1::GFP and JNK-1::GFP are highly enriched at the presynaptic terminals and colocalize with mCherry::RAB-3 (Figures S4B–S4J and S4K–S4S). This presynaptic enrichment requires the anterograde motor for STVs, UNC-104/KIF1A (Hall and Hedgecock, 1991); JKK-1::GFP and JNK-1::GFP are absent from the presynaptic terminals and accumulate in the cell body and dendrite in unc-104 mutants (data not shown).
Axonal transport is critical for the establishment and maintenance of presynapses (Hirokawa et al., 2010). Alterations in the cell-wide distribution of presynaptic components could originate from changes in their axonal transport dynamics. We previously showed that presynaptic protein mislocalization in arl-8 mutants likely results from premature clustering of STVs during trafficking (Klassen et al., 2010). To further investigate the role of the JNK pathway in regulating STV clustering during transport, we performed time-lapse imaging of STVs labeled with GFP::RAB-3 in vivo. Within the proximal axon of wild-type DA9, small mobile and stationary GFP::RAB-3 puncta can be visualized with a high-sensitivity CCD camera, evident as diagonal and vertical lines in the kymographs, respectively, with the mobile puncta representing trafficking STVs, which pause frequently en route and form stationary puncta. (Klassen et al., 2010; Figures 3A and 3B). arl-8 mutants did not exhibit changes in the directionality or velocity of STV movements (Klassen et al., 2010). However, we observed significant increases in the number and fluorescence intensity of stationary puncta (Figures 3C, 3F and and4G).4G). Meanwhile, there was a decrease in the number of moving events (Figure 3H). The jkk-1 mutation strongly alleviated the STV trafficking abnormalities in arl-8 mutants; the number of stationary puncta en route and their fluorescence intensity were significantly reduced in the arl-8; jkk-1 double mutants (Figures 3D, 3F and 3G). Accordingly, there was a significant increase in the number of moving events (Figure 3H). The jkk-1 single mutants showed no significant difference from wild-type animals (Figures 3E–3H).
The coexistence of stable and motile puncta is consistent with previous findings that STV undergo intermittent moving and stationary phases en route to the presynaptic terminals (Ahmari et al., 2000; Sabo et al., 2006). The transitions between these two states might serve as regulated switches to control the trafficking and aggregation of STVs. We observed two types of interactions between the motile and stable puncta: some motile puncta coalesce with existing stable puncta, whereas stable puncta can shed a fraction of their content to generate motile packets (Figures 3I and 3J). To better understand how these transitions are regulated, we quantified the frequency that moving particles dissociate from a stable punctum (dissociation rate) and the probability that mobile particles stop moving and cluster with a stable punctum (capture probability). We separated the arl-8 puncta into two groups: arl-8 puncta of fluorescence intensity within the wild-type range (wt intensity) and those with intensity values greater than the wild-type maximum (bright). The normal sized arl-8 stable puncta showed a significantly lower dissociation rate compared to wild type stable puncta, whereas the abnormally bright arl-8 stable puncta exhibited a significant increase in capture probability (Klassen et al., 2010; Figures 3I and 3J). The initial reduction in STV dissociation from small stable puncta and subsequent increase in STV capture at large stable puncta likely underlie the strong perturbation in presynaptic protein distribution in arl-8 mutants. In contrast, the stable puncta in arl-8; jkk-1 double mutants exhibited a significantly higher dissociation rate compared to arl-8 mutant puncta with similar intensity (Figure 3I). In addition, the capture probability of arl-8; jkk-1 stable puncta is comparable to that of the wild-type stable puncta and significantly lower than that of the abnormally bright arl-8 puncta (Figure 3J).
The above measurements were performed in the axon shaft where no mature synapses are found in wild-type animals. To investigate whether the same molecular regulation of trafficking and local assembly also takes place at mature synapses, we analyzed RAB-3 clusters in the dorsal presynaptic region (Figure 3K). Compared to wild-type clusters of similar intensity, the arl-8 mutant clusters exhibited a significantly lower dissociation rate (Figure 3L) and a significantly higher capture probability (Figure 3M), suggesting that arl-8 also promotes STV dissociation and inhibits STV capture at mature synapses. The jkk-1 mutation strongly suppressed the decrease in dissociation rate in arl-8 mutants (Figure 3L), without significantly affecting the capture probability (Figure 3M). Together, these findings suggest that loss of JKK-1 prevents excessive STV clustering in arl-8 mutants, likely by promoting dissociation of STVs from stationary clusters. This model whereby arl-8 and jkk-1 antagonistically regulate the dissociation of STVs is consistent with the changes in the cell-wide distribution of SV proteins in arl-8, jkk-1 and arl-8; jkk-1 mutants.
AZ proteins are thought to be transported in dense core vesicles, a vesicle population distinct from STVs (Zhai et al., 2001; Shapira et al., 2003; Maas et al., 2012), raising the question of how the transport of SV and AZ proteins can both be regulated by ARL-8 and JNK. To address this question, we performed two-color time-lapse imaging for UNC-10/RIM1::GFP and mCherry::RAB-3 in DA9. In wild-type animals, UNC-10::GFP movements are much less frequent than those of mCherry::RAB-3 (Figures 4B–4E), possibly reflecting a lower turnover rate of UNC-10 compared to SVs. However, we observed extensive association of the two markers during transport; 95.5% mobile UNC-10::GFP puncta also contained mCherry::RAB-3 and 22.9% mobile mCherry::RAB-3 puncta also contained UNC-10::GFP (Figure 4G). To determine whether this feature is unique to UNC-10::GFP, we also examined other AZ proteins, including GFP-tagged SYD-2/Liprin-α and the serine-threonine kinase SAD-1. The movements of these markers are even rarer than those of UNC-10::GFP, possibly due to their lower copy number on the trafficking packets and/or lower turnover rates. Nonetheless, when SYD-2/Liprin-α or SAD-1 movements were detected, association with mCherry::RAB-3 was also observed (Figures S5A–S5F). Furthermore, we also observed association of trafficking UNC-10::GFP with an integral SV protein, synaptogyrin (Figures S5G–S5I). These observations are consistent with previous Immuno EM and live imaging studies in cultured neurons (Tao-Cheng, 2007; Bury and Sabo, 2011).
Consistent with the model of STV/AZ co-transport, the ratio between anterograde and retrograde movements is similar for UNC-10::GFP and mCherry::RAB-3 (Figure 4F). In addition, the anterograde transport of STVs and several AZ proteins in DA9 are both dependent on UNC-104/KIF1A (Klassen et al., 2010). Together, our dynamic imaging analyses provide direct in vivo evidence that AZ proteins and STVs can be pre-assembled into transport complexes, providing a mechanism for the co-regulation of their axonal delivery. Therefore, the excessive aggregation of STVs in arl-8 mutants is likely accompanied by premature clustering of associated AZ proteins, resulting in defects in both STV and AZ protein distribution, which in turn can be simultaneously suppressed by JNK inactivation.
Several AZ molecules are critical for SV recruitment at presynaptic terminals (Jin and Garner, 2008; Owald and Sigrist, 2009). Loss-of-function mutations in the AZ molecules syd-2/liprin-α, syd-1 and sad-1 lead to a dramatic reduction in presynaptic SV cluster size and dispersal of SV clusters throughout the DA9 axon, and strongly suppress the enlarged size of SV clusters in arl-8 mutants (Klassen et al., 2010; Figures 6C–6I). Furthermore, we noticed that the stationary UNC-10::GFP and mCherry::RAB-3 puncta in the proximal axon showed almost complete co-localization, as evident by the double labeled vertical stripes in the kymographs (Figures 4B–4D and 4H), indicating that sites of AZ protein pause correspond to locations of STV aggregation during transport. These results, together with the observation that AZ proteins are co-trafficked with STVs, raise the possibility that AZ molecules may also promote STV clustering during trafficking, thus impacting STV axonal transport. To directly test this possibility, we performed time-lapse imaging of GFP::RAB-3 in syd-2 null mutants. Consistent with our hypothesis, we found the dissociation rate for stable GFP::RAB-3 puncta in the axon shaft in syd-2 mutants is significantly higher compared to wild-type stable puncta with similar intensity (Figure 5E). Accordingly, the number of moving events in syd-2 mutants is increased, reflecting an increase in STV motility (Figure 5G). In contrast, the capture probability was not affected in syd-2 mutants (Figure 5F). Similarly, for RAB-3 clusters at the mature synapses in the dorsal presynaptic region, the syd-2 mutation significantly increased the dissociation rate (Figure 5I), without significantly affecting the capture probability (Figure 5J). These results indicate that SYD-2 prevents the dispersion of mobile STV packets from anchored STV/AZ complexes. SYD-2 is also known to promote the clustering of several other AZ proteins at the presynaptic terminals (Patel et al., 2006), we therefore asked whether SYD-2 also promotes the association between other AZ proteins and STVs during transport. Indeed, in syd-2 mutants we found a moderate but significant decrease in the ratio of moving RAB-3 clusters associated with UNC-10 (17.8% in syd-2(wy5) vs. 22.9% in wild type, p<0.05, chi-square test, n=936–1212 moving events).
We further carried out time-lapse imaging in arl-8; syd-2 double mutants to better understand how the syd-2 mutation suppresses the arl-8 STV aggregation phenotype. As in syd-2 single mutants, stable puncta in arl-8; syd-2 double mutants exhibited an increased dissociation rate compared to those in arl-8 mutants (Figure 5E). Accordingly, there is a significant increase in the number of moving events and a decrease in the fluorescence intensity of stable puncta en route (Figures 5D, 5G and 5H). On the other hand, syd-2 did not affect the STV capture probability in arl-8 mutants (Figure 5F). Together, our data provide direct evidence that AZ proteins promote STV clustering during axonal transport and raise the possibility that ARL-8 and JNK might control STV aggregation via regulation of STV/AZ interaction during transport.
To determine whether ARL-8 and the JNK pathway regulate STV/AZ association during axonal transport, we performed two-color time-lapse imaging for UNC-10::GFP and mCherry::RAB-3 in the arl-8 and jkk-1 single and double mutants. As in wild-type animals, mobile UNC-10 clusters exhibited a high degree of association with RAB-3 during trafficking in these mutants (97/107, 106/114 and 196/206 mobile UNC-10 clusters associate with RAB-3 in arl-8, arl-8; jkk-1 and jkk-1 mutants, respectively) and the stable clusters also almost completely colocalized (data not shown). However, compared to wild-type animals, we found an increase in the proportion of moving RAB-3 puncta that carry UNC-10 in arl-8 mutants (Figure 6B), suggestive of increased STV/AZ association during transport. Since AZ proteins promote STV clustering during trafficking, the enhanced STV/AZ association may contribute to excessive STV aggregation in arl-8 mutants. This increased STV/AZ association was suppressed in arl-8; jkk-1 double mutants (Figure 6B), consistent with the hypothesis that ARL-8 and the JNK pathway may control STV aggregation during transport in part by regulating STV/AZ association. Further supporting this hypothesis, although the jnk-1 mutation led to strong suppression in arl-8 single mutants, in arl-8; syd-2 double mutants, in which AZ function is already severely defective, the same mutation only produced a subtle effect (Figures 6C–6I).
In light of these findings, we further examined whether ARL-8 and JNK-1 associate with STVs and/or AZ proteins during transport. Previous studies suggested that ARL-8 associates with SVs (Takamori et al., 2006; Klassen et al., 2010). Indeed, we observed that moving GFP::RAB-3 and UNC-10::GFP particles frequently associate with ARL-8::mCherry (Figures S6A–S6F) in the axon shaft. Notably, the stationary ARL-8 puncta also colocalized extensively with RAB-3 or UNC-10 (Figure S6A–S6F, vertical lines in the kymographs).
Dynamic imaging analyses showed that JNK-1 was also actively transported in the axon with pauses en route (Figures S6H and S6K). Interestingly, the majority of moving RAB-3 or UNC-10 packets were not associated with mobile JNK-1 puncta (Figures S6G–S6L). However, the stationary JNK-1 puncta still largely colocalized with the stationary RAB-3 and UNC-10 puncta (Figure S6G–S6L, vertical lines in the kymographs). Therefore, although JNK-1 and the STVs do not move together, they do pause at the same loci. Taken together, the colocalization of STVs, AZ proteins, ARL-8 and JNK-1 at common pause sites along the axon supports the notion that these sites represent regulatory points where ARL-8 and the JNK pathway control the switch between STV trafficking and aggregation.
Presynaptic proteins are transported to the synapses by molecular motors (Goldstein et al., 2008; Hirokawa et al., 2010). Regulation of motor activity may determine where presynaptic cargoes are deposited, thereby impacting synapse distribution. We previously found that overexpression of the kinesin motor UNC-104/KIF1A in DA9 strongly suppressed the arl-8 phenotype (Klassen et al., 2010). In our arl-8 suppressor screen, we further isolated a putative gain-of-function (gf) allele of unc-104, which suppressed the STV and AZ localization defects in arl-8 mutants (Figures 7A–7C and data not shown). We identified the molecular lesion as a G-to-R missense mutation at a highly conserved amino acid (Figure S7A). A mutation in the corresponding residue in human KIF1A (G631) disrupts inhibitory intramolecular interactions between the FHA and CC domains, resulting in increased KIF1A activity (Lee et al., 2004). These results raised the possibility that fine-tuning of motor activity may also serve as a mechanism to regulate STV clustering. Indeed, time-lapse imaging revealed that unc-104(gf) significantly reduced the capture probability in the axon shaft in both wild type and arl-8 mutant animals (Figure 7K). Interestingly, unlike the jkk-1 and syd-2 mutations, unc-104(gf) did not affect the dissociation rate (Figure 7L). Similarly, for RAB-3 clusters at the mature synapses, unc-104(gf) significantly reduced the capture probability (Figure 7M), with no significant effect on the dissociation rate (Figure 7N). Thus, control of UNC-104/KIF1A motor activity likely represents a mechanism that regulates STV capture. Consistent with this hypothesis, unc-104(gf) single mutants exhibited decreased presynaptic SNB-1::YFP puncta size (Figures 7D and 7O). Together, these findings suggest that both STV exit from and entry into stable clusters are subject to molecular regulation. While the exit process is regulated by ARL-8, the JNK-1 pathway and SYD-2, the entry process is controlled by ARL-8 and UNC-104/KIF1A. The fact that the abnormal distribution of presynaptic components in arl-8 mutants can be partially suppressed by either an increase in dissociation caused by the jkk-1 and syd-2 mutations or an decrease in capture efficiency caused by the unc-104(gf) mutation argues that it is the balance between the entry and exit processes that controls the position and size of presynaptic protein clusters. This model predicts that perturbation of both STV entry and exit may produce a stronger phenotype compared to manipulation of a single factor. Indeed, unc-104(gf); jkk-1 double mutants, in which capture is decreased and dissociation increased, showed enhanced reduction in the size of presynaptic SNB-1::YFP puncta compared to either single mutants (Figures 7A, 7D, 7E and 7O), supporting the notion that unc-104 and jkk-1 function in parallel to modulate STV clustering. Conversely, we built double mutants between two partial loss-of-function alleles, arl-8(tm2388) and unc-104(lf) (Figure S7A), in which capture is increased and dissociation decreased. While the single mutants showed much weaker phenotypes than their respective strong loss-of-function mutants (Figures 7G, 7H and 7P), the double mutants showed a strongly enhanced phenotype, with large RAB-3 puncta forming in the ventral axon and commissural regions (Figures 7I and 7P).
The downstream functions of small G proteins are mediated by effector proteins that bind specifically to the GTP-bound form of the G proteins (Donaldson and Jackson, 2011). The strong genetic interaction between arl-8 and unc-104 led us to test whether UNC-104/KIF1A is an ARL-8 effector. We first performed affinity chromatography with glutathione S-transferase (GST)-tagged human ARL8A and various GFP-tagged human KIF1A fragments (Figure 8A; CC1-FHA-CC2, 430–694 aa; CC3-UDR, 694–1209 aa; CC3, 694–775 aa; UDR, 776–1209 aa). We observed specific binding of the KIF1A CC3 domain to GST-ARL8A Q75L (GTP-locked, constitutively active) and to a less extent to wild-type ARL8A, and little binding to ARL8A T34N (GDP-locked, inactive) or GST alone (Figure 8B). Although the function of the CC3 domain is unknown, it is highly conserved between C. elegans and human (Figure S7B). We further confirmed that the binding of wild-type ARL8A to the KIF1A CC3 domain is dependent on GTP binding (Figure 8C). Similar results were also obtained for C. elegans ARL-8 and the UNC-104 CC3 domain (Figure 8D). Furthermore, in yeast two-hybrid assays, we detected interaction of the CC3 domain with ARL8A Q75L but not with ARL8A T34N (Figure 8E). Together, these results suggest that ARL-8/ARL8A physically interacts with the CC3 domain of UNC-104/KIF1A in a GTP-dependent manner.
If the interaction between ARL-8 and the CC3 domain is of functional importance, one prediction would be that overexpression of the UNC-104 CC3 domain may cause a dominant-negative effect and phenocopy the arl-8 mutants by competing with endogenous UNC-104/KIF1A for ARL-8 binding. We therefore overexpressed a C-terminal fragment of UNC-104 containing the CC3, UDR and PH domains in DA9. We included the PH domain as it is known to be required for SV binding (Klopfenstein et al., 2002). Indeed, when over-expressed in the wild-type background, this fragment caused an arl-8-like phenotype (Figures S7C–S7E), leading to a proximal shift of SNB-1::YFP signal. In addition, when over-expressed in the arl-8(tm2388) weak loss-of-function mutants, this fragment caused a strong enhancement of the phenotype (Figures 7J and 7P). The CC3 domain is required for generating this dominant-negative effect as a fragment containing only the UDR and PH domains did not cause any effect (data not shown). Together, these results suggest that UNC-104 functions as a downstream effector of ARL-8 in regulating synapse distribution.
Collectively, our findings provided new insights into how the axonal transport and local assembly of STVs and AZ proteins are coordinated to achieve the proper size, number and location of synapses (Figure S8): In wild-type animals, AZ proteins are associated with STVs during motor-driven axonal transport. STVs undergo frequent stops en route and form immotile STV/AZ clusters due to their intrinsic propensity of aggregation. Trafficking STV packets can cluster with the existing stable puncta, while the stable puncta can also shed a fraction of their content to generate mobile packets. At the pause sites, AZ assembly molecules and JNK promote STV aggregation by limiting dissociation of mobile STVs from the stable clusters, whereas ARL-8 inhibits excessive STV aggregation by promoting STV dissociation. On the other hand, ARL-8 and UNC-104/KIF1A negatively regulate the coalescence of mobile STVs with stable aggregates. In addition, UNC-104/KIF1A functions as an effector of ARL-8 by binding to the GTP-bound form of ARL-8. Together, these findings reveal a molecular network that regulates the balance between presynaptic protein transport and assembly to control the spatial distribution of presynaptic components.
Precise control of the size, number and location of synapses requires regulated distribution of SVs and AZ proteins, a process that is achieved through coordinated transport and assembly of presynaptic material. We identify novel molecular mechanisms that control the balance between presynaptic protein transport and assembly. We show that a JNK MAP kinase pathway and the small G protein ARL-8 antagonistically control a switch between aggregation and trafficking for STVs and AZ proteins, thereby determining their distribution. Interestingly, AZ proteins extensively associate with STVs and promote their aggregation at pause sites during transport. In addition, the anterograde motor UNC-104/KIF1A functions as an effector of ARL-8 and acts in parallel to JNK to control STV capture and synapse distribution. It is conceivable that the trafficking state of presynaptic proteins is favored in the proximal axon to facilitate efficient axonal transport, whereas aggregation prevails at sites of synaptogenesis due to the enhancement in pro-assembly forces and/or inhibition of anti-assembly mechanisms.
The evolutionarily conserved JNKs have been implicated in many critical processes including immunity, stress responses and tumorigenesis (Davis, 2000). In the vertebrate nervous system, JNKs have been involved in stress-induced cell death (Bozyczko-Coyne et al., 2002), regulation of motor binding to microtubules and cargoes (Morfini et al., 2006; Stagi et al., 2006; Horiuchi et al., 2007; Morfini et al., 2009), microtubule dynamics, commissure tract formation and optic fissure closure (Chang et al., 2003; Weston et al., 2003). In C. elegans, JKK-1 and JNK-1, homologs of mammalian MKK7 and JNK3, respectively, are required in the nervous system for coordinated locomotion (Kawasaki et al., 1999). JKK-1 and JNK-1 interact with the scaffold protein UNC-16/JIP3, an adaptor for the UNC-116/KIF5 motor (Byrd et al., 2001). Here we report a novel function of the JNK pathway in promoting presynaptic protein assembly. Inactivation of this pathway strongly suppressed the abnormal synapse distribution in arl-8 mutants. Live imaging revealed that a jkk-1 mutation promotes a trafficking identity for STVs by increasing their dissociation from immotile clusters during transport in arl-8 mutants. In addition, jkk-1 and jnk-1 single mutants also exhibited reduced SV and AZ protein clustering at presynaptic sites.
The C. elegans genome encodes a number of MAP kinases (Sakaguchi, 2004). Of note, the ubiquitin ligase RPM-1 was previously shown to inhibit a DLK-1/p38 MAP kinase pathway to regulate presynaptic development in the DD neurons (Schaefer et al., 2000; Zhen et al., 2000; Nakata et al., 2005). The Drosophila homolog of RPM-1, Highwire, restricts synapse number and size by attenuating a DLK/JNK MAP kinase pathway (Collins et al., 2006). We detected no suppression of arl-8 by dlk-1 mutations (data not shown). In addition, while mutations in the rpm-1 pathway produced dramatic effects on presynaptic organization in the DD motoneurons, they did not cause obvious presynaptic abnormalities in DA9 (data not shown). In contrast, arl-8/jkk-1 interaction strongly impacts synapse distribution in DDs, DA9, as well as multiple other neuronal types. Therefore, there is so far no data supporting a genetic interaction between the arl-8/jkk-1 and the rpm-1/dlk-1 pathway.
Two effectors of ARL8 in regulating lysosomal trafficking were recently identified, including the HOPS complex and the SKIP protein that links lysosomes to the KIF5 motor complex (Garg et al., 2011; Rosa-Ferreira and Munro, 2011). We have investigated their potential involvement in presynaptic development. Firstly, human ARL8B was reported to recruit the HOPS complex to direct lysosomal trafficking (Garg et al., 2011). We examined the phenotypes of deletion mutants in two of the core subunits of the HOPS complex in C. elegans, VPS-16 (ok719) and VPS-18 (tm1125), and an accessory subunit, VPS-39 (ok2442). Despite their lysosomal trafficking and/or lethality phenotypes (Hermann et al., 2005; Kinchen et al., 2008; Xiao et al., 2009), all three mutants appear normal in presynaptic development in DA9 (data not shown). Secondly, the human SKIP protein was proposed to bind to lysosome-localized ARL8B and the kinesin light chain, thus linking lysosomes to the KIF5 motor complex (Rosa-Ferreira and Munro, 2011). Knockdown of ARL8B, SKIP, or KIF5B generated similar changes in lysosome distribution in cultured cells. In the C. elegans genome, the gene Y51H1A.2 encodes the only protein sharing limited sequence homology with SKIP. However, a deletion mutation in this gene (K. Kontani, personal communication) did not cause abnormal SV protein localization (data not shown). In addition, loss-of-function mutations in klc-1 and klc-2, which encode the only two kinesin light chains in C. elegans, did not phenocopy arl-8 in DA9, nor did they enhance or suppress the arl-8 phenotype (data not shown). Furthermore, loss of function in UNC-116/KIF5 did not cause an arl-8-like presynaptic phenotype in DA9 either (data not shown). Collectively, these findings suggest that the JNK pathway represents a novel mechanism that strongly interacts with arl-8 in regulating presynaptic patterning.
It has been suggested that SV and AZ proteins are sorted into different vesicular cargoes at the Golgi. While SV proteins are transported in STVs (Matteoli et al., 1992; Ahmari et al., 2000; Tao-Cheng, 2007), the PTVs are thought to carry AZ material in vertebrate neurons (Zhai et al., 2001; Shapira et al., 2003; Maas et al., 2012). Interestingly, live imaging combined with retrospective EM analysis revealed that STVs are in proximity to dense core vesicles in the axon of cultured neurons (Ahmari, 2000). Immuno EM analyses using antibodies against the AZ proteins Piccolo and Basson and a number of SV proteins further provided ultrastructure evidence that various AZ and SV proteins may be preassembled into multi-vesicle transport aggregates (Tao-Cheng, 2007). In addition, two-color live imaging in cultured neurons also revealed that a proportion of STVs and PTVs are co-transported (Bury and Sabo, 2011). Consistent with these findings, we found extensive association between AZ proteins and STVs during transport in vivo. The association of various presynaptic components prior to synapse formation provides a mechanism for the co-regulation of their axonal transport and assembly, explaining the high degree of colocalization even in the absence of synaptic patterning cues and how the same molecular pathways regulate the distribution of both AZ and SV proteins.
Dynamic imaging analyses of STVs, AZ markers, ARL-8 and JNK-1 showed that all of them exhibited saltatory movements and largely shared identical pause sites during transport. These pause sites appear to represent regulatory points where the switch between the trafficking and aggregation states for STVs is controlled. Trafficking STV packets can stop moving and cluster with the existing stable puncta, potentially through interaction between presynaptic cargoes. The stable puncta can also shed motile packets. The balance between trafficking and aggregation is critically dependent on arl-8 and the JNK pathway. Interestingly, the AZ assembly proteins not only promote SV clustering at the presynaptic terminals, but also prevent STV dissociation from stable clusters en route. Furthermore, AZ/STV association during transport is antagonistically regulated by arl-8 and JNK. Together, these data are consistent with a model in which arl-8 and the JNK pathway control STV aggregation and trafficking by modulating STV/AZ interaction during transport. Interestingly, it has been shown that in cultured vertebrate neurons, synapses form preferentially at predefined STV pause sites upon axodendritic contacts (Sabo et al., 2006). Therefore, regulation of the balance between STV capture and dissociation at the pause sites may represent a general mechanism to control the distribution of presynaptic components.
SV and AZ components are delivered to the presynapses by motor proteins (Goldstein et al., 2008; Hirokawa et al., 2010). At the synapses, the motors may need to be inactivated in order to unload their cargoes. Therefore, regulation of motor activity may dictate where presynaptic cargoes are deposited, thereby determining the spatial pattern of synapses. As a critical motor for the axonal transport of presynaptic proteins, UNC-104/KIF1A is under several levels of intricate regulation. It is activated by phospholipid binding and dimerization (Hall and Hedgecock, 1991; Klopfenstein et al., 2002; Tomishige et al., 2002). In addition, intramolecular interactions between the NC and CC1 domains, FHA and CC2 domains, or FHA and CC1 domains have been shown to modulate UNC-104/KIF1A activity (Al-Bassam et al., 2003; Lee et al., 2004; Huo et al., 2012). Besides intramolecular regulation, the protein DENN/MADD has been identified as an adaptor between SVs and KIF1A (Niwa et al., 2008). SYD-2/Liprin-α has also been suggested to promote the clustering of monomeric UNC-104/KIF1A, thus enhancing its activity (Wagner et al., 2009). Here we show that ARL-8 likely represents a novel mechanism for UNC-104/KIF1A regulation. The GTP-bound form of ARL-8/ARL8A, but not the GDP-bound form, binds specifically to the CC3 domain of UNC-104/KIF1A. Overexpression of the UNC-104 CC3 domain phenocopies the arl-8 mutant in a wild-type background, and enhances the phenotype in a weak loss-of-function arl-8 mutant. Furthermore, overexpression of wild-type UNC-104 or a gain-of-function mutation in unc-104 partially and strongly suppressed the phenotype in arl-8 mutants. Dynamic imaging revealed that this gain-of-function mutation decreases the capture of mobile STV packets by stable clusters whereas the arl-8 mutation leads to increased capture. Conversely, a weak loss-of-function mutation in unc-104 strongly enhances the phenotype in weak loss-of-function arl-8 mutants. Together, these findings identify UNC-104/KIF1A as the first ARL-8 effector in regulating synapse distribution. The conformational changes in small G proteins triggered by GTP/GDP binding might serve as switches to control motor-cargo association, motor processivity, and/or motor binding to microtubules.
Collectively, our findings underlie an intimate link between transport regulation and the spatial patterning of synapses. We also uncovered molecular players that control the stop-go transitions for presynaptic cargoes to achieve appropriate synapse distribution. Interestingly, a recent study suggests that the even distribution of dense core vesicles among synaptic boutons at the Drosophila neuromuscular junction is also achieved by coordinating cargo transport and capture (Wong et al., 2012). Similar cellular strategies might also be utilized to achieve proper distribution of other cargoes, such as lysosomes, mitochondria and neurotransmitter receptors.
Worms were raised on OP50 E. coli seeded NGM plates at 20°C, excepting for the dynamic imaging experiments as detailed below. The mutant strains CZ5730 dlk-1(ju476)I, VC548 vps-16(ok719)III/hT2[bli-4(e937) let-?(q782) qIs48](I;III), VC8 jnk-1(gk7)IV, RB1975 klc-1(ok2609)IV, VC2542 vps-39(ok2442)V/nT1[qIs51](IV;V), and KU2 jkk-1(km2)X were obtained through the Caenorhabditis Genetics Center. wyIs292III (Punc-47::unc-10::tdTomato, Punc-129dorsal muscle::nlg-1::yfp) was kindly provided by G. Maro, klc-2(km11)V by K. Matsumoto, and krIs1V (Punc-47::snb-1::cfp, unc-49::YFP) by J. Bessereau. N2 Bristol was utilized as the wild-type reference strain.
Expression clones were made in the pSM vector, a derivative of pPD49.26 (A. Fire) with extra cloning sites (S. McCarroll and C. I. Bargmann, personal communication). The following plasmids and strains were generated as previously described: wyIs85 (Pitr-1 pB::gfp::rab-3), wyIs92 (Pmig-13::snb-1::yfp), wyIs109 (Pmig-13::cfp::rab-3, Pmig-13::snb-1::yfp, Pmig-13::sng-1::mcherry), wyEx771 (Pmig-13::gfp::ccb-1, Pmig-13::mCherry::rab-3), wyEx1199 (Punc-129DB neuron::gfp::rab-3). Other transgenic strains were generated using standard techniques as detailed in Supplemental Experimental Procedures.
Other methods are detailed in Supplemental Experimental Procedures.
This work was supported by the Howard Hughes Medical Institute, the National Institutes of Health NIH 5R01 NS048392, the National Major Basic Research Program of China (2011CB910503), and the National Natural Science Foundation of China (31070657 and 31190062). We thank the Caenorhabditis Genetics Center, the National Bioresource Project (Japan), G. Maro, P. Kurshan, and the labs of J. Bessereau, K. Kontani, S. Mitani, Y. Jin, and S. Munro for strains and reagents. We also thank S. Shaham for help with the analysis of the whole-genome sequencing data, C. Gao and Y. Fu for technical assistance, and A. McAllister, M. Klassen, and P. Kushan for thoughtful comments on the manuscript. Y. W. and K. S. designed experiments and wrote the paper. C. M. built some transgenes for dynamic imaging. L. H. and W. F. performed the experiments in Figure 8. Y. W. performed the other experiments and analyzed the data.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.