Within axons, molecular motors transport essential components required for neuronal growth and viability. Although many levels of control and regulation must exist for proper anterograde and retrograde transport of vital proteins, little is known about these mechanisms. We previously showed that presenilin (PS), a gene involved in Alzheimer's disease (AD), influences kinesin-1 and dynein function in vivo. Here, we show that these PS-mediated effects on motor protein function are via a pathway that involves glycogen synthase kinase-3β (GSK-3β). PS genetically interacts with GSK-3β in an activity-dependent manner. Excess of active GSK-3β perturbed axonal transport by causing axonal blockages, which were enhanced by reduction of kinesin-1 or dynein. These GSK-3β-mediated axonal defects do not appear to be caused by disruptions or alterations in microtubules (MTs). Excess of non-functional GSK-3β did not affect axonal transport. Strikingly, GSK-3β-activity-dependent axonal transport defects were enhanced by reduction of PS. Collectively, our findings suggest that PS and GSK-3β are required for normal motor protein function. Our observations propose a model, in which PS likely plays a role in regulating GSK-3β activity during transport. These findings have important implications for our understanding of the complex regulatory machinery that must exist in vivo and how this system is coordinated during the motility of vesicles within axons.
Glycogen synthase kinase-3 (GSK-3) is an ubiquitously expressed serine/threonine kinase that is found abundantly in the brain (1,2). The activity of GSK3 controls multiple neurodevelopmental processes, including neurogenesis, neuronal migration, neuronal polarization and axon growth and guidance (3). GSK-3β activity is highly regulated by phosphorylation (4): phosphorylation of GSK-3β at Ser9 results in the generation of an intra-molecular pseudo-substrate, which inactivates the enzymatic activity of GSK-3β (5), while phosphorylation at Tyr216 enhances GSK-3β activity (4,6). Although phosphorylation is the most common form of regulation, recent work has shown that protein complex formation and intracellular localization also play important regulatory roles in the activity of GSK-3β (7). Interestingly, changes in GSK-3 activity are associated with many psychiatric and neurodegenerative diseases, such as Alzheimer's disease (AD), schizophrenia and autism spectrum disorders (4).
An intriguing GSK-3β-binding protein is presenilin (PS), a key gene involved in AD. Although several reports indicate that GSK-3β binds to PS (8–10), the functional role this interaction plays is unclear. One study reported a three times increase in the amount of GSK-3β bound to mutant PS compared with wild-type (WT) PS (9), while another reported that mutant PS bound less GSK-3β than WT PS (10). Other work demonstrated that mutations of PS decreased its affinity for GSK-3β relative to WT PS (8). Despite these discrepancies, these studies postulate that GSK-3β and PS have an important functional relationship, although the significance of this relationship and its affect on disease are unknown.
Two recent studies demonstrated that endogenous PS or GSK-3β could regulate the bidirectional transport of APP vesicles within axons (11,12). Reduction of PS or GSK-3β increased both the anterograde and retrograde velocities of APP vesicles in Drosophila axons, suggesting that PS and GSK-3β together may directly influence motor protein function during axonal transport. Other studies done in cultured cells found that GSK-3β activity increased either in the presence of a FAD-linked PS mutation or with loss of PS function, leading to the proposal that deregulation of GSK-3β activity by loss of PS or familial AD-linked PS mutations can impair axonal transport (13,14). Further, increased GSK-3β activity appeared to increase the levels of kinesin light chain (KLC) phosphorylation leading to a reduction in membrane bound kinesin-1, disrupting anterograde transport (13,14). Intriguingly, kinesin heavy chain (KHC), the motor domain that interacts with microtubules (MTs) for motility is also phosphorylated (15,16) and KHC phosphorylation also induces membrane association (15). However, whether GSK-3β-mediated phosphorylation events on motor proteins control motor function during axonal transport in vivo and what role PS plays in this pathway remain allusive.
On the other hand, since GSK-3β is a tau kinase (17), perhaps the observed GSK-3β and PS-mediated effects on transport could arise due to abnormal tau function and associated MT defects. While excess of tau bound to MTs can physically impair axonal transport (18), destabilization of MTs by loss of tau function may also affect axonal transport. In vitro experiments found that the amount of tau associated with MTs can differentially modulate kinesin and dynein activities (19). Moreover, while tau phosphorylation can regulate its association with motor machinery (20), reductions in axonal transport mediated by kinesin can exacerbate tau hyperphosphorylation (21). However, whether PS also influences tau function and how MT abnormalities contribute to the observed transport defects remain unclear.
Here, we used genetics, in vivo and biochemical analysis in a simple model system to directly test the hypothesis that PS and GSK-3β function together during axonal transport. Our observations show that PS influences GSK-3β-activity during axonal transport via a pathway that affects both kinesin-1 and dynein function, but not tau. Our results propose a model in which PS may function to modulate GSK-3β activity, which may influence kinesin-1 and dynein membrane binding, perhaps by controlling proper motor function during fast axonal transport.
PS genetically interacts with GSK-3β during axonal transport
One simple prediction of the hypothesis that PS influences GSK-3β-mediated roles during axonal transport is that PS and GSK-3β should functionally interact with each other. To test this prediction, we analyzed Drosophila larvae carrying mutations of both the Drosophila homolog of GSK-3β, Shaggy (sgg) and PS (psn). Previous work demonstrated that Drosophila larvae containing mutations in genes encoding motor proteins show dramatic neuromuscular pathology and flip their tails (22,23). The segmental nerves from these mutant larvae contained prominently stained accumulations of synaptic vesicle proteins such as synaptotagmin (SYT) and cysteine string protein (CSP) (22,23). These synaptic accumulations are referred to as the axonal blockage phenotype. This phenotype which can be observed both in fixed (23–29) or live larval preparations (25,28,30) has successfully led to the identification of several novel genes involved in axonal transport (25,26,31). Under transmission EM, many types of identifiable axonal cargo (mitochondria, clear vesicles, dense core vesicles, large multi-vesicular bodies and large, dark prelysosomal vacuoles) are observed within these axonal swellings (22,23,28).
The Drosophila sgg gene shares >90% sequence similarity with human GSK-3β with the greatest similarity seen within the kinase domain (17). Sgg is also regulated by phosphorylation similar to human GSK-3β (4). Most loss-of-function mutations of sgg results in embryonic or larval lethality and do not eclose to adults. We used two loss-of-function sgg mutations; the amorphic loss-of-function mutant sgg[M11], which is a well-characterized spontaneously occurring mutation that causes death during embryogenesis (32), and the X-ray-induced amorphic early larval lethal mutant sgg (17). Larvae heterozygous for sgg [M11]−/+ or sgg−/+ do not show detectable levels of axonal blocks and are comparable with WT (Fig. 1A and B).
Unlike mammals, Drosophila has one PS gene, which produces two alternatively spliced forms. Drosophila PS is ∼50% amino acid sequence identity to human PS (33–35) and loss-of-function PS mutations result in lethal notch-like phenotypes (36,37). Similar to mammalian PS, two intramembranous aspartate residues are conserved in Drosophila PS (33,35). We used two loss-of-function mutations in the Drosophila PS gene; the amorphic psn[I2] mutation is a point mutation that produces a truncated product ending at W278 (37,38) causing a null mutation that results in death at early larval stages, and the psn, which is a deletion and/or splice site mutation resulting in the deletion of amino acids between the first transmembrane domain and the middle of the fourth transmembrane domain (Annette Parks communication to Flybase). This psn encodes a loss-of-function mutation and results in death at late larval stages. Quantitative PCR analysis confirmed that heterozygous combinations of psn[I2]−/+ or psn−/+ larvae express reduced amounts of PS mRNA compared with WT larvae (12). Homozygous psn −/− larvae or heterozygous psn−/+ or the psn[I2]−/+ larvae do not show detectable levels of axonal blocks and were comparable with WT, similar to our previous observations (Fig. 1A, C, E, (12)). These observations are similar to what was seen in larvae with reductions of kinesin-1 or dynein, and only larvae containing complete loss of function of motors showed axonal transport defects (22–24).
To test the prediction that PS and GSK-3β functionally interact with each other, we generated larvae that contained 50% reduction of SGG and 50% reduction of PS. The amorphic sgg alleles sgg[M11] or sgg[I] were crossed to amorphic PS alleles, psn[I2] or psn. We observed large amounts of axonal accumulations that contained CSP in nerves from larvae containing 50% reduction of SGG and 50% reduction of PS (sgg[M11]−/+; psn[I2]−/+, Fig. 1D or sgg[M11]−/+; psn−/+, Fig. 1F). Quantification analysis indicated that the number of blockages in these larvae (sgg[M11]−/+; psn[I2] or sgg[M11]−/+; psn−/+) was significantly increased compared with WT, heterozygous sgg (sgg[M11]−/+) or psn larvae (psn[I2] or psn−/+, Fig. 1G, P < 0.001). Larvae that were sgg−/+; psn[I2]−/+ or sgg−/+; psn−/+ also showed significant amounts of axonal blocks (Fig. 1G, P < 0.001). The extent of axonal blockages in these larvae was comparable with what was seen in larvae with complete loss of function of kinesin-1 or dynein (22–24). These observations suggest that GSK-3β and PS have an important functional relationship during axonal transport.
Since GSK-3β is a tau kinase, and tau is an MT-binding protein, it is possible that the axonal blockages we observed arose due to PS-dependent GSK-3β-mediated effects on tau. In this context, at least two possibilities could exist to cause these axonal defects: (i) changes in tau–MT interactions affecting MT stability could lead to blockages (39–41), and/or (ii) changes in tau phosphorylation could affect motor protein motility (20,21,42–45) causing blockages. To test these, we examined the stability of axonal MTs in vivo in the context of reduction of GSK-3β or reduction of PS using hTAU-GFP. Larvae expressing hTAU-GFP showed smooth GFP MT tracks (Supplementary Material, Fig. S1A), which also stained with Futsch, an MT-binding protein and acetylated tubulin (Supplementary Material, Fig. S1G and H). Motility of mRFP-tagged vesicles was observed moving along these GFP tracks (Supplementary Material, Fig. S1F). Although it was unclear as to how much of the expressed hTAU-GFP was bound to endogenous MTs, we were able to readily observe fragmented MTs under conditions of injury (Supplementary Material, Fig. S1B) similar to previous observations (46). However, reduction of PS or GSK-3β did not cause fragmented MTs as assayed in vivo in hTAU-GFP; psn[I2]−/+ (Supplementary Material, Fig. S1I) or hTAU-GFP;sgg[M11]−/+ (Supplementary Material, Fig. S1J) larval axons. Moreover, excess of GSK-3β also did not cause fragmented MTs (hTAU-GFP;sggWT, Supplementary Material, Fig. S1K). Further, our observations were consistent with a recent study that showed that only complete loss of GSK-3β altered MT stability, while reduction or over expression of GSK-3β had no effect on MTs (11). Therefore, while a complete loss of GSK-3β led to increased MT stability (11) and reduction of GSK-3β or PS did not have an observable effect on MTs (Supplementary Material, Fig. S1) despite impairing transport, the PS-dependent GSK-3β-mediated transport defects we observed may not be primarily caused by MT abnormalities. However, whether the observed transport defects were due to PS-dependent changes on GSK-3β-mediated tau phosphorylation remains unclear and needs further investigation.
PS influences active GSK-3β during axonal transport
Several signaling pathways have been found to regulate GSK3β activity by upstream regulators in response to stimuli (3). Similarly, PS may function as an upstream regulator that modulates GSK-3β activity during axonal transport. Indeed, previous work in cultured cells found that loss of function of PS or a mutation in PS increased phosphorylation of KLC by influencing GSK-3β activity, leading to a reduction of membrane bound kinesin-1 (13,14), postulating that PS may function to modulate GSK-3β activity during axonal transport. One prediction of this proposal is that loss of PS should stimulate GSK-3β activity. To test this, we used biochemical analysis to evaluate GSK-3β activity in heterozygous PS larvae (psn[I2]−/+), since null mutants of PS die at early larval stages. In contrast to previous observations in cells, we found that the level of active GSK-3β was significantly decreased in the postnuclear supernatant (PNS) of psn[I2]−/+ larvae compared with WT larvae (Fig. 2A and B, P < 0.005). Reduction of PS also decreased the amount of membrane bound active GSK-3β (Fig. 2A and B, P < 0.05). Strikingly, the amount of kinesin-1 and dynein bound to membranes was also decreased in psn[I2]−/+ membranes compared with WT membranes (Fig. 2A and B, (P < 0.05), suggesting that the binding of motors to membranes may be dependent on PS and active GSK-3β. Consistent with this, the amount of kinesin-1 and dynein in the soluble fraction was increased with reduction of PS (Supplementary Material, Fig. S2). Collectively, our results suggest that under physiological conditions PS can function to modulate GSK-3β activity to influence both kinesin-1 and dynein motor binding to membranes. Therefore, decreases in membrane bound motors mediated by reduction of GSK-3β activity could cause the axonal transport defects observed in sgg−/+; psn−/+ larvae (Fig. 1D and F).
The proposal that GSK-3β activity is necessary for axonal transport leads to the prediction that excess of active GSK-3β should perturb axonal transport by increasing motor binding to membranes. Indeed, excess of SGGACTIVE but not excess of a kinase-dead or non-functional form of SGG (SGGN) caused axonal transport defects (Fig. 3A–D). Similarly, excess of SGGWT, which had increased levels of active GSK-3β, also caused axonal transport defects (Fig. 3A and D). Excess of active GSK-3β was observed in both the PNS and the membrane fraction of larval brains expressing SGGACTIVE or SGGWT compared with WT larval brains (Fig. 4A and data not shown). To biochemically confirm that increased GSK-3β activity increases motor binding to membranes, we evaluated the amount of motor proteins bound to membranes in brains expressing SGGACTIVE and compared these levels with the levels observed in WT (Fig. 4A). Quantification of three independent experiments revealed a two- to three-fold increase of kinesin-1 in membranes from SGGACTIVE brains compared with WT brains (P < 0.05), while no significant change was seen in the level of kinesin-1 in the PNS (Fig. 4A). Intriguingly, we also observed a significant increase in the level of dynein bound to membranes from SGGACTIVE brains compared with WT brains (P < 0.05, Fig. 4A). Similar observations were also seen in brains expressing SGGWT (data not shown). In contrast, significant decreases were observed in the level of kinesin-1 and dynein in membranes from mutant sgg larval brains (sgg[M11]−/+) compared with WT brains (P < 0.05, Fig. 4B). Taken together, our results indicate that the activity of GSK-3β is required for motor protein binding to membranes. Therefore, excess GSK-3β activity can perturb axonal transport by increasing the amount of motors on moving vesicles.
PS-dependent GSK-3β activity negatively controls motor function during axonal transport
The proposal that PS influences GSK-3β activity leads to the prediction that reduction of PS in the context of excess GSK-3β should suppress GSK-3β activity-dependent transport defects by decreasing motor binding to membranes. In contrast to this prediction, we found that 50% reduction of PS with excess of SGGACTIVE failed to suppress axonal blocks (Fig. 5A–C), but instead caused an enhancement of axonal blockages compared with larvae expressing SGGACTIVE alone. Enhanced axonal blockages were observed in both SGGACTIVE; psn[I2]−/+ or SGGACTIVE; psn−/+ larvae (Fig. 5B and C). These larvae also showed the characteristic tail flip phenotype and did not eclose to adults (data not shown). Reduction of PS had no effect on the kinase-dead or non-functional form of GSK-3β (SGGN, Fig. 5D and E). Quantitative analysis revealed that the number of blockages in SGGACTIVE; psn[I2]−/+ and SGGACTIVE; psn−/+ larvae was significantly increased compared with SGGACTIVE alone (P < 0.001, Fig. 5F). Similar observations were also seen in SGGWT; psn[I2]−/+ and SGGWT; psn−/+ larvae (Fig. 5F). The extent of blockages in these larvae was comparable with the extent of axonal defects seen with complete loss of function of motor proteins (27). Collectively, these observations suggest that while GSK-3β activity is important for motor binding to membranes, active GSK-3β may also influence the proper control of motor protein function during vesicle motility in vivo. Indeed, recent work has suggested that GSK-3β and PS may act as negative regulators to control the number of active motors that are moving cargoes in vivo (11,12). Therefore, reduction of PS in the context of excess active GSK-3β may enhance axonal transport defects by improperly regulating motor protein function during cargo motility.
Although in vitro studies suggest that the binding of motors to membranes activates motor function (13), it is unclear how motor activity is initiated or controlled in vivo. The proposal that PS may influence GSK-3β-activity-dependent control of motor proteins leads to the prediction that activity changes in GSK-3β should significantly affect vesicle motility during transport. Therefore, both decreases and increases in GSK-3β activity should significantly influence vesicle velocities in vivo. Indeed, consistent with this proposal, reduction of PS decreased GSK-3β activity (Fig. 2), but increased both the anterograde and retrograde vesicle velocities (12) likely by improperly regulating motors. Similarly, reduction of GSK-3β also increased both the anterograde and retrograde vesicle velocities (11). Furthermore, excess of GSK-3β significantly decreased the average anterograde and retrograde duration-weighted segmental velocities of synaptobrevin vesicles (synb-GFP) or mitochondria (mito-GFP), two vesicles/organelles transport by kinesin-1 in the context of excess GSK-3β (Fig. 6A and B, Supplementary Material, Fig. S3A–Dand Table S1, P < 0.001 (47,48). In larvae expressing excess SGGWT, the average velocities of synaptobrevin vesicles significantly decreased to 0.118 and 0.099 µm/s from 0.283 and 0.294 µm/s in the anterograde and retrograde directions, respectively (Fig. 6C and D, Supplementary Material, Fig. S3A and B, P < 0.001; Supplementary Material, Table S1 and Movies S1 and S2). In larvae expressing excess SGGWT, the average velocities of mitochondria significantly decreased to 0.364 and 0.233 µm/s from 0.511 and 0.457 µm/s in the anterograde and retrograde directions, respectively (Fig. 6E and F, Supplementary Material, Fig. S3C and D, P < 0.001, Supplementary Material, Table S1 and Movies S3 and S4). Similar to our observations, Weaver et al. (11) also reported that excess GSK-3β decreased both the anterograde and retrograde velocities of APP vesicles, another vesicle transported by kinesin-1.
Substantial changes in pause durations, pause frequency and run lengths were also observed, indicating that vesicle velocity changes can occur due to changes in the frequency and duration of vesicle pauses causing changes in vesicle runs. Significant increases were seen in retrograde pause duration and pause frequency (P < 0.05), and in the anterograde pause frequency (P < 0.005) in synb-GFP, while only a mild change was detected in the anterograde pause duration in synb-GFP in larvae expressing SGGWT compared with larvae containing normal levels of SGG (P = 0.0647, Fig. 6G–J). Significant increases in retrograde pause duration and pause frequency were observed in mito-GFP, while no significant changes were detected in anterograde pause duration or pause frequency in mito-GFP in larvae expressing SGGWT compared with larvae containing normal levels of SGG (Fig. 6G–J). Consistent with the prediction that increased vesicle pausing should decrease vesicle run lengths, significant decreases in the anterograde synb-GFP (P < 0.05) and retrograde mito-GFP (P < 0.05) run lengths were seen in larvae expressing SGGWT compared with larvae containing normal levels of SGG (Fig. 6K and L), while a trend toward decreased run lengths were seen in retrograde synb-GFP (P = 0.158) and anterograde mito-GFP (P = 0.063). Such changes in the motility of synaptobrevin vesicles or mitochondria could be due to changes in motor protein function. Indeed, cargo population analysis showed increased stalled or static vesicles/organelles with excess SGG compared with normal SGG, consistent with decreased populations of anterograde, retrograde or reversing vesicles/organelles (Supplementary Material, Fig. S2E and F). The significant increases in both synb-GFP and mito-GFP (P < 0.05) switch frequencies observed in larvae expressing SGGWT compared with larvae containing normal levels of SGG (Fig. 6M) suggest that the corporative coordination of kinesin-1 and dynein motor function on synaptobrevin vesicles and mitochondria could be disrupted. Therefore, taken together these observations suggest that GSK-3β activity can negatively influence the coordinated function of both kinesin-1 and dynein motor function on vesicles during motility on MTs, and that this role may also be dependent on PS. Consistent with this, we found that reduction of motors with excess active GSK-3β enhanced axonal transport defects (Supplementary Material, Fig. S3). Therefore, while excess GSK-3β increases motor binding to membranes, motor function activities on vesicles are negatively regulated causing improper motor function and motility defects (Supplementary Material, Fig. S6).
We have identified a novel physiological role for PS and GSK-3β during axonal transport using genetics, in vivo imaging and biochemical analysis in Drosophila. Specifically, our observations lead us to two major conclusions: (i) PS and GSK-3β functionally interact with each other during axonal transport, and (ii) PS influences GSK-3β-activity for proper axonal transport via a pathway that affects both kinesin-1 and dynein function. These findings have important implications for our understanding of the complex regulatory machinery that is involved in the movement of vesicles by both kinesin-1 and dynein motors in vivo.
PS modulates GSK-3β activity during axonal transport
In vitro studies previously suggested that PS may influence GSK3β activity to affect kinesin-1-mediated axonal transport (13,14). Consistent with this proposal, we found that GSK-3β and PS functionally interacts with each other in the context of axonal transport in vivo. Our genetic analysis showed that while reduction of GSK-3β or PS alone did not cause axonal transport defects, reduction of both GSK-3β and PS caused axonal blockages (Fig. 1). There are at least three possibilities to explain the cause of these axonal transport defects. One possibility is that both PS and GSK-3β function together during axonal transport via an action on motors. PS is transported bi-directionally within PNS and CNS axons (49–51), and is contained within APP vesicles (52). In vivo, reduction of PS altered both the anterograde and retrograde velocities of vesicles by influencing both kinesin-1 and dynein motors (12). While the transport of GSK-3β has not been directly observed, GSK-3β is present in neurons and neuronal membranes that contain kinesin-1 (13,14) (Fig. 4) and dynein (Fig. 4). Similar to PS, GSK-3β also influences both anterograde and retrograde vesicle velocities in vivo (11). Since reduction of PS decreased GSK-3β activity and kinesin-1 and dynein binding to membranes (Fig. 2), and excess active GSK-3β decreased vesicle velocities by increasing pause frequencies/durations and switch frequencies, and decreasing run lengths (Fig. 6), perhaps PS functions to modulate GSK-3β activity for corporative and coordinated kinesin-1 and dynein motor function during transport (Supplementary Material, Fig. S6).
The second possibility is that the PS and GSK-3β-mediated axonal blockages we observed could result due to defects in MT stability. However, reduction of GSK-3β or PS, or excess of GSK-3β did not cause any observable change in MT stability as assayed using Tau-GFP despite impairing transport (Supplementary Material, Fig. S1; Figs 1 and 3). Similarly, recent observations using EB1-GFP also failed to detect MT defects with reduction or excess of GSK-3β (11). Perhaps, subtle changes in MTs could exist that escaped our analysis, which could impact vesicle motility leading to transport defects.
A third possibility is that PS may influence GSK-3β-mediated tau-dependent axonal transport defects in a phosphorylation-dependent manner. Mechanistically, GSK-3β-mediated tau-dependent transport defects could arise due to changes in the phosphorylation state of tau and the amount of tau associated with MTs, which could interfere with the binding of kinesin-1 and dynein motors to MTs (19,20,44). Indeed, Mudher et al. (18) found that excess active GSK-3β can enhance tau-dependent axonal defects, but this study did not assess the biochemical activity of GSK-3β or the binding of motors to MTs. Although currently little is known about a role for PS in GSK-3β-mediated tau phosphorylation, perhaps most likely some combination of these events led to the axonal transport defects we observed. While further studies would be needed to evaluate the contribution of each of these factors, our work raises the possibility that defects in regulatory pathways mediated by PS and GSK-3β directly impairs vesicle transport in vivo.
Is motor protein function influenced by GSK-3β phosphorylation?
It is possible that phosphorylation plays a key role in the regulation of motor protein activity and function. Previous work in cells suggested that the binding of kinesin-1 to membranes was dependent on the phosphorylation state of KLC, which is thought to be controlled by GSK-3β (14,16). While our in vivo findings also suggest that GSK-3β activity influenced membrane binding of motor proteins (Fig. 4), GSK-3β activity appears to also influence motor protein function (Fig. 6), perhaps via phosphorylation events. Although both KHC and KLC are phosphorylated under physiological conditions (15,53), the exact identity of the kinase(s) responsible for this action and the role differential phosphorylation of kinesin-1 subunits play in motor function are unclear. Furthermore, both KLC and KHC contain AMPK phosphorylation sites, in addition to GSK-3β phosphorylation sites (54). Sites for other kinases are also present on both KLC and KHC (53,55), indicating that phosphorylation of kinesin-1 and its subunits can play an important role in regulating its motor function in vivo (56). Unlike mammalian KLC, the protein sequence of Drosophila KLC lacks the XSXXXSX GSK-3β phosphorylation consensus sequence (57), while Drosophila KHC has two putative GSK-3β phosphorylation sequences. One of these putative sequences is within the kinesin motor domain and the other is within the stalk. Intriguingly, Lee et al. (15) previously proposed that phosphorylation of KHC, not KLC was required to induce membrane association, indicating that phosphorylation of KHC could influence membrane binding or motor function or both. Since we found that active GSK-3β influenced both motor binding to membranes (Fig. 4) and motor function (Fig. 6), perhaps GSK-3β-dependent phosphorylation of KHC could dictate the balance between these roles. It is conceivable that differential phosphorylation of the sites in the motor or stalk domains may coordinate the regulation of these functions. Although further study is needed to test these predictions, it is clear that under physiological conditions, GSK-3β plays an important role in controlling kinesin-1 function in vivo.
Our results also provide the first direct evidence for GSK-3β in regulating dynein-mediated transport. We found that GSK-3β activity influenced dynein binding to membranes and dynein-mediated retrograde vesicle velocities (Figs 4 and 6), indicating that these actions could also be controlled by GSK-3β-dependent phosphorylation of dynein. Indeed, phosphorylation has recently been shown to regulate targeting of cytoplasmic dynein to kinetochores during mitosis (58). A role for phosphorylation in the regulation of dynein light chain assembly has also been demonstrated (59). An early study reported that dynein associated with anterogradely moving organelles was phosphorylated, suggesting that differential phosphorylation could regulate dynein function (60). Intriguingly, Drosophila dynein heavy chain (DHC) has several putative GSK-3β phosphorylation consensus sequences, including one in DHC's motor domain. Although further study is needed, perhaps GSK-3β-dependent phosphorylation of dynein could dictate proper dynein motor function in vivo.
Does GSK-3β control PS function during axonal transport?
Although we found that the GSK-3β activity-mediated effects on transport were dependent on PS (Figs 1and 5), it is possible that GSK-3β may also control PS function during axonal transport and a feedback regulatory mechanism could exist under physiological conditions. Early work identified conserved GSK-3β consensus phosphorylation sites on PS (61,62). Phosphorylation of PS at one of these GSK-3β motifs mediated the interaction between PS and β-catenin (61), while phosphorylation at the second site altered the turnover of PS C-terminal fragment (CTF) (62). Moreover, endogenous GSK-3β was found to be part of a tetrameric complex with PS-CTF/NTF and β-catenin, indicating that a GSK-3β/PS complex plays an important role in the β-catenin pathway (63). Whether the PS-dependent GSK-3β activity changes we observed during axonal transport were also mediated through complex formation between PS, GSK-3β and motor proteins (64) remain to be examined.
In summary, our findings suggest that a complex regulatory mechanism exists in vivo and sheds new light on how this system is coordinated during the movement of vesicles within a living organism. Further, our work proposes that axonal transport defects induced by regulatory problems could contribute to tau abnormalities, and neuronal and synaptic defects observed in Alzheimer's and tau disease pathology (28,65,66). Thus, our work could highlight a potential novel therapeutic pathway for early treatment, prior to neuronal loss and the occurrence of clinical symptoms of disease.
MATERIALS AND METHODS
Two loss-of-function Drosophila Shaggy, sgg[M−11] and sgg , and two Drosophila PS, psnI2 and psn mutants and three transgenic Drosophila GSK-3β lines, UAS-SGGB (SGGWT), UAS-SGGS9A (SGGACTIVE) and UAS-SGGKK83-84MI (SGGN), were used (Bloomington Stock Center, (67)). For loss-of-function experiments, the KHC mutant, khc20 and the dynein light chain mutant, roblk, were used ((22–24). Expression of SGGWT, SGGACTIVE or SGGN was done by crossing these UAS lines to the pan neuronal GAL4 driver APPL-GAL4 (27] at 29°C. For genetic interaction experiments, APPL-GAL4;T(2:3) CyO TM6B, Tb/Pin88K was used. The chromosome carrying T(2:3) CyO TM6B, Tb is referred to as B3 and carries the dominant markers, Hu, Tb and CyO. For genetic interaction tests with kinesin and dynein motors, APPL-GAL4/APPL-GAL4;B3/Pin88K females were crossed to khc20/T(2:3) CyO TM6B or roblk/T(2:3) CyO TM6B,Tb males. Males that were APPL-GAL4/Y;khc20/B3 or APPL-GAL4/Y;roblk/B3 were crossed to females from the sgg transgenic lines and only females that were non-tubby were used for analysis. For genetic interaction tests with loss of function of sgg, sgg[M-11] or sgg  females were crossed to khc20/T(2:3) CyO TM6B or roblk/T(2:3) CyO TM6B,Tb males and females from this cross were used for analysis. For genetic interaction tests with PS, APPL-GAL4/APPL-GAL4;B3/Pin88K females were crossed to psn[I2] or psn /TM6C, Tb males. Males that were APPL-GAL4/Y; psn[I2] or psn /B3 were crossed to females from the sgg transgenic lines (SGGWT) or sgg  or sgg[M-11] and only females that were non-tubby were used for analysis. The transgenic human Tau line UAS-hTAU-GFP (Bloomington) was used. To evaluate MT stability males that were APPL-GAL4/Y; UAS-SGGWT or APPL-GAL4/Y; sgg [M-11] /B3 were crossed to virgin UAS-hTAU-GFP females and only females that were non-tubby was used for analysis.
Larval preparations, immunohistochemistry and quantification
Third instar larvae were dissected, fixed and segmental nerve immunostainings were done as described (68). Briefly, larvae were dissected in dissection buffer (2× stock contains 128 mm NaCl, 4 mm MgCl2, 2 mm KCl, 5 mm HEPES and 36 mm sucrose, pH 7.2). Dissected larvae were fixed in 4% formaldehyde and incubated overnight with antibodies against CSP (1:10, Developmental Studies Hybridoma Bank), GSK-3β (1:100, Cell Signaling) and/or HRP-TR (1:100 Invitrogen). Larvae were incubated in secondary antibodies (Alexa anti-mouse 568 or Alexa anti-rabbit 488, 1:100, Invitrogen) and mounted using Vectashield mounting medium (Vector Labs). Images were collected using an Leica TCS SP2 AOBS Spectral confocal microscope as described (27,68). Quantitative analysis on the extent of blockages was carried out by collecting six confocal optical images from larval neurons from the region directly below or posterior to the larval brain, where several segmental nerves are visible or come into focus through the optical series. For each genotype, four to six animals were imaged, and four nerves were analyzed over a length of 50 μm, using the threshold, density slice and particle analysis functions in NIH image software as previously described (27).
In vivo microscopy of vesicle movement and MT integrity
Drosophila transgenic lines expressing synb-GFP, Mito-GFP or hTAU-GFP were generated as previously described (26,27). Females from this line were crossed to males from APPL-GAL4 or D42GAL4, which express in all neurons. Female larvae were used for all in vivo imaging. To generate D42GAL4;synb-GFP/SGGWT larvae, females that were D42;synb-GFP were crossed to males that were SGGWT/B3 and non-tubby female larvae that were D42GAL;synb-GFP/SGGWT were used. To generate APPL-GAL4;Mito-GFP/SGGWT larvae, females that were APPL-GAL4;SGGWT/B3 were generated and crossed to Mito-GFP males and non-tubby female larvae from this cross was used for in vivo analysis. To express APPL-GAL4;hTAU-GFP/SGGWT, APPL-GAL4;hTAU-GFP/sgg[M11] or APPL-GAL4;hTAU-GFP/ psn[I2] larvae, females that were APPL-GAL4;SGGWT/B3, APPL-GAL4;sgg[M11]/B3, or APPL-GAL4; psn[I2]/B3 were generated and crossed to hTAU-GFP males and non-tubby female larvae from this cross was used for in vivo analysis. Third instar larvae were dissected on a sylgard platform using Ca2+-free buffer containing the following, Nacl (128 mm), EGTA (1 mm), MgCl2 (4 mm), KCl (2 mm), HEPES (5 mm) and sucrose (36 mm) as described in Kuznicki et al. (48). Dissected animals were inverted onto a cover slip and imaged using a Nikon Eclipse TE 2000-U inverted microscope with a Coolsnap HQ camera and a 100×/1.40NA oil objective. One hundred and fifty frames of videos were collected at 200 ms exposure under the control of Metamorph software. For each genotype, four time-lapsed movies were collected for each animal; five animals were imaged; a total of 20 movies were collected.
In vivo movement analysis
In vivo movement analysis was performed as described in Reis et al. (69) and Gunawardena et al. (12). Briefly, GFP-tagged vesicles/organelles in time-lapse movies were detected as single particles as described in Yang et al. (70). Their full trajectories were recovered using customized software based on the modification of the single particle tracking technique described in Ponti et al. (71). Since the particle tracking output contained mainly trajectory segments, a further computational process was performed to link these segments into full trajectories. Finally, a manual process was used to recover trajectories that the software was unable to recover. All recovered trajectories were manually inspected so that errors were either corrected or removed. For each genotype, individual cargoes were automatically classified as being either stationary, anterograde, retrograde or reversing. Cargo trajectories of each genotype were then analyzed by calculating different descriptors that characterize the overall distribution of cargo population and individual cargo behavior in terms of velocity, pause, run lengths and reversals (switches) (12,69). In particular, to determine the velocity of a specific cargo, its trajectory is first partitioned into segments that are uninterrupted by pause or reversal events. For a given direction, either anterograde or retrograde, duration-weighted segmental velocity of the cargo is defined by its total distance of movement divided by its total duration of movement in that direction. This definition effectively weights cargo velocities within different segments by their durations. All data analysis was conducted using customized software written in MATLAB (Mathworks) and C++ and in-depth details are provided in (12,69).
For immunofluorescence analysis of axonal blockages, statistical analysis was performed using Excel (Microsoft Corp.), using the two-sample two-sided Student's t-test, and two other multiple comparison procedures (the Bonferroni and Dunnett procedures) specifically designed to compare each treatment with a control. Differences were considered significant at a significance level of 0.05, which means a 95% statistically significant correlation. All three statistical methods revealed similar significant differences. For in vivo movement analysis, as previously done, data were first checked for normality using three different tests implemented in the nortest package of R: the Lilliefors test, the Anderson-Darling test and the Shapiro-Francis test (12,69). For those that generally follow normal distributions, their means were compared using the two-sample two-sided Student's t-test. For those following non-normal distributions, their means were compared using the permutation T-test (72) or the Wilcoxon rank-sum test.
Membrane floatation and western blot analysis:
As previously described, 5 mls of larval brains from each genotype (psn[I2]−/+, SGGACTIVE or sgg[M11]−/+) were homogenized in acetate buffer (10 mm HEPES, pH 7.4, 100 mm K acetate, 150 mm sucrose, 5 mm EGTA, 3 mm Mg acetate, 1 mm DTT) with proteinase and phosphatase inhibitors (25). Debris was removed by centrifugation at 1000g for 7 min, and the resulting PNS was brought to 40% sucrose, bottom loaded and overlaid with two cushions of 35 and 8% sucrose. The gradient was centrifuged at 50 000 rpm in a TLS55 rotor (Beckman Coulter, Fullerton, CA, USA) for 1 h. Light membranous organelles and membrane-associated proteins floated to the 35/8 interface, whereas heavier membranes and mitochondria were found in the pellet. Equal amounts of protein from the PNS, 35/8 interface, soluble and pellet fractions were analyzed by western blotting. Antibodies pSer9 GSK-3b monoclonal antibody (Abcam at 1:500), pY216 GSK-3b monoclonal antibody (Abcam at 1:500), GSK-3b total monoclonal antibody (Abcam at 1:100), anti-KHC polyclonal antibody (Cytoskeleton at 1:1000), anti-DIC74kd monoclonal antibody (Millipore at 1:100), anti-tubulin monoclonal antibody (Abcam at 1:1000 dilution) and anti-actin monoclonal antibody (Abcam at 1:1000 dilution) were used. Immunoreaction was detected using the ECL kit (Pharmacia) and imaged using QuantityOne (Bio-Rad). Quantification analysis was performed using NIH ImageJ software. Each lane of the gel was analyzed using plot lane, wand and label peaks gel analysis functions. Data obtained as percent values for each sample by Image J were analyzed in Excel (Microsoft Corp.). Relative intensity was calculated by dividing the percent value for each sample by percent value of syntaxin and then normalized to WT, so that WT was 1. Using the two-sample two-sided Student's t-test, differences were considered significant at a significance level of 0.05, which means a 95% statistically significant correlation from three separate membranes from three different experiments.
This work was supported by funds from the John R. Oishei Foundation to SG and a Fulbright Scholarship (Tibetan Fund) to K.D. G.J.I., K.H.Z. and E.S. were supported by fellowships from the University at Buffalo (UB) Center for Undergraduate Research and Creative Activities (CURCA).
We thank Tymish Ohulchanskyy for assistance with the Leica Confocal microscope, Stefan Roberts for assistance with biochemical assays, Ge Yang and Minhua Qiu for assistance with in vivo data analysis, Margaret Hollingsworth for critical reading of the manuscript, Priyantha Karunaratne for constant support and members of the Gunawardena laboratory for constructive discussions.