Introduction

Kinesin-1 is the founding member of the kinesin superfamily of microtubule motor proteins. It moves vesicles in squid giant axons (Vale et al, 1985; Brady et al, 1990; Hirokawa et al, 1991) and has been implicated in the dynamics of the endoplasmic reticulum (ER), Golgi apparatus, secretory vesicles, endocytic structures, mitochondria and mRNP particles (Wozniak et al, 2004). This raises the question as to how one motor can have so many different cargoes, and this may be due to the existence of different motor isoforms. Kinesin-1 comprises two motor subunits (heavy chains, KHCs) and two light chains (KLCs), and in vertebrates, both subunits are encoded by three genes (Xia et al, 1998; Rahman et al, 1999; Junco et al, 2001). In most non-neuronal vertebrate cells, one heavy chain isoform, KIF5B or ubiquitous KHC (uKHC), is expressed along with KLC1 and KCL2. There are many different splicing isoforms of KLC1 (Khodjakov et al, 1998; Gyoeva et al, 2000; McCart et al, 2003), although only one KLC isoform type is present in any given kinesin-1 molecule (Gyoeva et al, 2004).

Both heavy and light chains have been suggested to play a role in attaching kinesin-1 to cargo. The C-terminal region of KHC interacts with a number of proteins, and in fungi such as Neurospora crassa, which lack light chains altogether, this region is sufficient to bind cargo (Seiler et al, 2000). A variety of proteins interact with the KHC C-terminal domain, between amino acids 813–963, while the KLCs interact with amino acids 771–813 (Adio et al, 2006). Importantly, both the KLCs and most potential cargo molecules can bind simultaneously to KHC (Huang et al, 1999; Diefenbach et al, 2002). An exception to this model is milton, which competes with KLCs for binding to the same KHC domain (Glater et al, 2006).

KLCs are important components of kinesin-1 in higher eukaryotes, as loss of KLC function leads to severe defects in neurons and eventually to death (Gindhart et al, 1998; Rahman et al, 1999). One possible role for KLCs is to allow the inhibition of KHC function when kinesin-1 is not attached to cargo (Verhey et al, 1998). Native kinesin-1 can exist in a folded, inactive state where the KHC C-terminal domain interacts with the neck region close to the motor domain of KHC, so inhibiting its ATPase activity (Hackney et al, 1992; Coy et al, 1999; Friedman and Vale, 1999; Stock et al, 1999; Hackney and Stock, 2000; Seiler et al, 2000) and its ability to move along microtubules (Coy et al, 1999). The presence of light chains has been proposed to allow kinesin-1 to switch between the inactive, folded state when it is free in the cytosol, and the extended, active form when it is attached to cargo (Verhey et al, 1998).

KLCs interact with a number of potential cargo molecules. KLCs contain an N-terminal coiled-coil domain, termed the heptad repeat, that binds to KHC. Upstream of this region is a short conserved stretch that interacts with the glycogen synthase kinase 3-binding protein GBP/Frat (Weaver et al, 2003). However, the most important region for cargo binding is thought to be a domain C-terminal to the heptad repeat that contains five tetratricopeptide repeats (TPR), which bind to a wide range of proteins (Adio et al, 2006). Consistent with this idea, an antibody to the TPR domain inhibits kinesin-1-driven motility in squid axoplasm by releasing kinesin-1 from the membrane (Stenoien and Brady, 1997). The TPR domains in the various KLC isoforms are almost identical (Rahman et al, 1998), and therefore both KLC1 and 2 would be expected to interact equally with all the cargo molecules identified so far. In contrast, there is considerable variability in the very C-terminal region of KLCs, both between the KLC1 and 2 genes, and within KLC1 populations, due to alternate splicing (McCart et al, 2003). It has been suggested that this C-terminal region is important for targeting kinesin-1 to different cargoes, since KLC1B or C were enriched on mitochondria (Khodjakov et al, 1998) while KLC1D or E were located on the Golgi apparatus (Gyoeva et al, 2000). Furthermore, the C-terminus of KLC2 was shown to be important for kinesin-1 binding to axonal vesicles, since phosphorylation of serine residues in the very C-terminus of KLC2 leads to release of kinesin-1 from the membrane via the action of Hsc-70 (Tsai et al, 2000; Morfini et al, 2002). The only protein identified so far that interacts with the C-terminus of KLCs is the 14-3-3 protein, which binds to KLC2 when it is phosphorylated on Ser575 (Ichimura et al, 2002).

Here, we aimed to investigate the role of different KLC isoforms in kinesin-1 function. Using in vitro motility assays, we followed the kinesin-1-driven movement of rough endoplasmic reticulum (RER) membranes, and of vesicles present in a Golgi membrane fraction. We show that KLC1 isoform-specific C-terminal domains play an important role in this process.

Results

Golgi and ER membranes have specific KLC splicing variants

Different KLC1 isoforms, generated by alternate splicing (Figure 1A), have been proposed to target kinesin-1 to specific cargoes (Gyoeva et al, 2000). We decided to use in vitro assays for kinesin-1-driven membrane movement to test the role of specific KLC1 isoforms in kinesin-1 function on different organelles.

Our previous work showed that there is plentiful uKHC in a rat liver Golgi fraction (Robertson and Allan, 2000). As expected, uKHC was also present in a rat liver RER fraction (Figure 1B, upper panel). Reblotting the same nitrocellulose membranes with an antibody (KLC_ALL) that recognises all KLC forms (Stenoien and Brady, 1997) revealed that the RER and Golgi membranes contain KLC proteins with different molecular weights (Figure 1B). The RER fraction had a single KLC band, while the Golgi fraction contained one major and two minor bands, in keeping with the RER fraction being more homogeneous than the stacked Golgi fraction (Leelavathi et al, 1970; Allan and Vale, 1991). Based on the size differences, the large isoform present in the Golgi fraction could be KLC1D or E, or KLC2. The RER fraction band could correspond to any of the smaller KLC1 isoforms (A, B, C and F: Figure 1A).

Based on the relative mobilities of the KLC isoforms present in the RER and Golgi fractions, coupled with the published observations that KLC1D/E is associated with the Golgi apparatus (Gyoeva et al, 2000), we decided to test the effects of KLC1B and D, and KLC2, using functional assays. GST-fusion proteins for KLC1B, KLC1D and KLC2 were generated that lacked the heptad repeats that bind to KHCs but contained TPR domain, which is thought to bind to cargo, and the specific carboxyl-terminus (Figure 1C). These fusion proteins are referred to as BTC, DTC and 2TC throughout. In addition, we prepared a carboxyl-terminal fragment of rat uKHC (aa 771–963), which included the KLC-binding domain (Figure 1C). The Coomassie stained purified proteins are shown in Figure 1D.

Exogenous KLC1B, but not KLC1D or KLC2, inhibits motility of RER membranes

Interphase cytosol obtained from Xenopus laevis eggs promotes microtubule-based motility of both ER and Golgi membranes isolated from rat liver (Allan and Vale, 1991, 1994; Robertson and Allan, 2000). The movement can be analysed in real-time using video enhanced differential interference contrast microscopy (VE-DIC). The motility is MT-based, since cytochalasin D is added to prevent actin polymerisation. Virtually no movement occurred in the absence of cytosol (Supplementary Figure 1A and B).

When the RER fraction is combined with cytosol, membrane tubules extend along microtubules and fuse with each other to form an extensive two-dimensional network (Allan and Vale, 1994; Supplementary Figure 1C). The fusing tubules form three-way junctions and counting these junctions provides a simple indication of the extent membrane tubule movement (Allan, 1995). We used this feature to analyse the effects of the GST-fusion proteins on the motility of RER tubules.

RER membranes were first incubated with GST-KLC fusion protein, or GST as a control, then mixed with Xenopus egg cytosol and analysed as described in the Materials and methods. There was a significant reduction in RER membrane network formation if BTC was used, while incubation with DTC had no effect (Figure 2A; Supplementary Figure 1C), suggesting that the inhibition was KLC1 isoform-specific. In support of this conclusion, no inhibition was observed with 2TC, the KLC2-derived fusion protein (Figure 2B; Supplementary Figure 1C).

As a further test that RER movement is driven by kinesin-1, we incubated membranes with the C-terminal domain of rat uKHC fused to GST (uKHCct), since the C-terminal segment has previously been shown to inhibit kinesin-1-driven microtubule gliding and ATPase activity (Coy et al, 1999). While GST alone had no effect (data not shown), uKHCct caused a very strong dose-dependent inhibition (Figure 2C; Supplementary Figure 1C). Since over 90% of motility was inhibited with larger amounts of uKHCct, this indicates that kinesin-1 is likely to be the only motor protein that transports rat liver RER.

Motility of vesicles in a rat liver Golgi fraction depends on KLC1D

In contrast to the RER membranes, under the conditions used here the Golgi fraction membranes do not form tubules: instead many vesicles are seen moving along MT in the presence, but not the absence, of cytosol (Supplementary Figure 1B). We tested the effect of the KLC and uKHC fusion proteins on this vesicle movement. There was no reduction in vesicle movement if the membranes were preincubated with BTC as compared to GST-treated samples (Figure 3A; Supplementary Figure 1D). However, DTC reduced the number of moving vesicles by 40–50%, depending on the amount of protein used.

The motility of Golgi vesicles was not inhibited completely, which could indicate either that DTC is an inefficient inhibitor or that another motor protein drives a proportion of the motility. Therefore, we used uKHCct protein, which almost completely inhibited the motility of RER membranes (Figure 2B). However, uKHCct only reduced the number of moving vesicles by about 50% (Figure 3B, compare bars 1, 2, 7 and 8). Moreover, no further inhibition was seen when membranes were preincubated with a mixture of uKHCct and DTC (Figure 3B, bar 4), even when maximally inhibitory levels of uKHCct were used (Figure 3B, compare bars 8 and 9). An intermediate level of inhibition was observed when membranes were treated with uKHCct and BTC together (Figure 3B, bar 3), which is to be expected since the B isoform of KLC does not inhibit motility of the Golgi membranes, and acts to dilute the uKHCct. Finally, as was the case for RER movement, we observed that 2TC has no effect on the motility of Golgi membranes (Figure 3B, compare bars 1, 5 and 6).

Taken together, these data suggest that kinesin-1 is responsible for approximately 50% of vesicle motility in the Golgi fraction and that the carboxyl-terminus of KLC1D is most likely to be involved in the attachment of kinesin-1 to these vesicle membranes.

Cytosolic kinesin-1 is not required for membrane movement

In both motility assays, the presence of Xenopus egg cytosol greatly stimulates membrane movement (Supplementary Figure 1A and B). Since immunoblotting of Xenopus egg cytosol with anti-uKHC reveals plentiful soluble kinesin-1 (Figure 4A), it was possible that Xenopus kinesin-1 was being recruited to the membranes to drive the motility we observe, and that recombinant KLC and/or KHC prevented this recruitment. In Xenopus there are two uKHC bands, one of which migrates more slowly than the rat liver uKHC, which allowed us to test if Xenopus uKHC is recruited to rat liver membranes. As shown in Figure 4A, Xenopus egg uKHC remains in the supernatant and no recruitment is observed to either RER or Golgi fraction membranes.

As another approach to test the role of Xenopus kinesin-1, we immunodepleted kinesin-1 from Xenopus egg cytosol using the SUK4 antibody (Robertson and Allan, 2000), or MYC antibody as a control (Figure 4B). There was no significant difference in the motility of RER membranes or Golgi fraction vesicles in uKHC or MYC-depleted versus untreated cytosols (Figure 4C). Moreover, the kinesin-1 fusion proteins had similar inhibitory effects in SUK4-immunodepleted cytosol (Figure 4D and E) as they did when untreated cytosol was used (Figures 2 and 3).

One possible explanation for the effects of KLC1 and KHC fusion proteins might be that they block the ability of Xenopus egg cytosol to activate rat kinesin-1 on the membrane by sequestering an important cytosolic component. To test this possibility, cytosol was pre-incubated with uKHCct, BTC, DTC or GST coupled to glutathione beads to remove any interacting proteins, and then the depleted cytosol was collected and used in motility assays as before. This treatment had no effect on the motility of RER tubules or Golgi fraction vesicles (Supplementary Figure 1E and F).

From these data, we conclude that neither Golgi fraction vesicle nor RER membrane movement needs Xenopus kinesin-1, and that the inhibitory effects of our recombinant proteins likewise do not depend on the presence of Xenopus kinesin-1, or on their ability to interfere with the activation of rat kinesin-1 by cytosolic factors.

Kinesin-1 fusion proteins do not release the motor from the membrane

The most obvious explanation for the observed inhibitory effect of the KLC and uKHC fusion proteins is that they cause the release of rat kinesin-1 from the membrane. To test this possibility, RER membranes were incubated with recombinant protein and collected by centrifugation. Neither heavy nor light chains were removed from the membranes by incubation individually with BTC, DTC, uKHCct or GST (Figure 5A). This result could be explained if both KHCs and KLCs contribute to kinesin-1 attachment to membranes, in which case interference with one interaction might not be enough to cause release from the membrane. However, when both KLC1 and uKHC reagents were used together, kinesin-1 still remained membrane-associated (Figure 5A). Similarly, when this experiment was repeated with Golgi membranes, neither subunit was released from the membrane following incubation either with individual fusion proteins (Figure 5B, left panel) or with both uKHC and KLC reagents (Figure 5B, right panel).

Mechanism of uKHCct inhibition

Soluble kinesin-1 is usually in a folded, inactive conformation, where the C-terminal domain of KHC interacts with the heavy chain neck region, leading to inhibition of kinesin-1's motor and ATPase activity. This interaction is thought to be relieved by cargo binding or by incubating soluble kinesin-1 at high ionic strength, which generates an extended, active conformation (Hackney et al, 1992; Coy et al, 1999; Friedman and Vale, 1999; Stock et al, 1999; Hackney and Stock, 2000; Seiler et al, 2000). Since the combination of uKHCct and KLC1 fusion proteins did not release kinesin-1 from membranes, we considered the possibility that uKHCct inhibited membrane-bound kinesin-1 by binding directly to the KHC neck, so inhibiting its ATPase activity. To test this, we generated a fusion of the N-terminal half of uKHC (amino acids 1–565) with GFP and expressed it in COS7 cells. Cell lysates were incubated with purified uKHCct bound to glutathione beads, or with beads coated with BTC, DTC or GST (proteins that should not be able to interact with the N-terminal domain of uKHC). The N-terminal of uKHC was seen to interact with uKHCct, but not with the other proteins (Figure 6A), suggesting that uKHCct might indeed inhibit motility by binding directly to the KHC neck.

As a further test of this hypothesis, we examined the effect of uKHCct on the conformation of kinesin-1 using sucrose density gradient centrifugation. We predicted that exogenous uKHCct would compete with the endogenous KHCct for binding to the neck of native kinesin-1, so generating a pool of motor in the extended conformation. When kinesin-1 was solubilised from membranes preincubated with GST, BTC or DTC, it migrated in the mid-region of a 5–20% sucrose density gradient (fractions 6/7–10; Figure 6B). In contrast, kinesin solubilised from membranes that were pretreated with uKHCct had a much broader distribution across the gradient, with a proportion being found in the less dense fractions 11–13. This suggests that the incubation with uKHCct had indeed converted some folded kinesin-1 to the extended state.

We then tested the effects of uKHCct and the other fusion proteins on kinesin-1's ability to bind to microtubules in the presence of ATP. RER membranes were preincubated with fusion proteins, then solubilised and incubated in the presence of 5 mM ATP, with or without microtubules (Figure 7A). As a control, membrane lysates were incubated with 400 M AMP-PNP to induce rigor binding of kinesin-1 to microtubules. In parallel, the same experiment was performed with Xenopus egg cytosol (Figure 7B). As expected, both rat RER and Xenopus cytosolic kinesin-1 pelleted efficiently with microtubules in the presence of AMP-PNP. Kinesin-1 that had been preincubated with BTC, DTC or GST did not bind to microtubules in the presence of 5 mM ATP. However, preincubation with uKHCct led to a proportion of kinesin-1 that was sufficiently tightly bound to microtubules in the presence of 5 mM ATP to appear in the microtubule pellet, indicating that the presence of uKHCct slows or inhibits the release of kinesin-1 from microtubules in the presence of ATP. The result is specific to kinesin-1, since kinesin-2 was not affected by the presence of uKHCct (Figure 7B, middle panel). Preincubation of rat or Xenopus kinesin-1 with uKHCct did not interfere with the ability of AMP-PNP to promote the binding of kinesin-1 to microtubules (data not shown), further confirming that this recombinant protein does not prevent kinesin-1-microtubule interactions.

Identification of kinesin-1 vesicular cargoes

Although the RER is easy to identify by VE-DIC on morphological grounds, the vesicles present in the Golgi membrane fraction are heterogeneous. Since VLDL particles are easily recognised by EM (Allan and Vale, 1994), we used immuno-EM to determine whether VLDL-containing secretory vesicles possessed kinesin-1. KLC and uKHC antibodies bound to vesicles of 70–200 nm in diameter, which were often aggregated together, but did not contain VLDL particles (data not shown). This was confirmed by labelling fixed and permeabilised Golgi membranes (Allan and Vale, 1994) with antibodies to kinesin, apolipoprotein B and albumin (data not shown).

To identify the kinesin-1 positive vesicles, we used an immunofluorescence approach that used unfixed membranes (Bananis et al, 2000, 2004). Membranes were labelled with H2, a monoclonal anti-KHC antibody, in combination with a polyclonal KHC antibody (HD) or a polyclonal KLC antibody. Although HD, which was raised against Drosophila KHC, also recognises other kinesin superfamily members, it reacts most strongly with uKHC (Rodionov et al, 1993). Both polyclonal antibodies showed a similar staining pattern and overlapped with the H2 antibody almost completely (Supplementary Figure 2A). In addition, H2 and HD both gave very similar labelling patterns in normal rat kidney (NRK) cells (Supplementary Figure 3A).

Many kinesin-1-positive vesicles contained p58/ERGIC58 (Figure 8A), a marker for the intermediate compartment/ERGIC that is involved in the transport of material in both directions between the ER and Golgi apparatus. We found occasional overlap on vesicles between kinesin-1 and p115 (Figure 8B), a protein that targets incoming transport vesicles to the Golgi apparatus (Waters et al, 1992). Vesicles within larger clumps of membrane were often labelled with both HD and antibodies to formiminotransferase cyclodeaminase (FTCD) (Figure 8C), a Golgi-localised protein that cycles through the ERGIC (Bashour and Bloom, 1998; Gao et al, 1998; Hennig et al, 1998). There was no colocalisation, however, between HD or H2 and other ERGIC markers such as the KDEL receptor (Supplementary Figure 2B), or COP (data not shown), or with the COPII component Sec23p (data not shown). In addition, kinesin-1 did not overlap with GM130, a cis-Golgi structural protein that cycles between the late ERGIC and the Golgi apparatus (Marra et al, 2001) (data not shown), nor with Golgin-84 (Supplementary Figure 2C).

We also tested whether any of the kinesin-1-positive vesicles were derived from the trans-Golgi or endocytic compartments. While there was no colocalisation between kinesin-1 and TGN38, some vesicles possessed both -adaptin and kinesin-1 (Figure 8D). There was no overlap between kinesin-1 and EEA1 or transferrin receptor (data not shown), whereas a few vesicles were positive for mannose-6-phosphate receptor (M6PR) and kinesin-1 (Figure 8E).

As a control, we labelled NRK cells with the above antibodies. Kinesin-1 antibodies labelled the perinuclear region, and showed general overlap with antibodies to p58/ERGIC58 or FTCD (Supplementary Figures 3B, 4A and B), -adaptin (Supplementary Figure 3C) and GM130 (Supplementary Figure 4C). This distribution of kinesin-1 in rat cells is consistent with previous work (Marks et al, 1994).

Discussion

Many different cellular processes are orchestrated by the numerous members of the kinesin superfamily. However, even a single motor type, such as kinesin-1, may transport multiple distinct cargoes, but how this is achieved is poorly understood. Here, we provide functional evidence that KLC isoforms determine the specificity of kinesin-1 association with different cargoes.

In vitro assays for studying kinesin-1 function

We have used two in vitro motility assays for these studies. Rat liver RER movement in the presence of Xenopus egg cytosol has been described previously (Allan and Vale, 1994), and we now demonstrate that this motility is driven by kinesin-1. This is in keeping with in vivo and in vitro observations of ER dynamics or distribution in different systems (Feiguin et al, 1994; Lane and Allan, 1999; Bannai et al, 2004).

The second motility assay uses a rat liver Golgi stack fraction combined with Xenopus egg cytosol, and under the conditions used here, this generates active vesicle movement. We find that both the DTC and uKHCct fusion proteins inhibit 50% of this motility, identifying kinesin-1 as the motor for a substantial proportion of moving vesicles. The remaining vesicle translocation is driven by kinesin-2 (Wozniak and Allan, unpublished data), with a minor contribution from as yet unidentified kinesin family members, which might include KIF1C (Dorner et al, 1998). Interestingly, the kinesin family member that drives the extension of membrane tubules in the Golgi fraction in the presence of brefeldin A is clearly distinct from the vesicle motors, since not only is tubule extension much slower than vesicle movement, it is also inhibited by the H1 monoclonal antibody (Robertson and Allan, 2000), whereas vesicle movement is not (data not shown).

The kinesin-1-positive vesicles in the Golgi fraction often contain ERGIC58, suggesting that they are involved in the Golgi–ERGIC–ER recycling pathway. The occasional presence of p115 would be consistent with this hypothesis. We saw no obvious overlap with other ERGIC components such as COPI or the KDEL receptor, however, suggesting that the vesicles were not fragments of the ERGIC: instead, they might be involved in traffic between the Golgi apparatus and late ERGIC, or between the ERGIC and the ER. This is in keeping with previous studies on the complexity and dynamics of the ERGIC (Marra et al, 2001). Larger vesicles were sometimes labelled with antibodies to FTCD, a Golgi protein that also cycles between the ERGIC and Golgi apparatus: while it partially overlaps with ERGIC58, p115 and COPI, its localisation is clearly distinct (Gao et al, 1998; Hennig et al, 1998).

Surprisingly, kinesin-1 was not present on vesicles containing secretory markers, even though kinesin-1 in other systems clearly transports vesicles between the TGN and plasma membrane (Wozniak et al, 2004). A small percentage of vesicles did possess both kinesin-1 and -adaptin or M6PR, suggesting that kinesin-1 may be involved in a subset of post-TGN traffic in rat liver. Taken together with the localisation of kinesin-1 to the Golgi region in vivo (this work; Marks et al, 1994; Gyoeva et al, 2000), these data support a role for kinesin-1 in traffic away from the Golgi apparatus back towards the ER, and for selected onwards traffic.

Function of specific KLC1 isoforms

It has been proposed, based on the use of antibodies that recognise limited sets of KLC1 isoforms (Khodjakov et al, 1998; Gyoeva et al, 2000) and on the molecular weight of KLCs present on particular cargo (Liao and Gundersen, 1998), that specific KLC isoforms target kinesin-1 to distinct structures. Here, we provide functional evidence in support of this model. We have made use of GST fusion proteins that contain the TPR domain and C-terminal regions of KLC1B, D and KLC2, but which lack the N-terminus and KHC-binding heptad repeat domain. Strikingly, we found that only the KLC1B fusion protein inhibited RER movement, while KLC1D inhibited Golgi vesicle movement.

An antibody that recognises both KLC1D and E has previously been shown to label the Golgi apparatus in cultured cells (Gyoeva et al, 2000). Since KLC1D is identical to KLC1E except for the inclusion of an additional nine amino acids (Figure 1B), it is possible that both isoforms are present in the rat liver Golgi fraction, and that the KLC1D fusion protein inhibits both KLC1D and E function. Indeed, immunoblotting of the Golgi membranes revealed two KLC bands of similar molecular weight (Figure 1A). An interesting possibility is that these two isoforms correspond to two distinct kinesin-1-positive vesicle populations: those containing p58 and those positive for markers such as FTCD, M6PR or -adaptin.

Our functional data suggest that KLC1B is important for RER motility. In contrast, an antibody to this isoform labels mitochondria, rather than ER (Khodjakov et al, 1998). However, since KLC1B and C differ only by the insertion of nine amino acids in the KLC1C c-terminus (Figure 1A), the anti-KLC1B antibody will also recognise KLC1C. It is possible that one of these splicing variants is specific for mitochondria, while the other is restricted to ER membranes, but this hypothesis remains to be tested.

Since the sole difference between the KLC1B and D constructs used here lies in the C-terminus, this demonstrates the importance of this region in cargo-specific functions of KLCs. It also suggests that the TPR domain, which interacts with many potential cargo proteins, has a more generic function in cargo binding. The only protein identified so far that interacts with the C-terminal region of a KLC is 14-3-3, which has been reported to bind to phosphorylated KLC2 C-terminal (Ichimura et al, 2002). We are currently searching for KLC1B and D-interacting proteins in the RER and Golgi fractions.

A key question is how do the KLC1 fusion proteins lead to the inhibition of membrane-associated kinesin-1? If KLCs are needed for kinesin-1 to bind to cargo, then a simple possibility is that the appropriate fusion protein would compete for binding to a specific receptor on the membrane, releasing kinesin-1, and inhibiting movement. This would parallel the effects of the KLC_ALL antibody, which binds to the TPR domain and leads to loss of motor binding to axonal vesicles (Stenoien and Brady, 1997). This was not the case for KLC1 fusion proteins, however, as kinesin-1 remained membrane-bound although membrane motility was inactivated. This result could be explained if both KHCs and KLCs participate in cargo binding. Indeed, KHC dimers may be able to bind to membranous cargo in the absence of KLCs (Skoufias et al, 1994; Palacios and St Johnston, 2002), and fungal kinesin-1 lacks KLCs altogether (Seiler et al, 2000). Moreover, two potential kinesin-1 receptors have been identified on the ER—kinectin (Toyoshima et al, 1992) and the related p180 ribosome receptor (Diefenbach et al, 2004)—both of which interact with the KHC C-terminal domain (Ong et al, 2000; Diefenbach et al, 2004). If both heavy and light chains bind simultaneously to different cargo molecules, this could explain the tight membrane association that has been reported for kinesin-1 (Muresan et al, 1996; Tsai et al, 2000), and would be consistent with the undetectable level of kinesin-1 turnover on membranes that we have observed. Specific regulatory mechanisms would then be needed to control kinesin-1 cargo association, with one example being the phosphorylation of the C-terminal domain of KLC2 by glycogen synthase kinase 3, which enables Hsc70 to remove kinesin-1 from the membrane in neurons (Morfini et al, 2002).

Regulation of kinesin-1 activity

If release of kinesin-1 from the membrane is not the mechanism of inhibition, then what is? A likely alternative involves the ability of kinesin-1 to fold up so that the C-terminal domain of KHC interacts with the neck region, so inhibiting ATPase activity (Coy et al, 1999; Friedman and Vale, 1999; Stock et al, 1999; Hackney and Stock, 2000) and the ability to move along microtubules (Coy et al, 1999). Although KHC dimers alone can fold in this way, the suppression of ATPase activity is even greater in the presence of KLCs (Coy et al, 1999; Friedman and Vale, 1999; Stock et al, 1999; Hackney and Stock, 2000), and it has been suggested that KLCs are vital to allow correct in vivo regulation of this folding process (Verhey et al, 1998). We propose that the appropriate KLC reagent (KLC1B for the RER, KLC1D for vesicles in the Golgi fraction) competes with endogenous KLC1 for binding to a cargo-specific receptor (protein X in Figure 9, I II). Although kinesin-1 still remains membrane-bound by virtue of the uKHC-protein Y interaction, the release of KLC1 from protein X leads to a conformational change that exposes the critical region in the uKHCct, allowing kinesin-1 to fold up and therefore inhibiting motor activity (Figure 9, II).

When both heavy and light chain fusion proteins are combined, kinesin-1 is still not released from the membrane (Figure 5). This suggests that uKHCct is not able to compete for binding to protein Y (Figure 9). Instead, based on our results (Figures 6 and 7) and those of others (Coy et al, 1999; Friedman and Vale, 1999; Stock et al, 1999; Hackney and Stock, 2000), we propose that uKHCct inhibits kinesin-1-driven membrane movement by binding to the neck region of the membrane-bound motor (Figure 9, III), thereby directly inhibiting its ability to hydrolyse ATP and move along microtubules. Interestingly, we find that this interaction leads to some kinesin-1 pelleting with microtubules in the presence of ATP (Figure 7), which fits with the observation that a short C-terminal fragment does not inhibit the binding of Drosophila kinesin-1 to microtubules, but instead actually increased the time spent associated with the microtubule three-fold (Coy et al, 1999). One possibility is that the interaction of exogenous C-terminus locks kinesin-1 in a different kinetic state compared to the ADP-bound form that binds only weakly to microtubules, and that is seen when the endogenous tail interacts with the neck (Hackney and Stock, 2000). Alternatively, the association between native kinesin-1 and microtubules may be promoted by the ability of the C-terminal fragment to bind to microtubules directly (Coy et al, 1999; Hackney and Stock, 2000; Figure 7B). In full-length KHC, this microtubule binding site is thought to remain hidden in the presence of light chains (Navone et al, 1992; Verhey et al, 1998), although it may be active in fungal kinesin-1, which lacks KLCs (Straube et al, 2006). We can rule out that uKHCct is inhibiting kinesin-1 activity simply by coating the entire microtubule surface, since it had no effect on the 50% of vesicle movement driven by kinesin-2.

The model we propose for the inactivation of kinesin-1 while it is membrane-bound is based on our observations following the addition of exogenous protein. It has previously been suggested that kinesin-1 is folded in the absence of cargo (i.e. when kinesin-1 is cytosolic), and that it becomes unfolded and activated following binding to its correct target (Verhey et al, 1998; Coy et al, 1999; Friedman and Vale, 1999; Stock et al, 1999; Hackney and Stock, 2000). However, since our data show that it is possible to inactivate membrane-associated kinesin-1, it is tempting to speculate that reversible folding may occur while kinesin-1 is attached to cargo, and that this may constitute a mechanism of regulation. Indeed, it is known that ER-associated kinesin-1 is inactive in Xenopus eggs and early embryos, but becomes activated when the membranes are incubated in cytosol prepared from somatic cells (Lane and Allan, 1999). What might trigger a regulated conformational change is not clear, but phosphorylation of the motor or an alteration in the conformation of protein X would be two possibilities. Indeed, if protein X were a transmembrane protein, this might provide a means of coordinating motor activity with the concentration of cargo in the vesicle or tubule.

In summary, we propose that KLCs must be correctly associated with cargo-specific molecules in order for kinesin-1 to be in the unfolded, active state (Figure 9). Given the KLC isoform-specific effects we observe, the variable C-terminal domain of KLCs is very likely to be responsible for targeting kinesin-1 to particular cargoes. The existence of at least 19 splice forms of KLC1 (McCart et al, 2003) suggests that there are easily enough isoforms to account for the wide range of structures that this motor transports within the cell.

Materials and methods

Additional Materials and methods (antibodies and reagents, cloning and protein purification, immunofluorescence) are provided as Supplementary data.

Preparation and manipulation of Xenopus extracts

Interphase-arrested Xenopus egg extracts were prepared from metaphase-arrested CSF extracts (Murray, 1991), or by preparation of extracts from laid eggs in the absence of EGTA (Newmeyer and Wilson, 1991). Cytosol fractions for motility assays were generated by centrifugation as described (Allan and Vale, 1991).

To immunodeplete kinesin-1 from Xenopus egg cytosol, protein G magnetic beads (Dynal) were washed with A/S buffer (100 mM K-acetate, 3 mM Mg-acetate, 5 mM EGTA, 10 mM HEPES, 150 mM sucrose, pH 7.4) and incubated with SUK4 or MYC antibodies for 1 h at 4°C. The beads were washed with A/S buffer then 150 l Xenopus cytosol was added and incubated for 1 h at 4°C. To test the effect of GST-fusion proteins on Xenopus cytosol, 30 g recombinant protein was bound to glutathione beads for 1 h. Beads were washed twice with A/S buffer and incubated for 30 min at 4°C with 100 l of interphase Xenopus cytosol. In both cases, cytosols were used immediately for motility assays.

Preparation of rat liver membrane fractions and motility assays

RER and stacked Golgi fractions were prepared from rat livers as described (Allan and Vale, 1991, 1994). Membranes were incubated for 10–13 min on ice with the indicated recombinant proteins before assaying motility. A measure of 0.5–0.8 l RER (4.5–7.2 g) or 1 l Golgi fraction membranes (2.7 g) were added to 9 l of Xenopus cytosol and incubated in a microscope flow cell in a humid chamber (Allan and Vale, 1991) that had been pre-coated with microtubules. Taxol-stabilised porcine brain tubulin microtubules (0.2 mg/ml in BRB80: 80 mM K-PIPES, 2 mM MgCl₂, 1 mM EGTA pH 6.8) were incubated in the flow cell for 20–30 min and washed three times with A/S buffer with energy mix (7.5 mM creatine phosphate, 1 mM MgATP, 0.1 mM EGTA) before flowing in the cytosol-membrane mixture. For Supplementary Figure 1A and B, cytosol was replaced by A/S buffer with energy mix. Golgi fraction motility assays were incubated for 10 min and then 20 random fields were observed for 30 s each by VE-DIC microscopy (Lane and Allan, 1999). The average number of moving vesicles per field was determined. RER motility assays were incubated for 60 min in the flow cell to allow ER network formation and then 30 random fields were recorded, and the number of three-way junctions in the ER network was counted (Allan, 1995).

Biochemical analysis of membranes

To test if Xenopus egg kinesin-1 was recruited onto rat liver membranes, 10 l of membranes were incubated with 50 l of Xenopus egg cytosol for 10–20 min at RT in the presence of nocodazole, and then pelleted through a 0.8 M sucrose cushion in a Beckman TLA-100 rotor at 55 000 r.p.m., 4°C for 30 min. To test the influence of fusion proteins on kinesin-1 membrane association, 10 l of membranes was incubated with 20 l of fusion protein for 30 min on ice and centrifuged as above. The pellets were washed with A/S buffer analysed by immunoblotting.

To test whether GST-fusion proteins affect kinesin-1 conformation, 80 g Golgi membranes were incubated with 20 g recombinant proteins for 10 min on ice, then lysed for 10 min on ice by the addition of an equal volume of 2% Triton TX-100, 400 mM NaCl, 100 mM ascorbic acid and protease inhibitors in acetate buffer. After centrifugation at full speed in a microfuge for 10 min at 4°C, the lysates were loaded onto 650 l 5–20% sucrose gradients (in acetate buffer), centrifuged in a Beckman SW55 rotor with adapters at 4°C, 46 600 r.p.m. for 5 h. Fractions were collected from the bottom of the tubes and analysed by immunoblotting.

Microtubule binding assays

RER membranes (400 g) or interphase Xenopus cytosol (200 g) were incubated for 15 min at RT with 200 l of A/S buffer or A/S buffer containing 0.3 mg/ml GST-fusion proteins. Samples were lysed for 10 min on ice after the addition of an equal volume of 2% Triton TX-100, 400 mM NaCl plus protease inhibitors and centrifuged for 10 min at 4°C, full speed in a microfuge. Cleared lysates were supplemented with either 5 mM ATP, 20 M taxol, 1 mM DTT, 1 g/ml cytochalasin D (lysates preincubated with recombinant proteins) or 10 U/ml hexokinase, 20 M glucose, 20 M taxol, 400 M AMP-PNP, 1 mM DTT, 1 g/ml cytochalasin D (lysates preincubated with buffer) and left at RT for 5 min. Microtubules (50 g) were added and incubated for 30 min at 30°C. Samples were centrifuged for 30 min at 32 000 r.p.m., 4°C in Beckman TLS55 rotor through a 40% (w/v) sucrose cushion in BRB80 buffer supplemented with 1 mM DTT, 4 M taxol, 1 g/ml cytochalasin D and protease inhibitors. Pellets were washed in BRB80 buffer with 1 mM DTT, 4 M taxol, 1 g/ml cytochalasin D and protease inhibitors and spun as above for 20 min. Pellets were resuspended in 15 l BRB80 and analysed by immunoblotting.

GST pull-downs from COS7 cell lysates

KIF5B_NT-GFP (Supplementary Methods) was expressed in COS7 cells after transfection using JetPei (Polyplus-transfection, France). Cells were washed once with PBS, lysed in lysis buffer (10% glycerol, 1% Triton TX-100, 1.5 mM MgCl₂, 50 mM HEPES, pH 7.5, 150 mM NaCl, 1 mM EGTA, 50 mM ascorbic acid) and spun for 10 min in a microfuge at full speed, 4°C. Recombinant kinesin-1 GST-fusion proteins (5 g each) were immobilised on glutathione beads for 1 h, then washed twice with acetate buffer, 1% Triton TX-100, 150 mM NaCl. Beads were then incubated with 100 g of COS7/KIF5B_NT-GFP lysate in a total volume 100 l for 2 h at 4°C. Beads were washed three times as above and mixed with SDS–PAGE loading buffer.

Supplementary data

Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).

Acknowledgements

We are grateful to all our colleagues who generously provided reagents. In particular, we would like to thank Dr Fatima Gyoeva for providing KLC1 constructs. We are grateful to Dr Aleksandr Mironov Jr for help with electron microscopy. This work was funded by the Biotechnology and Biological Sciences Research Council (Grant C16691) and The Wellcome Trust (Grant 078825/Z/05/Z).

References

Adio S, Reth J, Bathe F, Woehlke G (2006) Regulation mechanisms of kinesin-1. J Muscle Res Cell Motil 27: 153–160 | Article | PubMed | ChemPort |

Allan V (1995) Protein phosphatase 1 regulates the cytoplasmic dynein-driven formation of endoplasmic reticulum networks in vitro. J Cell Biol 128: 879–891 | Article | PubMed | ISI | ChemPort |

Allan V, Vale R (1994) Movement of membrane tubules along microtubules in vitro: evidence for specialised sites of motor attachment. J Cell Sci 107: 1885–1897 | PubMed | ISI | ChemPort |

Allan VJ, Vale RD (1991) Cell cycle control of microtubule-based membrane transport and tubule formation in vitro. J Cell Biol 113: 347–359 | Article | PubMed | ISI | ChemPort |

Bananis E, Murray JW, Stockert RJ, Satir P, Wolkoff AW (2000) Microtubule and motor-dependent endocytic vesicle sorting in vitro. J Cell Biol 151: 179–186 | Article | PubMed | ChemPort |

Bananis E, Nath S, Gordon K, Satir P, Stockert RJ, Murray JW, Wolkoff AW (2004) Microtubule-dependent movement of late endocytic vesicles in vitro: requirements for Dynein and Kinesin. Mol Biol Cell 15: 3688–3697 | Article | PubMed | ChemPort |

Bannai H, Inoue T, Nakayama T, Hattori M, Mikoshiba K (2004) Kinesin dependent, rapid, bi-directional transport of ER sub-compartment in dendrites of hippocampal neurons. J Cell Sci 117: 163–175 | Article | PubMed | ChemPort |

Bashour A-M, Bloom G (1998) 58K, a microtubule-binding Golgi protein, is a formiminotransferase cyclodeaminase. J Biol Chem 273: 19612–19617 | Article | PubMed | ISI | ChemPort |

Brady ST, Pfister KK, Bloom GS (1990) A monoclonal antibody against kinesin inhibits both anterograde and retrograde fast axonal transport in squid axoplasm. Proc Natl Acad Sci USA 87: 1061–1065 | Article | PubMed | ChemPort |

Coy DL, Hancock WO, Wagenbach M, Howard J (1999) Kinesin's tail domain is an inhibitory regulator of the motor domain. Nat Cell Biol 1: 288–292 | Article | PubMed | ISI | ChemPort |

Diefenbach RJ, Diefenbach E, Douglas MW, Cunningham AL (2002) The heavy chain of conventional kinesin interacts with the SNARE proteins SNAP25 and SNAP23. Biochemistry 41: 14906–14915 | Article | PubMed | ISI | ChemPort |

Diefenbach RJ, Diefenbach E, Douglas MW, Cunningham AL (2004) The ribosome receptor, p180, interacts with kinesin heavy chain, KIF5B. Biochem Biophys Res Commun 319: 987–992 | Article | PubMed | ChemPort |

Dorner C, Ciossek T, Muller S, Moller PH, Ullrich A, Lammers R (1998) Characterization of KIF1C, a new kinesin-like protein involved in vesicle transport from the Golgi apparatus to the endoplasmic reticulum. J Biol Chem 273: 20267–20275 | Article | PubMed | ISI | ChemPort |

Feiguin F, Ferreira A, Kosik KS, Caceres A (1994) Kinesin-mediated organelle translocation revealed by specific cellular manipulations. J Cell Biol 127: 1021–1039 | Article | PubMed | ISI | ChemPort |

Friedman DS, Vale RD (1999) Single-molecule analysis of kinesin motility reveals regulation by the cargo-binding tail domain. Nat Cell Biol 1: 293–297 | Article | PubMed | ISI | ChemPort |

Gao Y-S, Alvarez C, Nelson D, Sztul E (1998) Molecular cloning, characterization and dynamics of rat formiminotransferase cyclodeaminase, a Golgi-associated 58-kDa protein. J Biol Chem 273: 33825–33834 | Article | PubMed | ISI | ChemPort |

Gindhart Jr JG, Desai CJ, Beushausen S, Zinn K, Goldstein LS (1998) Kinesin light chains are essential for axonal transport in Drosophila. J Cell Biol 141: 443–454 | Article | PubMed | ISI | ChemPort |

Glater E, Megeath L, Stowers R, Schwarz T (2006) Axonal transport of mitochondria requires milton to recruit kinesin heavy chain and is light chain independent. J Cell Biol 173: 545–557 | Article | PubMed | ChemPort |

Gyoeva FK, Bybikova EM, Minin AA (2000) An isoform of kinesin light chain specific for the Golgi complex. J Cell Sci 113: 2047–2054 | PubMed | ISI | ChemPort |

Gyoeva FK, Sarkisov DV, Khodjakov AL, Minin AA (2004) The tetrameric molecule of conventional Kinesin contains identical light chains. Biochemistry 43: 13525–13531 | PubMed | ChemPort |

Hackney DD, Levitt JD, Suhan J (1992) Kinesin undergoes a 9 S to 6 S conformational transition. J Biol Chem 267: 8696–8701 | PubMed | ISI | ChemPort |

Hackney DD, Stock MF (2000) Kinesin's IAK tail domain inhibits initial microtubule-stimulated ADP release. Nat Cell Biol 2: 257–260 | Article | PubMed | ISI | ChemPort |

Hennig D, Scales S, Moreau A, Murley L, De Mey J, Kreis T (1998) A fromiminotransferase cyclodeaminase isoform is localized to the Golgi complex and can mediate interaction of trans-Golgi network-derived vesicles with microtubules. J Biol Chem 273: 19602–19611 | Article | PubMed | ChemPort |

Hirokawa N, Sato-Yoshitake R, Kobayashi N, Pfister KK, Bloom GS, Brady ST (1991) Kinesin associates with anterogradely transported membranous organelles in vivo. J Cell Biol 114: 295–302 | Article | PubMed | ISI | ChemPort |

Huang JD, Brady ST, Richards BW, Stenolen D, Resau JH, Copeland NG, Jenkins NA (1999) Direct interaction of microtubule- and actin-based transport motors. Nature 397: 267–270 | Article | PubMed | ISI | ChemPort |

Ichimura T, Wakamiya-Tsuruta A, Itagaki C, Taoka M, Hayano T, Natsume T, Isobe T (2002) Phosphorylation-dependent interaction of kinesin light chain 2 and the 14-3-3 protein. Biochemistry 41: 5566–5572 | Article | PubMed | ISI | ChemPort |

Junco A, Bhullar B, Tarnasky HA, van der Hoorn FA (2001) Kinesin light-chain KLC3 expression in testis is restricted to spermatids. Biol Reprod 64: 1320–1330 | Article | PubMed | ChemPort |

Khodjakov A, Lizunova EM, Minin AA, Koonce MP, Gyoeva FK (1998) A specific light chain of kinesin associates with mitochondria in cultured cells. Mol Biol Cell 9: 333–343 | PubMed | ISI | ChemPort |

Lane JD, Allan VJ (1999) Microtubule-based endoplasmic reticulum motility in Xenopus laevis: activation of membrane-associated kinesin during development. Mol Biol Cell 10: 1909–1922 | PubMed | ISI | ChemPort |

Leelavathi DE, Estes LW, Feingold DS, Lombardi B (1970) Isolation of a Golgi-rich fraction from rat liver. Biochim Biophys Acta 211: 124–138 | ChemPort |

Liao G, Gundersen GG (1998) Kinesin is a candidate for cross-bridging microtubules and intermediate filaments. Selective binding of kinesin to detyrosinated tubulin and vimentin. J Biol Chem 273: 9797–9803 | Article | PubMed | ISI | ChemPort |

Marks DL, Larkin JM, McNiven MA (1994) Association of kinesin with the Golgi apparatus in rat hepatocytes. J Cell Sci 107: 2417–2426 | PubMed | ISI | ChemPort |

Marra P, Maffucci T, Daniele T, Di Tullio G, Ikehara Y, Chan E, Luini A, Beznoussenko G, Mironov A, De Matteis M (2001) The GM130 and GRASP65 Golgi proteins cycle through and define a subdomain of the intermediate compartment. Nat Cell Biol 3: 1101–1113 | Article | PubMed | ISI |