A new role for kinesin-directed transport of Bik1p (CLIP-170) in Saccharomyces cerevisiae

Microtubules (MTs) are fibrous structures in the cytoplasm of eukaryotic cells and play a vital role in cell organization, motility and division. MTs are intrinsically polar polymers with a fast-growing plus end and a slow-growing minus-end (Carvalho et al., 2003; Desai and Mitchison, 1997). In fungi and animals, the minus-ends of MTs are usually at or adjacent to the MT organizing centre and the plus ends are oriented peripherally. Members of a group of proteins called plus-end-tracking proteins or +TIPs specifically associate with the plus ends of MTs (Galjart and Perez, 2003). In live-cell experiments, green fluorescent protein (GFP)-labelled +TIPs appear as comet- or dot-like structures that remain on the MT plus ends (Perez et al., 1999). +TIPs form structural links between MT plus ends and polarized membrane sites or kinetochores (Akhmanova et al., 2001; Carvalho et al., 2003; Coquelle et al., 2002; Dujardin and Vallee, 2002; Fukata et al., 2002; Howard and Hyman, 2003; Lin et al., 2001; Pierre et al., 1992; Tai et al., 2002), and are important for the regulation of MT dynamics (Han et al., 2001; Maiato et al., 2003; Rogers et al., 2002; Tirnauer et al., 1999).

The first discovered +TIP, CLIP-170, contains specific MT-binding domains (CAP-Gly domains) that are shared by related proteins such as CLIP-115 and P150^Glued in mammalian cells, or its orthologues Bik1p and Tip1p in budding and fission yeasts. The CAP-Gly-containing proteins are crucial for astral MT interactions with cortical dynein-dynactin patches during mitosis and for the delivery of motors such as dynein to their site of action (Coquelle et al., 2002; Lansbergen et al., 2004; Niccoli et al., 2004; Tai et al., 2002; Vaughan et al., 2002; Xiang et al., 2000). In budding yeast, dynein is delivered to the cortex on the plus ends of polymerizing MTs. Furthermore, dynein recruitment at MT tips is thought to involve dynein interactions with Bik1p at MT plus ends (Sheeman et al., 2003).

How CLIP-170 and related proteins `recognize' MT plus ends is not completely clear. In mammalian cells, there is evidence that CLIP-170 accumulates at MT ends both by co-polymerizing with tubulin (treadmilling mechanism) and by associating with another +TIP, EB1 (hitch-hiking mechanism) (Galjart, 2005). An additional mechanism has been proposed in yeast in which Bik1p and Tip1p are thought to be principally localized at MT plus ends by kinesin motor proteins (Kip2p in budding yeast, Tea2p in fission yeast) (Busch et al., 2004; Carvalho et al., 2004).

We have previously demonstrated that, in budding yeast, removal of the C-terminal phenylalanine residue of α-tubulin dramatically impairs the recruitment of Bik1p at MT plus ends (Badin-Larcon et al., 2004; Erck et al., 2005; Peris et al., 2006). We have recently reported related observations in mammalian cells, in which the association of CLIP-170 and other CAP-Gly +TIPs with MT plus ends is severely impaired in cells expressing mutant tubulin lacking the C-terminal amino acid (Glu tubulin; lacking Tyr) (Erck et al., 2005; Peris et al., 2006). Several studies have provided a structural basis for our observations by demonstrating a crucial requirement of the C-terminal aromatic residue of α-tubulin for tubulin interaction with CAP-Gly domains (Honnappa et al., 2006; Mishima et al., 2007; Weisbrich et al., 2007). In budding yeast models in which Bik1p is principally localized as a Kip2p cargo, direct interaction of Bik1p with tubulin is not required. There is an apparent conflict between such models and the massive mis-localization of Bik1p caused by the Glu mutation in budding yeast.

Here, we provide evidence that, whereas Kip2p mediates transport of Bik1p towards MT plus ends, it is not sufficient to sustain subsequent Bik1p plus-end tracking, which apparently requires interaction of Bik1p with tubulin. Additionally, our data indicate that dynein is transported at MT plus ends in a Kip2p-Bik1p-dependent manner. Apparently, dynein can subsequently track MT plus ends even when Bik1p is severely depleted at those ends. Thus, although Kip2p might not be sufficient to mediate Bik1p plus-end tracking, it seems to be central in a new pathway for dynein localization at plus ends, which can support apparently normal dynein function even when the interaction of Bik1p with tubulin is disrupted.

We have previously shown that Bik1p accumulation at MT plus ends is decreased about threefold in yeast strains expressing phenylalanine-deleted Glu tubulin (tub1-Glu) compared with strains expressing wild-type Phe tubulin (TUB1) (Badin-Larcon et al., 2004). However, these results were observed using overexpressed Bik1p. In such conditions, Kip2p might be saturated, and we could have looked principally at the localization of excess free Bik1p. To examine the localization of Bik1p expressed at physiological levels, we imaged a functional fusion between Bik1p and a 3GFP cassette (Bik1p-3GFP) in cells in which the spindle pole body (SPB) was also labelled with Spc42p-RedStar. Fluorescence analysis in the TUB1 strain showed a Bik1p-3GFP signal as dots at MT plus ends (Fig. 1A), as previously described (Lin et al., 2001). Bik1p dots were still visible in the tub1-Glu strain, but quantitative analysis indicated a threefold decrease of the Bik1p signal at plus ends compared with the TUB1 strain, as previously observed with plasmid-overexpressed Bik1p (Fig. 1B). Thus, the Glu mutation affects Bik1p localization, even when expressed at physiological levels.

We then searched for factors responsible for persistent localization of some Bik1p at MT plus ends in the tub1-Glu strain. We tested the role of Bim1p. Bim1p orthologues can localize CAP-Gly proteins at plus ends in many systems (Lansbergen and Akhmanova, 2006). In previous studies, budding yeast looked to be an exception, because Bik1p localization was insensitive to BIM1 deletion. However, we thought that the situation might be different in tub1-Glu strains. We examined the consequences of BIM1 deletion on Bik1p-3GFP localization in our strains. Strains TUB1 bim1Δ and tub1-Glu bim1Δ were both viable and grew normally (data not shown). MT length was somewhat decreased in both TUB1 bim1Δ and tub1-Glu bim1Δ strains (Table 1). As expected, BIM1 deletion had no detectable effect on Bik1p-3GFP localization in the TUB1 bim1Δ strain (Fig. 1A and for quantitative analysis Fig. 1B). By contrast, in the tub1-Glu bim1Δ strain, Bik1p-3GFP signal was reduced about fivefold compared with the tub1-Glu strain, to reach barely detectable levels (Fig. 1A,B). Bik1p localization at SPBs was also reduced (Fig. 1A,B). Blot analysis showed similar levels of Bik1p-3GFP expression in TUB1 and tub1-Glu strains deleted or not for BIM1 (Fig. 1C), indicating that Bik1p depletion at MT plus ends in Glu tubulin strains did not reflect variations in protein expression levels. Thus, in the tub1-Glu strain, Bim1p is a determinant for Bik1p plus-end tracking.

Why should the Glu tubulin mutation suppress Bik1p tracking of plus ends? In the context of the current models, in which Bik1p end tracks plus ends principally as a cargo of Kip2p, Bik1p plus-end tracking could be affected because of abnormal interactions of Kip2p with Glu MTs. To test this possibility, we examined the behaviour of a Kip2p-3YFP fusion protein expressed at physiological levels. In both TUB1 and tub1-Glu strains, Kip2p-3YFP was visible as fluorescent dots at MT ends (Fig. 2A). In a quantitative analysis, Kip2p-3YFP signals at MT tips were not affected by the Glu tubulin mutation (Fig. 2B). Kip2p has been shown to end-track both growing and shrinking MTs (Carvalho et al., 2004). Accordingly, in video-microscopy experiments, Kip2p-3YFP tracking of growing and shrinking MT plus ends was visible in both TUB1 and tub1-Glu strains (data not shown and Fig. 2C,D, supplementary material Movies 1, 2). The number of Kip2p-decorated MT ends as well as MT plus-end dynamics were similar in TUB1 and tub1-Glu strains (Table 2). In previous studies, Bim1p removal enhanced Kip2p MT plus-end labelling (Carvalho et al., 2004). Accordingly, in both TUB1 bim1Δ and tub1-Glu bim1Δ strains, the Kip2p-3YFP signal at MT plus ends increased (Fig. 2A,B). BIM1 deletion induced modifications of MT plus-end dynamics, as previously observed (Tirnauer et al., 1999), but the MT plus-end dynamics were not significantly affected by the Glu tubulin mutation (Tables 1, 2). Thus, Kip2p localization and dynamic behaviour at MT plus ends was not detectably affected by the Glu tubulin mutation.

In the wild-type strain, the motion of fluorescent Kip2p along astral MTs towards MT plus ends can be visualized as low-intensity speckles moving along MTs towards the plus end at a rate that exceeds that of MT polymerization (Carvalho et al., 2004). Such speckles could be detected in the tub1-Glu strain (Fig. 2E, supplementary material Movie 3). In the wild-type strain, Bik1p was transported towards MT plus ends as a cargo of Kip2p. Accordingly, we observed Bik1p-3GFP low-intensity speckles sliding along Glu MTs in a manner that suggested motor-dependent transport (Fig. 2F, supplementary material Movie 4). Thus, Kip2p plus-end-directed motion along astral MTs, as well as Bik1p transport on astral MTs, were apparently present in the Glu-tubulin-expressing strain.

Our results demonstrate that the formation of Bik1p comets is suppressed at plus ends in Glu tubulin strains despite the persistence of normal Kip2p comets. Such results indicate that MT plus-end tracking by Bik1p and Kip2p are distinct processes.

In current models, dynein is recruited at MT plus ends by Bik1p. One would then expect dynein to be depleted at Glu MT plus ends. However, when we analyzed the localization of Dyn1p-3GFP fusion proteins expressed at physiological levels, Dyn1p-3GFP was visible as fluorescent dots at the plus ends of growing or shrinking MTs in both TUB1 and tub1-Glu strains (Fig. 3 and supplementary material Movies 5, 6). Furthermore, BIM1 deletion produced no detectable difference in Dyn1p-3GFP dot signals and intensity even in the tub1-Glu bim1Δ strain, in which Bik1p-3GFP was not detectable at MT plus ends (Fig. 3A,B). It could be that, whereas Bik1p is required for dynein localization at MT plus ends in wild-type strains, Bik1p is no longer needed for dynein localization at MT plus ends in the Glu tubulin mutant. We tested this possibility by examining dynein localization in TUB1 bik1Δ and tub1-Glu bik1Δ strains. In both cases, dynein became undetectable at MT plus ends (Fig. 3B). Thus, Bik1p is strictly required for dynein positioning at MT plus ends even in the Glu tubulin mutant.

A possibility was that, in tub1-Glu and tub1-Glu bim1Δ strains, dynein could be delivered at MT ends as a partner of Bik1p, in a complex with Kip2p, and could subsequently end-track MTs independently of Bik1p. To test for the existence of dynein transport along astral MTs, we searched for dynein fluorescent speckles moving along MTs in the tub1-Glu strain. We could indeed observe such speckles (Fig. 3E and supplementary material Movie 7), which is strongly indicative of motor-dependent dynein transport towards MT plus ends.

To test whether Kip2p was indeed involved in dynein transport, we constructed a double-deleted tub1-Glu bim1Δ kip2Δ strain. In this strain, cell viability was compromised, with 56% spore lethality. However, viable cells grew well (Fig. 4A) and astral MTs were readily observable (Fig. 4B and Table 1). Interestingly, dynein was no longer detectable at MT plus ends (Fig. 3B). Correlatively, in video-microscopy experiments, spindle positioning was conspicuously perturbed with spindle elongation occurring in the mother cell and a marked delay of spindle penetration in the bud (Fig. 4C-E). There are obviously interesting questions concerning the mechanisms that finally position the spindle in tub1-Glu bim1Δ kip2Δ strains. Within the scope of our study, we principally conclude that both Bik1p and Kip2p are essential for dynein localization in the tub1-Glu bim1Δ strain; this finding is compatible with the idea that dynein is transported to plus ends in a complex with Bik1p and Kip2p.

Apart from this main conclusion, we note that astral MTs of sizeable length persisted the in KIP2-deleted strain, which contrasted with previous observations of drastic MT-length reduction following KIP2 deletion (Cottingham and Hoyt, 1997). We believe that this phenotypic variation, which we observed in TUB1 and tub1-Glu strains, might result from the absence of TUB3 in our strains (Table 1 and supplementary material Table S1).

The data shown above demonstrate the existence of a pathway for dynein positioning in which Bik1p is only needed as a linker between Kip2p and dynein, being otherwise dispensable at MT plus ends. However, such results did not exclude the possibility of dynein recruitment at plus ends by Bik1p, as in previous models. To test for such recruitment in our strains, we examined strains lacking Kip2p but still displaying variable amounts of Bik1p at MT plus ends. As stated above, we note that, in our KIP2-deleted strains, MT length was not affected as dramatically as previously described on different genetic backgrounds (Table 1).

We examined the effect of KIP2 deletion in the tub1-Glu strain. Bik1p was detectable at MT plus ends in such strains, which is compatible with Bim1p-mediated localization (see above). Dynein was also detectable at tub1-Glu kip2Δ MT ends.

We similarly examined the effect of KIP2 deletion in the TUB1 strain. KIP2 deletion induced proportional depletion of Bik1p and dynein at MT plus ends (Fig. 3B). Additional deletion of BIM1 in these wild-type-tubulin strains did not induce any detectable changes in Bik1p or dynein signals at MT ends (Fig. 3B).

Collectively, these results indicate that dynein can be directly recruited at plus ends by Bik1p, when the latter protein is present at MT ends.

In this study, we provide strong evidence that, although Bik1p is brought to MT plus ends by Kip2p, MT end tracking by Bik1p and Kip2p are distinct processes.

A key factor for deciphering the molecular interactions involved in Bik1p plus-end tracking has been the use of the Glu tubulin mutation, which, in recent studies, has been shown to specifically inhibit the interaction of CAP-Gly MT-binding domains with the C-terminus of α-tubulin (Honnappa et al., 2006; Mishima et al., 2007; Weisbrich et al., 2007). In our previous study, the Glu mutation only induced a threefold decrease of the Bik1p signal at MT ends. The existence of a sizeable remaining Bik1p signal could suggest incomplete inhibition of Bik1p interaction with tubulin by the Glu mutation in budding yeast. However, in this study we demonstrate that, in the tub1-Glu bim1Δ strain, the Bik1p signal at MT ends decreases to reach barely detectable levels. In many cell types, the localization of Bik1p orthologues depends on EB1 orthologues (Busch and Brunner, 2004; Komarova et al., 2005; Lansbergen and Akhmanova, 2006). Budding yeast looked to be an exception, with Bik1p localization being insensitive to BIM1 deletion in wild-type strains (Lin et al., 2001) (Fig. 1). Apparently, the contribution of Bim1p to Bik1p localization is minimal in wild-type strains, but becomes readily apparent in Glu tubulin mutant strains.

Previous work has demonstrated that Bik1p is transported towards MT ends in a complex with Kip2p and that such Kip2p-dependent transport is important for Bik1p localization at MT plus ends (Carvalho et al., 2004). Accordingly, in the present study, we found similar Kip2p-dependent transport of Bik1p both in wild-type and Glu tubulin strains. Additionally, our data indicate that the Bik1p-Kip2p complex dissociates upon arrival at plus ends, with further MT plus-end tracking by Bik1p depending principally upon Bik1p interaction with tubulin. What could trigger the dissociation of the Bik1p-Kip2p complex at plus ends? It is possible that the conformation of Kip2p and its interactions with tubulin are different when the motor walks on MTs and when it end-tracks MTs. The two processes actually occur with profoundly different velocities, with Kip2p motion on MTs being three- to five-fold faster than MT growth (Carminati and Stearns, 1997; Carvalho et al., 2004; Shaw et al., 1997). In the case of another plus-end-directed +TIP kinesin, MCAK, plus-end tracking involves domains distinct from the motor domain, which is apparently inactive during end tracking (Moore et al., 2005). Kip2p might similarly be inactive as a motor during plus-end tracking, with a conformation that does not allow interaction with Bik1p. How could Bik1p end-track MTs once it is dissociated from Kip2p? It is possible that Bik1p binds to free tubulin and co-polymerizes with tubulin exactly as suggested in the case of CLIP-170 in mammalian cells (Galjart and Perez, 2003). However, such a model would only account for Bik1p tracking of growing plus ends, whereas Bik1p also tracks shrinking plus ends in budding yeast (Carvalho et al., 2004), suggesting the possibility of a direct association of Bik1p with plus ends whether growing or shrinking. In any case, the concentration of Bik1p at plus ends implies that Bik1p recognizes a tubulin conformation specific to MT ends, in addition to the tubulin C-terminus, as previously suggested in the case of other +TIPS (for reviews, see Carvalho et al., 2003; Galjart and Perez, 2003; Lansbergen and Akhmanova, 2006).

Collectively, our data suggest that, in wild-type strains, Bik1p can track MT plus ends owing to its intrinsic tubulin-binding capacity, in a manner very much similar to that proposed for mammalian cells, and then functions to recruit dynein at MT plus ends. Kip2p transport of Bik1p apparently facilitates Bik1p concentration close to plus ends, without being strictly required.

Kip2p has been previously shown to be important for dynein localization at MT plus ends (Carvalho et al., 2004), which could be explained by dynein recruitment at MT plus ends in a complex with Bik1p. We provide evidence that Kip2p also sustains another, functionally redundant, pathway for dynein localization and transport. In this pathway, dynein is apparently transported along MTs, to plus ends, in a Bik1p-Kip2p-dependent manner, and can then tracks MT plus ends even in the absence of Bik1p. Previous studies have indicated that dynein might normally track plus ends in a ternary complex with Pac1p and Bik1p (Lee et al., 2003; Li et al., 2005; Sheeman et al., 2003). Maybe, when Bik1p is lacking, dynein interaction with Pac1p is sufficient to sustain dynein tracking of plus ends. One could then wonder why dynein could not be recruited directly by Pac1p. Maybe, dynein interaction with Bik1p releases an inhibition that otherwise precludes dynein interaction with Pac1p. A schematic summary of these models and conclusion is shown Fig. 5.

Finally, it will be of obvious interest to know whether a kinesin-dependent pathway for CLIP-170 mediated localization of dynein at plus ends exists in mammalian cells. One way to know could be to test for the existence of dynein comets in tubulin-tyrosine-ligase-deficient cells (Glu tubulin cells). The observation of such comets would strongly favour the possibility of kinesin-mediated transport of both CLIP-170 and dynein in mammalian cells, as in yeast.

The plasmids and strains used in this study were isogenic to the S288c background and are listed in supplementary material Table S1. Deletion mutants were obtained by one-step gene replacement and verified by PCR. pFC1 is the integrating Bik1p-3GFP plasmid, the hphMX4 cassette from pMJ696 [pAG32 in Goldstein and McCusker (Goldstein and McCusker, 1999)] was cloned AccI-NotI in place of TRP1 in pB1587 (Lin et al., 2001). pFC2 is the 3′region of DYN1 fused to 3GFP of pB1761 (Sheeman et al., 2003) cloned EcoRI-NotI in pRS406. pFC3 is the 3′region of KIP2 fused to 3YFP of pMG108 (Carvalho et al., 2004) cloned SalI-NotI to pRS406. pAFS125-Glu is a tub1-Glu C-terminal mutation cloned SacI-BstEII from pUC18-tub1-Glu (Badin-Larcon et al., 2004) in pAFS125.

Western blots were performed on whole-cell extract from the indicated strain with anti-GFP antibody (Molecular Probes). Correction for loading equality was achieved with whole α-tubulin signal quantification using mAb YOL1/34 (Sera-Lab, Crawley Down, Sussex, UK).

Cell imaging was performed on a Zeiss Axiovert microscope equipped with a Cool Snap ES CCD camera (Ropper Scientific). All images were captured using 2×2 binning and 17 sequential z-axes collected at 0.3-μm, the exposure time varying between constructions. Time-lapse video microscopy for Kip2p-3YFP and Dyn1p-GFP were collected with five sequential z-axes at 0.5 μm steps and an exposure time of 800 ms every 40 seconds. For MT-length measurements: images of maximal intensity projection of cells bearing GFP-TUB1 or GFP-tub1-Glu fusions integrated at the URA3 locus are used to visualize MTs. For MT-dynamics measurements, Kip2p-3YFP is the reporter and images were collected with five sequential z-axes at 0.5 μm steps every 10 seconds. Speckles were observed using a single plane exposed for 1 second every 5 seconds for Dyn1p-3GFP, and every 2.5 seconds for Bik1p-3GFP and Kip2p-3YFP. All image manipulations and fluorescence-intensity measurements were performed using MetaMorph software (Universal Imaging). For technical consideration and fluorescence intensity, those experiment are done with only one protein tagged. SPBs and plus ends are distinguish based on their movement during the whole time-lapse analysis (SPB exhibit very little movement, plus ends exhibit intensive movements). Fluorescent counting was performed with maximal intensity projections computed from the original datasets and corrected for cytoplasmic background fluorescence. Bik1p-3GFP quantifications were done on pre-anaphase cells. Plus-end fluorescence quantifications were an average of MT plus-end signal over cell number. We proceeded to statistical analysis (Student's t-test) and determined that all obvious differences were significant; P<0.001.

We thank D. Pellman for the generous gift of plasmids; A. W. Murray for providing the GFP-tubulin plasmid; Y. Saoudi for help in imaging; J. M. Soleilhac for preliminary work; A. Fourest-Lieuvin and J. Brocard for help with MT dynamics analysis; E. Denarier for help in submission. This work was supported in part by a grant from la Ligue Nationale contre le Cancer to D.J., ANR `Tyr-Tips' to D.J., la région Rhône-Alpes to F.C. and C.B., and ARC to F.C.