We describe the identification of GIM1/YKE2, GIM2/PAC10, GIM3, GIM4 and GIM5 in a screen for mutants that are synthetically lethal with tub4‐1, encoding a mutated yeast γ‐tubulin. The cytoplasmic Gim proteins encoded by these GIM genes are present in common complexes as judged by co‐immunoprecipitation and gel filtration experiments. The disruption of any of these genes results in similar phenotypes: the gim null mutants are synthetically lethal with tub4‐1 and super‐sensitive towards the microtubule‐depolymerizing drug benomyl. All except Δgim4 are cold‐sensitive and their microtubules disassemble at 14°C. The Gim proteins have one function related to α‐tubulin and another to Tub4p, supported by the finding that the benomyl super‐sensitivity is caused by a reduced level of α‐tubulin while the synthetic lethality with tub4‐1 is not. In addition, GIM1/YKE2 genetically interacts with two distinct classes of genes, one of which is involved in tubulin folding and the other in microtubule nucleation. We show that the Gim proteins are important for Tub4p function and bind to overproduced Tub4p. The mammalian homologues of GIM1/YKE2 and GIM2/PAC10 rescue the synthetically lethal phenotype with tub4‐1 as well as the cold‐sensitivity and benomyl super‐sensitivity of the yeast deletion mutants. We suggest that the Gim proteins form a protein complex that promotes formation of functional α‐ and γ‐tubulin.
Microtubules are hollow cylinders with a diameter of 25 nm. They are part of the cytoskeleton of eukaryotic cells and play an important role in numerous cellular processes, including chromosome segregation in mitosis and meiosis, organelle positioning and intracellular transport. The wall of the microtubule cylinder consists of 13 protofilaments which are strings of alternating α‐ and β‐tubulin subunits pointing in the same direction (reviewed by Mandelkow and Mandelkow, 1993). Therefore, microtubules are polar structures with biochemically distinct ends. Microtubule formation is a complex process that includes the post‐translational folding of α‐ and β‐tubulin, their assembly into the heterodimer tubulin and finally the formation of microtubules from tubulin subunits (Solomon, 1991). All of these steps are assisted by proteins in vivo.
The first post‐translational step in the pathway leading to the formation of the tubulin heterodimer is the folding of α‐ and β‐tubulin. Genetic (Ursic and Culbertson, 1991; Chen et al., 1994; Vinh and Drubin, 1994; Archer et al., 1995) and biochemical (Frydman et al., 1992; Yaffe et al., 1992; Melki et al., 1993; Sternlicht et al., 1993; Tian et al., 1996) evidence suggests that tubulins undergo facilitated folding via their interaction with cytoplasmic chaperonins. Conditional lethal mutations in TCP1, BIN2, BIN3 and ANC2, coding for subunits of cytoplasmic chaperonin (TRiC), affect microtubule as well as actin functions (Ursic and Culbertson, 1991; Chen et al., 1994; Li et al., 1994; Miklos et al., 1994; Vinh and Drubin, 1994). For example, tcp1‐1 cells are cold‐sensitive for growth, and accumulate multi‐nucleated and anucleated cells with abnormal microtubule structures. This mutant is also sensitive towards the microtubule‐depolymerizing drug benomyl (Ursic and Culbertson, 1991). Remarkably, bin3‐1 cells display a normal distribution of unbudded, small‐budded and large‐budded cells, although the microtubules of bin3‐1 cells disassemble at 14°C (Chen et al., 1994).
While the cytoplasmic chaperonin is sufficient for the production of native actin in an ATP‐dependent manner (Gao et al., 1992), folding of α‐ and β‐tubulin requires the participation of additional cofactors. Such proteins, named cofactors A, B, C, D and E, have been identified by biochemical approaches (Tian et al., 1996, 1997). Potential yeast homologues of cofactors A, B, D and E are Rbl2p, Alf1p, Cin1p and Pac2p (Archer et al., 1995; Tian et al., 1996; Geiser et al., 1997). Surprisingly, ALF1, RBL2, CIN1 and PAC2 are not essential for growth of yeast cells (Hoyt et al., 1990; Stearns et al., 1990; Archer et al., 1995). However, their deletion causes sensitivity towards benomyl and they show multiple genetic interactions with genes coding for components of the microtubule system. In addition, Rbl2p interacts directly with yeast β‐tubulin (Archer et al., 1995).
In many cell types, microtubule assembly is initiated at centrosomes, basal bodies, spindle pole bodies (SPBs) or nucleus‐associated bodies, structures for which Pickett‐Heaps (1969) coined the generic term microtubule‐organizing centre (MTOC). MTOCs assist the assembly of the first tubulin monomers into oligomers, a step known as microtubule nucleation (Mitchison and Kirschner, 1984). A universal component of MTOCs involved in microtubule nucleation is γ‐tubulin (Oakley et al., 1990; Horio et al., 1991; Stearns et al., 1991). γ‐Tubulin forms a complex with additional proteins (Stearns and Kirschner, 1994; Moudjou et al., 1996; Akashi et al., 1997), and purification of such a complex from Xenopus laevis eggs identified α‐, β‐ and γ‐tubulin and proteins with mol. wts of 195, 133, 109 and 75 kDa (Zheng et al., 1995).
In the yeast Saccharomyces cerevisiae, the MTOC is known as the SPB. It is a multi‐layered structure which is embedded in the nuclear envelope (Byers and Goetsch, 1975; Byers, 1981). The yeast γ‐tubulin is encoded by the essential TUB4 gene (Sobel and Snyder, 1995; Marschall et al., 1996; Spang et al., 1996). Tub4p forms a 6S complex with the SPB components Spc98p and Spc97p (Geissler et al., 1996; Spang et al., 1996; Knop et al., 1997). Purification of this complex suggests that it contains only one molecule of Spc98p and Spc97p, but two or more molecules of Tub4p (Knop and Schiebel, 1997). Fractionation experiments suggest that the Tub4p complex assembles in the cytoplasm followed by its nuclear import via an essential nuclear localization sequence in Spc98p (G. Pereira et al., 1998). Finally, Spc98p and Spc97p of the Tub4p complex interact in the nucleus with the amino‐terminal domain of the SPB component Spc110p (Knop and Schiebel, 1997). The binding site for the Tub4p complex at the outer plaque is still unknown.
To identify genes that are involved in yeast γ‐tubulin functions, we performed a genetic screen for mutants that are synthetically lethal with tub4‐1. We identified SPC98 and SPC97 coding for components of the yeast γ‐tubulin complex and five additional genes, which we named GIM1–GIM5. Their products are not associated with the SPB, instead they form cytoplasmic multi‐protein complexes which promote formation of functional α‐tubulin and Tub4p. Most interestingly, the Gim proteins are phylogenetically conserved proteins, and mouse and human homologues function in yeast, indicating that the mammalian proteins fulfil a very similar role.
Identification of GIM1/YKE2, GIM2/PAC10, GIM3, GIM4 and GIM5 in a screen for mutants that are synthetically lethal with a mutated yeast γ‐tubulin
In the yeast S.cerevisiae, the γ‐tubulin Tub4p forms a complex with the SPB components Spc98p and Spc97p. Physical interaction between Tub4p, Spc98p and Spc97p is reflected by multiple genetic interactions, including synthetic lethality (Geissler et al., 1996; Knop et al., 1997). The latter is an indication of a functional relationship of two gene products. To identify further components that functionally interact with TUB4, we performed a genetic screen for mutants that are synthetically lethal with tub4‐1, yielding 12 mutants with this phenotype. Subsequent analysis showed that two of these mutants were defective in SPC98 and two in SPC97. The other eight belong to five complementation groups which we named GIM1–GIM5 (genes involved in microtubule biogenesis). Plasmids containing GIM1–GIM5 were isolated by transforming the mutants with a yeast genomic library. Subcloning of DNA fragments and complementation analysis showed that GIM1 is identical to YKE2 (Shang et al., 1994), coding for a protein of 114 amino acids (Table I). GIM2 had already been identified as PAC10, a gene that becomes essential in the absence of the CIN8‐encoded kinesin motor (Geiser et al., 1997). GIM3 corresponds to the open reading frame (ORF) YNL153c, encoding a protein of 129 amino acids. The DNA fragment complementing gim4 cells contained three overlapping ORFs, one of which was interrupted by an intron. The various coding regions were cloned directly behind the ADH promoter of plasmid p415‐ADH. Only the ADH promoter fusion with the intron‐containing ORF YEL003w complemented gim4 cells (data not shown). Gim4p is a protein of 132 amino acids. GIM5 (ORF YML094w) also contains a small intron and encodes a protein of 163 amino acids.
The Gim proteins are phylogenetically conserved coiled‐coil proteins
As reported for Gim2p/Pac10p (Geiser et al., 1997), all Gim proteins have a high probability of forming coiled‐coils (Lupas, 1996). Sequence comparisons have shown that Gim1p/Yke2p, Gim2p/Pac10p, Gim3p, Gim4p and Gim5p are related proteins. They also have relatives in mammals, Caenorhabditis elegans and Schizosaccharomyces pombe, revealing that they are phylogenetically conserved (Figure 1). Remarkably, Gim1p/Yke2p and Gim5p are homologous to two proteins of the archaebacterium Methanococcus jannaschii, while no relatives were found in eubacteria. We also noticed that the Gim proteins are more closely related to their homologues from other species than to each other (Figure 1).
Deletion of the GIM genes causes microtubule defects
To understand the function of the GIM gene products, we investigated whether they are essential for growth of yeast cells. We disrupted the entire coding regions of the GIM genes in the diploid yeast strain YPH501. As shown for Δgim5 (Figure 2A), spore analysis revealed that GIM1/YKE2, GIM3, GIM4 and GIM5 are not essential for growth. In agreement with a previous report (Geiser et al., 1997), we observed that Δpac10 cells were defective in spore germination. Such a defect was not apparent for the other GIM mutants, raising the possibility that GIM2/PAC10 has an additional specialized function in spore germination. The disruption of any of the GIM genes resulted in a slow growth phenotype at 30°C (Figure 2A and B). A more detailed analysis indicated that the doubling time of Δgim1/yke2, Δgim2/pac10, Δgim3 and Δgim5 cells grown in liquid medium at 30°C was increased by a factor of 1.4 compared with the wild‐type, while the doubling time of Δgim4 cells was only affected 1.1‐fold. It has been reported previously that a deletion mutant of GIM1/YKE2 shows normal growth over a wide range of temperatures (Shang et al., 1994). This discrepancy with our study is explained either by strain differences or by the fact that Shang et al. (1994) did not disrupt the entire coding region of YKE2.
Common phenotypes of mutants affecting the microtubule cytoskeleton are a cold‐sensitive growth defect and an increased sensitivity towards the microtubule‐depolymerizing drug benomyl (Neff et al., 1983; Huffaker et al., 1988). Therefore, we investigated whether the growth defects of the haploid gim null strains were intensified at lower temperatures. The gim null strains showed an enhanced growth defect at reduced temperatures in comparison with the wild‐type (Figure 2B; compare 30°C with 23°C and 14°C plates). An exception was Δgim4 cells, which did not show an enhanced growth defect at lower temperatures (Table I). In contrast to the wild‐type, all gim null strains were super‐sensitive to 2.5 μg/ml benomyl (Figure 2C). Such a strong benomyl sensitivity has only been observed for mutants defective in microtubule biogenesis (Stearns et al., 1990; Chen et al., 1994; Archer et al., 1995; Tian et al., 1997).
We investigated whether the Gim proteins function in parallel pathways. In this case double mutants should have more pronounced defects compared with the single mutants. However, the double deletion mutants of Δgim1/yke2 together with any one of Δgim2/pac10, Δgim3, Δgim4 or Δgim5 were viable and as benomyl‐sensitive as the single mutants. We also tested whether overexpression of any of the GIM genes rescued the synthetically lethal phenotype of gim tub4‐1 cells. This was the case for GIM1 which weakly suppressed the defect of gim4 tub4‐1 cells (data not shown).
The strong benomyl super‐sensitivity of the haploid gim null strains suggests that the gene products are needed for either tubulin formation or microtubule stability. To analyse these possibilities, we investigated, by indirect immunofluorescence, the microtubule arrays of Δgim1/yke2, Δgim2/pac10 and Δgim3 cells incubated at 14°C for 20 h. The phenotype is shown for Δgim1/yke2 cells as a representative, since it is very similar to that of the other mutants. Δgim cells lost most of their cytoplasmic and nuclear microtubules, leaving only a small spot near the nucleus, most likely at the SPB (Figure 3A). This observation suggests that microtubule attachment to the SPB via the Tub4p complex is still taking place in Δgim cells. Instead microtubule stability seems to be impaired. Furthermore, cells with a large bud (Figure 3A, arrow) had only one 4′,6'‐diamidino‐2‐phenylindole (DAPI) staining region in one of the two cell bodies, a phenotype that is consistent with the observed defect in microtubule organization. Despite these defects, gim null cells incubated for 20 h at 14°C were still as viable as wild‐type cells when then grown at 30°C (data not shown). Based on the distributions of the DNA content (Figure 3B) and of the cell morphologies (Figure 3A), Δgim cells did not arrest at a defined stage of the cell cycle. This observation was surprising, since many mutants with a defect in spindle formation arrest in mitosis due to a mitotic check point (Hoyt et al., 1990; Li and Murray, 1991). The non‐arrest phenotype could point to a role for the Gim proteins in mitotic check point control. However, all GIM deletion mutants arrested in the cell cycle like wild‐type cells in response to the microtubule‐depolymerizing drug nocodazole, while the mitotic check point control mutant bub2 did not (Hoyt et al., 1990; Li and Murray, 1991; Geiser et al., 1997; data not shown). This led us to conclude that the GIM gene products are not part of a mitotic check point. Taken together, the strong benomyl sensitivity and the microtubule defects of the gim null mutants at reduced temperature are most consistent with a role for the Gim proteins in tubulin biogenesis.
gim null mutants are sensitive towards the actin inhibitor latrunculin‐A and are osmotically sensitive
We noticed that the volume of cells of the GIM deletion mutants incubated at 14°C was ∼2.2‐ to 2.4‐fold increased compared with the wild‐type (Table I). Such a phenotype has been reported, among others, for mutants affecting the actin cytoskeleton (reviewed by Drubin, 1990). Many mutants with actin defects are sensitive towards the actin‐binding drug latrunculin‐A (Ayscough et al., 1997) and are osmotically sensitive (Drubin, 1990). We found that the gim null strains were more sensitive towards latrunculin‐A than the wild‐type, which indicates an influence of the Gim proteins on the actin cytoskeleton (Figure 3C). This sensitivity was stronger compared with that of the cold‐sensitive mutant tcp1‐1, but was, however, clearly weaker compared with the actin mutant act1‐2. TCP1 encodes a component of TRiC which is required for actin folding (Ursic and Culbertson, 1991). Furthermore, similarly to actin mutants, Δgim1/yke2, Δgim2/pac10 and Δgim5 cells did not grow on plates containing 1.4 M sorbitol. In contrast, Δgim3 cells grew slowly, while Δgim4 grew as wild‐type cells (Figure 3D). At 1.8 M sorbitol, Δgim3 cells failed to grow and a weak growth defect of Δgim4 cells was observed (Figure 3D). We sought additional evidence for a role for the Gim proteins in actin function. However, deletion of GIM1/YKE2 was synthetically lethal with neither act1‐2 nor act1‐3. In addition, Δgim1/yke2, Δgim2/pac10, Δgim3 and Δgim4 cells did not show an obvious defect in actin organization as judged by indirect immunofluorescence using anti‐actin antibodies (data not shown), suggesting that the gim null mutants have only subtle actin defects. In conclusion, the GIM genes may have multiple functions, since their deletion causes microtubule defects as well as an increased sensitivity towards the actin‐specific drug latrunculin‐A and towards high osmolarity.
The GIM genes interact genetically with genes involved in microtubule biogenesis
To gain further insight into the role of the GIM genes, we investigated their genetic interactions with other genes involved in microtubule function. We first addressed the question of whether total loss of GIM gene function is synthetically lethal with mutants of the Tub4p complex. Using the approach described in Figure 4A, we could show that this is indeed the case for tub4‐1. This result was confirmed by tetrad analysis using a GIM4/Δgim4::HIS3MX6 TUB4/tub4‐1 strain. No viable temperature‐sensitive HIS3 spores were obtained, confirming that Δgim4::HIS3MX6 combined with tub4‐1 is lethal. Microscopic inspection of the non‐colony‐forming spores identified several hundred cells. Therefore, Δgim4 tub4‐1 cells died only after multiple duplications. A similar result was obtained with GIM3 (data not shown).
In a similar manner, we tested whether GIM1/YKE2, GIM2/PAC10 and GIM4 showed interactions with SPC98 and SPC97, coding for components of the yeast γ‐tubulin complex. Δgim1/yke2 was synthetically lethal with spc98‐2 and synthetically toxic when combined with spc98‐1 (Figure 4B). Allele‐specific genetic interactions of GIM1/YKE2 were also observed with SPC97: deletion of GIM1/YKE2 was synthetically toxic in spc97‐20 cells, while it did not affect growth of the spc97‐14 mutant (Figure 4C). In contrast to Δgim1/yke2, Δgim2/pac10 or Δgim4 combined with spc98‐1, spc98‐2 or spc97‐20 hardly affected growth (Table II). It is important to note that the spc98 and spc97 alleles have distinct phenotypes and are suppressed differently by TUB4 (Geissler et al., 1996; Knop et al., 1997). In particular, spc97‐20 has an SPB duplication defect, while spc97‐14 fails to form a mitotic spindle (Knop et al., 1997).
Finally, GIM1/YKE2, GIM2/PAC10 and GIM4 were in part tested for their genetic interactions with TUB1, TUB2, BIN2, BIN3 and RBL2. TUB1 and TUB2 code for α‐ and β‐tubulin in yeast (Neff et al., 1983; Schatz et al., 1986a), while RBL2 encodes a β‐tubulin‐binding protein, the homologue of mammalian cofactor A (Archer et al., 1995; Tian et al., 1996). Bin2p and Bin3p are subunits of TRiC (Chen et al., 1994). Δgim1 showed synthetic lethality with tub1‐4, tub2‐403, tub2‐405, bin2‐1, bin3‐1 and Δrbl2 (Table II). In contrast, Δgim4 was only synthetically lethal in combination with bin2‐1, bin3‐1 and Δrbl2, but not with tub2‐403 or tub2‐405. In summary, the GIM genes show a broad range of genetic interactions with genes involved in tubulin biogenesis and microtubule nucleation.
Deletion of the GIM genes results in reduced levels of α‐tubulin, which explains the benomyl super‐sensitivity, but not the synthetic lethality with tub4‐1
A defect of the gim null mutants in tubulin biogenesis may result in reduced levels of α‐, β‐ or γ‐tubulin. The amount of these proteins was determined by immunoblotting using specific antibodies. While the α‐tubulin content was about half to a third compared with wild‐type, β‐tubulin and Tub4p levels were approximately the same (Figure 5A and B). It is noteworthy that deletion of GIM4 affected α‐tubulin less severely than deletion of the other GIM genes. We also noticed that the reduction of α‐tubulin levels was independent of the incubation temperature of the cultures (data not shown).
An increase of β‐tubulin over α‐tubulin levels, which is the case in the gim null mutants, is toxic for yeast cells (Burke et al., 1989; Katz et al., 1990; Schatz et al., 1986b). Therefore, increasing the α‐tubulin content by overexpression of TUB1 or reducing the pool of free β‐tubulin by overproduction of the β‐tubulin‐binding protein Rbl2p may suppress the tubulin defects. Indeed, the benomyl super‐sensitivity of the five Δgim mutants was fully rescued when TUB1 or RBL2 were overexpressed using either elevated gene dosage (2μ plasmids) or the strong ADH promoter (Table I), while overexpression of TUB4 showed no effect (Figure 5C; shown for Δgim1/yke2). We then investigated whether the reduced level of α‐tubulin is also the cause for the synthetically lethal phenotype with tub4‐1. However, overexpression of either TUB1 or RBL2 did not suppress the synthetic lethality, while plasmid‐encoded TUB4 and the GIM genes complemented the mutants (Figure 5D; shown for Δgim1/yke2). In summary, the benomyl super‐sensitivity of the GIM mutants is the result of a reduced level of α‐tubulin; however, this is not the reason for the synthetically lethal phenotype with tub4‐1. This suggests that the Gim proteins have at least two functions, one related to α‐tubulin and the other to Tub4p.
Gim1p/Yke2p, Gim2p/Pac10p, Gim3p and Gim5p bind to overproduced Tub4p
The failure of overexpressed TUB1 and RBL2 to suppress the synthetically lethal phenotype with tub4‐1 and the allele‐specific interaction of GIM1/YKE2 with SPC98 and SPC97 point to a more direct role for the Gim proteins in Tub4p complex function. This notion was supported further by the finding that overexpression of TUB4 was toxic in the Δgim1/yke2 null strain (Figure 6A, sector 1), while it hardly affected growth of the wild‐type (Figure 6A, sector 3) (Spang et al., 1996; Knop et al., 1997). Δgim1/yke2 cells overexpressing TUB4 frequently were large‐budded and showed abnormal bud morphologies (Figure 6A), phenotypes which were observed neither in Δgim1/yke2 cells nor in wild‐type cells overexpressing TUB4 (data not shown). Similarly, simultaneous overexpression of chromosomally integrated TUB4, SPC98 and SPC97 was not toxic for wild‐type cells; however, it strongly affected growth of the gim1/yke2 null mutant (Figure 6B). As expected, this defect was rescued by plasmid‐encoded Gim1p.
This result, together with the observed synthetically lethal phenotype with tub4‐1, suggests that the GIM genes become essential either after overexpression of TUB4 or in the presence of a mutated TUB4. This consideration led us to test whether the Gim proteins bind to overproduced Tub4p. A gene fusion consisting of TUB4 and three repeats of the haemagglutinin (HA) epitope‐encoding sequence was used for this experiment. TUB4‐3HA was overexpressed in wild‐type cells (Figure 6C), followed by the immunoprecipitation of Tub4p‐3HA using anti‐HA antibodies. Interestingly, the precipitate not only contained Tub4p‐3HA (lane 3), but also Gim1p/Yke2p (lane 5), Gim2p/Pac10p (lane 7), Gim3p (lane 9) and Gim5p (lane 11). This co‐immunoprecipitation was specific as no, or hardly any, Gim protein was detected when the anti‐HA precipitation was performed with an extract containing overproduced Tub4p (lanes 6, 8, 10 and 12) instead of Tub4p‐3HA. Gim4p was not tested due to the lack of an antibody. In summary, our results show that GIM1/YKE2 becomes essential after TUB4 overexpression and that the Gim proteins bind to overproduced Tub4p.
The Gim proteins interact with each other and are present in common complexes
The interactions between the Gim proteins were investigated by immunoprecipitation experiments, using functional, epitope‐tagged versions of the Gim proteins (Figure 7A). For example, co‐immunoprecipitation of Gim2p/Pac10p‐MYC with Gim1p/Yke2p‐3HA was observed. Similarly, Gim3p‐MYC co‐precipitated with Gim2p/Pac10p‐3HA and Gim4p‐3HA precipitated Gim2p/Pac10p‐MYC. The result of the immunoprecipitation was confirmed using a functional gene fusion between GIM5 and protein A (GIM5‐ProA). Gim5p‐ProA was enriched from cell lysates with IgG–Sepharose. We detected not only Gim5p‐ProA in the precipitate, but also Gim1p/Yke2p, Gim2p/Pac10p, Gim3p and Gim4p‐3HA (Figure 7B). Similar results were obtained with Gim2p/Pac10p‐ProA (data not shown). Multiple interactions between the Gim proteins were also observed using the yeast two‐hybrid system (Fields and Song, 1989) (data not shown).
The results of the immunoprecipitations do not allow a decision to be made as to whether the Gim proteins form only one or several complexes. To address this point, yeast extract from wild‐type cells was fractionated by gel filtration (Figure 7C). Immunodetection of the Gim proteins revealed that Gim1p/Yke2p, Gim2p/Pac10p, Gim3p and Gim5p co‐fractionated over a wide range, suggesting that the complexes were heterogeneous in their composition. However, all four proteins peaked in the same fraction corresponding to an apparent mol. wt of ∼170–230 kDa. It is noteworthy that Gim2p/Pac10p was also detected in earlier fractions corresponding to mol. wts up to 2 MDa. In conclusion, the immunoprecipitation experiments demonstrate that the Gim proteins interact with each other and that they are present in common complexes.
The Gim complexes are localized mainly in the cytoplasm of yeast cells
In order to obtain a better understanding of the Gim complexes, we investigated their cellular localization. Localization studies by indirect or direct immunofluorescence using epitope‐tagged or green fluorescent protein (GFP) fusion proteins excluded a localization of the Gim proteins at the SPB. Instead, a cytoplasmic staining was observed (data not shown). This result was confirmed by fractionation experiments (Figure 8). Yeast cell lysates (lane 1) were separated by differential centrifugation into a soluble cytoplasmic fraction (lane 2) and a sediment (lane 3) containing nuclei, mitochondria and other organelles. The distribution of the α‐ and β‐subunits of the cytoplasmic fatty acid synthase (Fas) and the nuclear protein nucleolin (Nop1p) established that the two fractions were hardly contaminated by each other. Gim1p/Yke2p, Gim2p/Pac10p, Gim3p, Gim4p‐3HA and Gim5p were detected predominantly in the cytoplasmic fraction (lane 2). Taken together, the localization studies suggest that the Gim complexes are localized predominantly in the cytoplasm of yeast cells.
The mammalian homologues of Gim1p/Yke2p and Gim2p/Pac10p are fully functional in yeast
Gim1p/Yke2p is 47% identical to the mouse KE2 gene product (Shang et al., 1994) and Gim2p/Pac10p is 43% identical to the human VBP‐1 protein (Geiser et al., 1997). We were interested to see whether the KE2 and VBP‐1 gene products can substitute for Gim1p/Yke2p and Gim2p/Pac10p in yeast. The KE2 and VBP‐1 genes were amplified by PCR from a mouse and a human cDNA library, respectively. Both genes were cloned behind the yeast ADH promoter. Like plasmid‐encoded GIM1/YKE2, the ADH–KE2 promoter fusion complemented the synthetically lethal phenotype of Δgim1/yke2 tub4‐1 (data not shown) as well as the cold‐ and benomyl‐sensitive growth defects of Δgim1/yke2 cells (Figure 9A). Similar results were observed with the human VBP‐1 gene using the Δgim2/pac10 mutant (Figure 9B). Our results indicate that the Gim proteins are not only structurally conserved, but that they also fulfil very similar functions in mammalian and in yeast cells.
Gim1p/Yke2p, Gim2p/Pac10p, Gim3p, Gim4p and Gim5p interact with each other and form complexes
In this study, we performed a genetic screen for mutants that are not viable in combination with a mutated TUB4. We expected to find genes coding either for components of the Tub4p complex or for proteins that are involved in the folding of Tub4p or the assembly of the Tub4p complex. Since the Tub4p complex is associated with SPB substructures that are located in the nucleus (inner plaque) and in the cytoplasm (outer plaque) (Rout and Kilmartin, 1990; Spang et al., 1996; Knop et al., 1997), we could also expect to identify factors that play a role in the nuclear import of the Tub4p complex (G.Pereira et al., 1998). Finally, we expected to find either α‐ or β‐tubulin, since it has been proposed that these proteins interact with γ‐tubulin (Oakley, 1992).
Our genetic screen identified SPC98 and SPC97, which was not surprising, since previous results indicated that they interact genetically with TUB4 (Geissler et al., 1996; Knop et al., 1997). In addition, we identified five novel genes that we named GIM1–GIM5. Two of the GIM genes have been described previously: YKE2 was analysed due to its homology to the mouse KE2 gene, located on the mouse MHC locus (Shang et al., 1994), and PAC10 was found in a screen for genes which become essential in the absence of CIN8, encoding a kinesin‐related protein (Geiser et al., 1997). The co‐immunoprecipitation and the fractionation studies indicate that all five Gim proteins function together and that they are part of either one complex or several complexes containing different Gim proteins and possibly as yet unidentified proteins. In retrospect, our screen was quite selective, as it only identified components of the Tub4p or the Gim complexes.
Gim proteins have multiple functions related to the yeast cytoskeleton
All GIM deletion mutants had a reduced level of α‐tubulin, while the amounts of β‐tubulin and Tub4p were normal. The imbalance of α‐ and β‐tubulin results in an elevated pool of free β‐tubulin which is probably the cause of the microtubule defects. Our conclusion is supported by the observation that the deletion of TUB3, coding for the minor α‐tubulin in yeast, also results in benomyl super‐sensitivity (Schatz et al., 1986b) and that overexpression of TUB2, coding for β‐tubulin, is toxic (Katz et al., 1990). We reasoned, that any manipulation that increases the amount of α‐tubulin or decreases the pool of free β‐tubulin should suppress the tubulin defects. This was indeed the case, as overexpression of TUB1 or RBL2, both of which reduce the level of free β‐tubulin (Archer et al., 1995), cured the benomyl super‐sensitivity of the gim null strains. We also believe that the synthetically lethal defects of GIM deletions with Δrbl2 are the result of too little α‐tubulin, as loss of Rbl2p will increase the pool of free β‐tubulin further, which is then lethal for the cells (Katz et al., 1990; Archer et al., 1995). Furthermore, the reduced level of α‐tubulin in Δpac10 cells may explain why this gene becomes essential in the absence of the microtubule‐binding protein Cin8p (Geiser et al., 1997). Alternatively, Pac10p may have additional, motor protein‐related functions.
GIM deletion mutants are slightly more sensitive towards the actin inhibitor latrunculin‐A, than tcp1‐1 which codes for a mutated subunit of TRiC. Besides having other functions, the latter plays an essential role in actin folding (Ursic and Culbertson, 1991; Chen et al., 1994; Miklos et al., 1994; Vinh and Drubin, 1994). An additional hint of a function for the Gim proteins in actin‐related processes comes from the sensitivity of GIM deletion mutants towards high osmolarity, which is also a common phenotype of actin mutants (Drubin, 1990). Since we did not observe direct defects of the actin cytoskeleton in gim null mutants and since latrunculin‐A is highly specific towards actin (Ayscough et al., 1997), the actin defects in the gim strains are probably only subtle.
A number of results point to specific functions of the Gim proteins with respect to Tub4p. First, all GIM genes genetically interact with TUB4. Second, GIM1/YKE2 shows allele‐specific genetic interactions with SPC98 and SPC97, coding for components of the yeast γ‐tubulin complex. Third, overexpression of TUB4 or co‐overexpression of TUB4, SPC98 and SPC97 is toxic in a gim1/yke2 null strain, while it does not affect growth of wild‐type cells (Spang et al., 1996; Knop and Schiebel, 1997). This suggests that the GIM genes become essential in the presence of a mutated Tub4p protein or after overproduction of Tub4p. Under the latter condition, we could show that the tested Gim1p/Yke2p, Gim2p/Pac10p, Gim3p and Gim5p proteins were associated with Tub4p.
Do all of the Gim proteins have the same function? The very similar phenotypes of the gim null strains as well as complex formation suggest that the Gim proteins have at least overlapping functions. However, we also obtained evidence for specialized functions of the Gim proteins. For example, only Δpac10 cells but not the other Δgim mutants have a spore germination defect (Geiser et al., 1997). In addition, Δgim4 cells have a less severe growth defect and are not as osmotically sensitive as the other GIM deletion mutants. Finally, GIM1/YKE2, GIM2/PAC10 and GIM4 interact differently with SPC98 and SPC97.
The GIM genes encode phylogenetically conserved proteins which may function in folding or assembly processes
How do the Gim proteins function? The cellular localization, the effects of gim null mutants on the actin and tubulin systems as well as the phylogenetic conservation of the Gim proteins are reminiscent of the cytoplasmic chaperonin TRiC. Our localization studies suggests that the Gim proteins, like the TRiC proteins, form cytoplasmic multiprotein complexes. While TRiC contains eight subunits (Kubota et al., 1995; Willison and Horwich, 1996; Nitsch et al., 1997), it is still unclear whether additional Gim proteins exist. The tubulin defects of TRiC mutants and gim null strains are remarkably similar. TRiC mutants are also frequently super‐sensitive towards benomyl, and their microtubules depolymerize at reduced temperatures. In addition, some of the TRiC mutants also do not arrest in the cell cycle, although microtubule structures are impaired (Ursic and Culbertson, 1991; Chen et al., 1994; Miklos et al., 1994). tcp1‐1 was similarly sensitive towards latrunculin‐A as the gim null strains. Finally, a similar phylogenetic conservation as for the Gim proteins has been observed for the subunits of TRiC (Kubota et al., 1995; Nitsch et al., 1997): like TRiC, the Gim proteins have relatives in human, mouse, C.elegans and S.pombe (Figure 1). As for the Gim proteins, only two homologues of TRiC subunits were identified in archaebacteria and none in eubacteria. In addition to this, the Gim proteins as well as the subunits of TRiC bear a stronger homology to their relatives from other species than to each other (Kubota et al., 1995; Waldmann et al., 1995; Nitsch et al., 1997).
The Gim complexes may play a role in the biogenesis of α‐ and γ‐tubulin and probably actin, possibly as a molecular chaperone. A number of our results are consistent with this notion. First, the GIM genes show genetic interactions with components of TRiC, which is expected if both complexes function in related processes. Secondly, the Gim proteins bind to overproduced Tub4p. Thirdly, the reduced level of α‐tubulin may be the consequence of a folding defect. Fourthly, the phenotypes of TRiC and GIM mutants are very similar. The Gim complexes may assist protein folding either before or after TRiC function. They could also bind to folded monomeric Tub4p or actin, thereby reducing potential toxic effects of the monomers. In any case, the Gim proteins only become essential for growth under certain circumstances, such as in the presence of a mutated Tub4p protein or overexpressed Tub4p.
Our results demonstrate that the mammalian homologues of Gim1p/Yke2p and Gim2p/Pac10p, the KE2 and VBP‐1 gene products, are functional in yeast. Their expression complemented the synthetically lethal defects of Δgim1/yke2 or Δgim2/pac10 with tub4‐1 as well as the benomyl super‐sensitivity and cold sensitivity. Interestingly, KE2 mRNA is most abundant in brain and testis (Abe et al., 1988), tissues which have a high content of tubulin. It is expected that these tissues have a high level of proteins that are required for tubulin biogenesis, as is the case for TRiC (Silver et al., 1979, 1987). The homologue of Gim2p/Pac10p, the VBP‐1 protein, has been identified in a two‐hybrid screen using the von Hippel–Lindau tumor suppressor gene as a bait (Tsuchiya et al., 1996). Whether this interaction has any physiological significance remains to be determined. Homologues of Gim3p and Gim5p were also found; however, we have not yet tested whether they function in yeast. The fact that the KE2 and VBP‐1 proteins can substitute for Gim1p/Yke2p and Gim2p/Pac10p in yeast raises the possibility that the homologues of the Gim proteins also form complexes that promote the formation of functional α‐ and γ‐tubulin in mammals and other species.
Materials and methods
Media and general methods
Basic yeast methods and growth media were as described (Guthrie and Fink, 1991). For Gal1‐controlled gene expression, yeast strains were grown in synthetic complete medium (SC) containing raffinose (2%) as carbon source. Glucose (2%) or galactose (2%) was added for the repression or induction of the Gal1 promoter, respectively. Yeast strains were transformed by the lithium acetate method (Schiestl and Gietz, 1989). Escherichia coli strains were transformed by electroporation (Dower et al., 1988). PCR was performed with Vent polymerase (New England Biolabs). Recombinant DNA methodology was performed as reported (Sambrook et al., 1989).
Plasmids and yeast strains
Plasmids and yeast strains used and constructed during the course of this study are listed in Table III. The mouse KE2 gene and the human VBP‐1 gene were amplified by PCR from a mouse and human cDNA library, respectively and the PCR products obtained were cloned into p415‐ADH. Epitope‐tagged versions of GIM genes were generated by recombinant PCR, cloned into pRS315 (Sikorski and Hieter, 1989) and tested for functionality in gim mutants. The coding regions of the intron containing the GIM4 ORF were amplified by PCR from chromosomal yeast DNA and cloned into p415‐ADH (Mumberg et al., 1995). The intron of GIM5 was removed by recombinant PCR. The coding region of GIM5 was cloned into expression vector p415‐ADH. Double deletion mutants of Δgim1/yke2 together with any one of Δgim2/pac10, Δgim3, Δgim4 or Δgim5 were generated by subsequent deletion of the respective ORFs using either kanMX4 (Wach et al., 1994) or HIS3MX6 (Wach et al., 1997) as selectable markers. Chromosomally integrated C‐terminal fusions of the GIM genes with GFP, protein A and 3HA were generated by homologous recombination with PCR‐amplified GIM‐GFP‐kanMX4 (Wach et al., 1997), GIM‐ProA‐kanMX4 and GIM‐3HA‐kanMX4 cassettes (E.Schiebel, unpublished). TUB1 and RBL2 were amplified by PCR and cloned into p415‐ADH and pRS425 (Christianson et al., 1992), respectively.
Isolation of mutants which are synthetically lethal with tub4‐1
In order to identify mutants which are synthetically lethal with tub4‐1, we used the ade2/ade3 red/white colony sectoring system (Koshland et al., 1985; Huffaker et al., 1987). The haploid strain K2348 (ade2 ade3) was transformed with an ADE3/URA3 plasmid containing TUB4 (pSG24). This plasmid had a mutated centromere rendering it unstable under non‐selective growth conditions. The chromosomal TUB4 gene was disrupted by insertion of the HIS3 marker using the Δtub4::HIS3 cassette of plasmid pSM219 (Spang et al., 1996). This strain was transformed with a TRP1‐based plasmid carrying the temperature‐sensitive tub4‐1 allele (pSG23), yielding strain SGY33. The tub4‐1 allele harbours a single amino acid substitution which causes cells to die at 37°C while they stay alive at 30°C (Spang et al., 1996). Strain SGY33 grows with red colony colour, since it is genotypic ade2, resulting in the accumulation of a red intermediate of the adenine biosynthetic pathway. Loss of the ADE3 plasmid pSG24 under non‐selective conditions changes the genotype from ade2 to ade2 ade3, resulting in white or red/white sectoring colonies. Strain SGY33 was mutagenized with methane sulfonic acid ethyl ester (EMS, Sigma) to ∼30% survival. Mutations that conferred synthetic lethality with tub4‐1 were identified on plates without selection for the plasmid pSG24 as red, non‐sectoring colonies, since these cells depend on TUB4 of the ADE3‐containing plasmid. Among 60 000 screened colonies, 12 colonies with this phenotype were obtained. These were unable to grow on plates containing 5‐fluoro‐orotic acid (5‐FOA), a drug that kills cells harbouring a URA3 gene. They all regained the red/white sectoring phenotype and their ability to grow on 5‐FOA when transformed with a LEU2‐based plasmid containing the TUB4 gene but not when transformed with a plasmid carrying tub4‐1. Although all 12 mutants showed a red/white sectoring phenotype when transformed with an SPC98‐containing plasmid, it was shown by crossings to strain SGY88 (SPC98‐LEU2) and subsequent linkage analysis (see below) that the mutation causing synthetic lethality resided only in two cases in SPC98. In contrast, the red/white sectoring phenotype of the other mutants was due to suppression of the tub4‐1 phenotype by SPC98 (Geissler et al., 1996). The mutants were back‐crossed twice to strain SGY24, which revealed that the phenotypes were caused by a single, recessive mutation. Crosses between the mutants identified seven complementation groups.
Identification and disruption of the GIM genes
Two of the mutants from our screen were complemented by SPC97, recently identified as the third component of the γ‐tubulin complex in yeast (Knop et al., 1997). The other eight mutants were transformed with DNA from a LEU2/CEN‐based yeast genomic library (kindly provided by Dr K.Nasmyth), and the transformants were screened for regaining of the red/white sectoring phenotype. The genomic inserts of plasmids allowing cells to lose the TUB4‐containing ADE3/URA3 plasmid after re‐isolation and re‐transformation were sequenced. Subcloning and sequencing identified five genes which we named GIM1–GIM5.
The GIM, SPC98 and SPC97 genes on pRS305 (Sikorski and Hieter, 1989) were targeted to their chromosomal locations in strain SGY24. The resulting strains were then crossed with the corresponding mutants. Random spore analysis of the diploid strains revealed that, for example, GIM1 was linked to the mutation conferring the synthetically lethal phenotype with tub4‐1, since all sectoring haploid progenies did not contain the labelled GIM1‐LEU2. The GIM genes were disrupted by insertion of the heterologous markers kanMX4 (Wach et al., 1994) or HIS3MX6 (Wach et al., 1997), resulting in the deletion of the entire ORFs. The kanMX4 and HIS3MX6 disruption cassettes were generated by PCR and transformed into diploid YPH501 or haploid YPH499 (in the case of GIM2/PAC10). The deletion of the GIM genes in the resulting strains was confirmed by PCR.
Anti‐Gim1p/Yke2p, anti‐Gim2p/Pac10p, anti‐Gim3p, anti‐Gim5p antibodies, immunoblots, immunoprecipitation experiments and immunofluorescence microscopy
The entire coding regions of GIM1/YKE2, GIM2/PAC10, GIM3 and GIM5 were cloned into pGEX expression plasmids (Pharmacia). The GST fusion proteins were expressed and purified by affinity purification using glutathione–Sepharose according to the manufacturer's recommendations. Antibodies against the purified proteins were raised in rabbits as described (Harlow and Lane, 1988). Antibodies against Gim1p/Yke2p and Gim2p/Pac10p were affinity purified as described (Spang et al., 1995). The specificity of the antibodies was demonstrated by performing immunoblots with extracts from GIM and Δgim strains.
For immunoprecipitations of the Gim proteins, cells of strains carrying 3HA‐ or MYC‐tagged derivatives were grown in selective media at 30°C to a density of 2×107 cells/ml. Washed cells were resuspended in H‐buffer [50 mM HEPES, pH 7.5, 100 mM NaCl, 2.5 mM MgCl2, 1 mM EGTA, 5% glycerol, 1 mM phenylmethylsulfonyl fluoride (PMSF), 5 mM benzamidine, 10 μM leupeptin, 20 μg/ml chymostatin and 20 μg/ml E64] and lysed by vortexing with glass beads on ice until >95% of the cells were lysed. Cell debris were removed by centrifugation (5 min, 20 000 g). Cell lysates were incubated with anti‐HA antibodies (12CA5) covalently bound to protein A–Sepharose for 3 h at 4°C. The anti‐HA antibodies were cross‐linked to protein A–Sepharose as reported (Harlow and Lane, 1988). The precipitates were washed twice with H‐buffer, resuspended in HU‐buffer (Knop et al., 1996) and heated for 10 min at 65°C prior to SDS–PAGE (Laemmli, 1970). Gim5p‐ProA was precipitated identically using IgG–Sepharose. As molecular weight standards for SDS–PAGE, we used myosin (200 kDa), β‐galactosidase (116 kDa), phosphorylase B (97.4 kDa), serum albumin (66 kDa), ovalbumin (45 kDa), carbonic anhydrase (31 kDa), trypsin inhibitor (21.5 kDa), lysozyme (14.4 kDa) and aprotinin (6.5 kDa).
For the overexpression of TUB4‐3HA and TUB4 from the Gal1 promoter, yeast cells were grown in SC medium with 2% raffinose as sole carbon source at 30°C to a cell density of 5×106 cells/ml. The Gal1 promoter was induced by the addition of 2% galactose for 6 h. Harvested cells were resuspended in L‐buffer (20 mM Tris–HCl, pH 7.5, 100 mM NaCl, 10 mM EDTA, 1 mM EGTA, 5% glycerol, 1 mM GTP and protease inhibitors) and lysed with glass beads as described above. Cell lysates were incubated with anti‐HA antibodies covalently bound to protein A–Sepharose for 1 h at 4°C.
For the quantifications of α‐tubulin, β‐tubulin and Tub4p, cells were lysed as described by Knop et al. (1996). The primary antibodies used for immunoblots were mouse monoclonal anti‐HA (12CA5, Hiss Diagnostics), mouse monoclonal anti‐MYC (9E10; Boehringer Ingelheim), rabbit anti‐Gim1p/Yke2p, rabbit anti‐Gim2p/Pac10p, rabbit anti‐Gim3p, rabbit anti‐Gim5p, rabbit anti‐Tub4p (Geissler et al., 1996), rabbit anti‐α‐tubulin, rabbit anti‐β‐tubulin (α‐ and β‐tubulin antibodies were a gift of Dr F.Solomon), rabbit anti‐Nop1p (Dr E.Hurt) and rabbit anti‐Fas antibodies (Dr R.Egner). As secondary antibodies, rabbit anti‐mouse or goat anti‐rabbit IgGs coupled to horseradish peroxidase (Jackson Immuno Research Laboratories) were used. The immunoreaction was visualized by an ECL Kit from Amersham.
Immunofluorescence of formaldehyde‐fixed yeast cells was performed as described (Knop et al., 1996) with 1 h fixation time. The primary antibodies were mouse monoclonal anti‐β‐tubulin WA3 (kindly provided by U.Euteneuer‐Schliwa) or mouse monoclonal anti‐actin (Amersham). Secondary antibodies were goat anti‐mouse IgGs coupled to CY3 (Jackson Immuno Research Laboratories). DNA was stained with DAPI (Boehringer Mannheim).
Nocodazole arrest, flow cytometry, determination of cell size and protein determination
For nocodazole arrest, cells were treated with 15 μg/ml nocodazole (Sigma) in liquid medium at 30°C for 3 h. For flow cytometry, cells were prepared as reported (Hutter and Eipel, 1979). The DNA content of 20 000 cells was determined using a flow cytometer (Facscalibur, Becton‐Dickinson). Mean cell volumes were determined using a CASY1 Cell Counter and Analyzer System (Schärfe System GmbH). Protein concentrations were determined according to Bradford (1976).
Size exclusion chromatography
Cell lysates of YPH499 were prepared in H‐buffer as described above. Proteins were separated according to their apparent size using a Smart System from Pharmacia with a Superose 6 column (PC 3.2/30). The column was equilibrated with H‐buffer, and molecular weight determination was carried out according to a calibration curve obtained with standard proteins (thyroglobulin, apoferritin, β‐amylase, alcohol dehydrogenase, bovine serum albumin, ovalbumin, carbonic anhydrase, RNase A, blue dextran; Sigma) separated in H‐buffer. Separation was performed with a flow rate of 15 μl/min. The eluate was collected in fractions of 50 μl each. Each three consecutive fractions were pooled and analysed by immunoblotting.
We are very grateful to Dr A.Lupas for sequence analysis of the Gim proteins and to Dr D.Jenne for the PCR amplification of VBP‐1. Katrin Grein, Elvira Glatt and Tanja Hilsendegen are acknowledged for excellent technical assistance. We would like to thank Dr F.Solomon for the gift of anti‐α‐tubulin and anti‐β‐tubulin antibodies and Drs D.Botstein, K.Nasmyth and T.Stearns for providing us with strains. We thank S.Elliott and Dr M.Knop for carefully reading the manuscript and Dr F.‐U.Hartl for helpful discussions. This work was supported by a grant of the BMBF.
- Copyright © 1998 European Molecular Biology Organization