|Home | About | Journals | Submit | Contact Us | Français|
A previously uncharacterized yeast gene (YER016w) that we have named BIM1 (binding to microtubules) was obtained from a two-hybrid screen of a yeast cDNA library using as bait the entire coding sequence of TUB1 (encoding α-tubulin). Deletion of BIM1 results in a strong bilateral karyogamy defect, hypersensitivity to benomyl, and aberrant spindle behavior, all phenotypes associated with mutations affecting microtubules in yeast, and inviability at extreme temperatures (i.e., ≥37°C or ≤14°C). Overexpression of BIM1 in wild-type cells is lethal. A fusion of Bim1p with green fluorescent protein that complements the bim1Δ phenotypes allows visualization in vivo of both intranuclear spindles and extranuclear microtubules in otherwise wild-type cells. A bim1 deletion displays synthetic lethality with deletion alleles of bik1, num1, and bub3 as well as a limited subset of tub1 conditional-lethal alleles. A systematic study of 51 tub1 alleles suggests a correlation between specific failure to interact with Bim1p in the two-hybrid assay and synthetic lethality with the bim1Δ allele. The sequence of BIM1 shows substantial similarity to sequences from organisms across the evolutionary spectrum. One of the human homologues, EB1, has been reported previously as binding APC, itself a microtubule-binding protein and the product of a gene implicated in the etiology of human colon cancer.
In budding yeast (Saccharomyces cerevisiae) the microtubule cytoskeleton has been implicated in a limited number of cellular functions (for a recent review see (Botstein et al., 1997 )). In addition to the separation of chromosomes during mitosis, only two other functions clearly have been shown to require intact microtubules: movement of the nucleus to the bud neck just prior to separation of the chromosomes and nuclear fusion (karyogamy) following cellular fusion during mating (Huffaker et al., 1988 ). Yeast microtubules are always found attached to a spindle pole embedded in the nuclear membrane (Byers, 1981 ). The intranuclear microtubules project into the nucleus and appear to be responsible for chromosome separation, whereas the extranuclear microtubules have been functionally implicated in both the premitotic nuclear movements and karyogamy (Huffaker et al., 1988 ).
In most eukaryotic cells, the diversity in function of different types of microtubules in the same cell can be attributed to differences in the tubulin isoforms comprising microtubules or to differences in the bound associated proteins, or both. In S. cerevisiae, there is only one gene encoding β-tubulin, and either of the two α-tubulin-encoding genes has repeatedly been shown to suffice for all of the normal microtubule-associated functions (Neff et al., 1983 ; Schatz et al., 1986 , 1988 ). Thus, it appears that the basis for the diversity of functions must lie in the associated proteins and not in the tubulin polymer itself. For this reason, searches for authentic microtubule-binding proteins have been carried out in yeast for many years, with the resulting discovery of a number of such proteins (Meluh and Rose, 1990 ; Barnes et al., 1992 ; Hoyt et al., 1992 ; Roof et al., 1992 ; Pasqualone and Huffaker, 1994 ; Interthal et al., 1995 ; Irminger-Finger et al., 1996 ; Botstein et al., 1997 ).
Among these there are several that decorate microtubules when examined in colocalization experiments. Some of these are nonessential for growth, although their absence does produce a phenotype (Hoyt et al., 1992 ; Roof et al., 1992 ; Interthal et al., 1995 ; Pellman et al., 1995 ). The protein product of the BIK1 gene is a good example: originally identified serendipitously as a karyogamy-defective mutant, the bik1 null mutant also was found to have more subtle defects in spindle morphology (Trueheart et al., 1987 ). However, bik1 mutants display synthetic lethality with tubulin mutations (Berlin et al., 1990 ) as well as mutations in other genes. A particularly interesting genetic characteristic of BIK1 is that overexpression of the gene has a strong phenotype, resulting in the disappearance of microtubule structures and arrest of cell division (Berlin et al., 1990 ). This suggests that the stoichiometry of Bik1p is somehow important and supports a role for Bik1p in microtubule cytoskeleton structure as well as function.
Here, we characterize another gene encoding a microtubule-binding protein that appears to have a structural and functional role in the microtubule cytoskeleton. This gene (called BIM1 for binding to microtubules; the open reading frame is YER016w) emerged from a two-hybrid screen in which TUB1, the major gene encoding yeast α-tubulin, was used as the “bait.” In addition to BIM1, the screen identified TUB2 (encoding β-tubulin) and BIK1. We show that Bim1p colocalizes with both intranuclear and extranuclear microtubules; that deletion mutants are viable but have obvious microtubule phenotypes, including a strong bilateral karyogamy defect and synthetic lethality with tub1 and bik1 mutations; and that overexpression of BIM1 results in a readily scorable microtubule phenotype including cell cycle arrest.
Finally, the sequence of BIM1 is similar to the sequence of human EB1, a putative ligand of APC, the adenomatous polyposis tumor suppressor protein implicated in the etiology of inherited colon cancer (Groden et al., 1991 ). Human EB1 was originally identified in a two-hybrid screen using APC as bait and the homology to the then uncharacterized YER016w open reading frame was noted (Su et al., 1995 ). It is particularly interesting that wild-type (but not mutant) APC has been associated both structurally and functionally with the microtubule cytoskeleton in mammalian systems (Munemitsu et al., 1994 ; Smith et al., 1994 ). In addition, APC localization depends on intact microtubules (Smith et al., 1994 ; Nathke et al., 1996 ). Our findings provide context for these observations.
Yeast strains are listed in Table Table1,1, plasmids in Table Table2.2. Standard methods were used for growth, sporulation, and genetic analysis of yeast (Guthrie and Fink, 1991 ). Strains DBY7830 and DBY7834 are the products of the tub2–201 allele from DBY4869 backcrossed six times to DBY6654 or a similar congenic strain.
DNA cloning was performed using standard methods (Sambrook et al., 1989 ). Oligonucleotide sequences are listed in Table Table3.3. To construct pRB2510, the ACT1 terminator was excised from pTS161 as a BamHI-SphI fragment and inserted into BamHI-SphI sites of pRB1508. To construct pRB2514, TUB1 was amplified by polymerase chain reaction (PCR) from the genomic DNA as a template using Vent polymerase (New England Biolabs, Beverly, MA) and primers TUB1–1 and TUB1–2. The PCR fragment was subsequently digested with NcoI and inserted into the NcoI site of pRB2510. Mutant tub1 alleles were amplified using the same primers (except for tub1–801 which required primer tub1–3, and alleles tub1–851 and tub1–852 which required primers tub1–4 and tub1–5, respectively) and cloned into pRB2510 in duplicate. Plasmid pRB2639 was constructed similarly to pRB2514, except that primers for TUB3 amplification were TUB3–1 and TUB3–2. For pRB2637, the ADE2 gene was excised as a BglII fragment from pASZ10 (Stotz and Linder, 1990 ) and blunt-end ligated into the EcoRV and StuI sites of pJJ244 (Jones and Prakash, 1990 ). To make pRB2654, BIM1, amplified by PCR with Vent polymerase and primers BIM1–1 and BIM1–2, was digested with BamHI and XbaI and cloned into the BamHI and XbaI sites of pRB2138. To construct pRB2652, the same PCR product as for pRB2654 was inserted into the BamHI and XbaI sites of pTS161. pRB2663 was constructed by inserting the BglII fragment from pASZ10 (Stotz and Linder, 1990 ), containing the ADE2 gene, into the BamHI site of pUC19 (Sambrook et al., 1989 ).
Disruptions were constructed by double-fusion PCR (Amberg et al., 1995 b). The bim1Δ::URA3 allele was created using primers bim1Δ-1, bim1Δ-2, bim1Δ-3, and bim1Δ-4. The URA3 marker was amplified by using plasmid pJJ242 (Jones and Prakash, 1990 ) as a template and M13 “forward” and “reverse” primers. Since such deletions are viable (see below), the bim1Δ::URA3 disruption cassette was introduced directly into the haploid DBY7826 (a derivative of DBY6592 that no longer contains pRB326) by transformation, producing DBY7300. For the purpose of genetic analysis, the bim1Δ::ura3::ADE2 allele was created by transforming DBY7300 with the SmaI-PstI fragment from pRB2637, producing DBY7301. Strains DBY7303, DBY7305, and DBY7306 were progeny of DBY7301 mated to DBY6654.
The bik1Δ::ADE2 allele was created using primers bik1Δ-1, bik1Δ-2, bik1Δ-3, and bik1Δ-4. The ADE2 marker was amplified by using plasmid pRB2663 as a template and M13 forward and reverse primers (as above). Since such deletions are viable, the bik1Δ::ADE2 disruption cassette was introduced directly into the haploid DBY6592 by transformation, producing DBY7827.
The num1Δ::URA3 allele was created using primers num1Δ-1, num1Δ-2, num1Δ-3, and num1Δ-4. The URA3 marker was amplified by using plasmid pJJ242 (Jones and Prakash, 1990 ) as a template and M13 forward and reverse primers (as above). Since such deletions are viable, the num1Δ::URA3 disruption cassette was introduced directly into the haploid DBY7826 by transformation, producing DBY7828.
The bub3Δ::ADE2 allele was created using primers bub3Δ-1, bub3Δ-2, bub3Δ-3, and bub3Δ-4. The ADE2 marker was amplified by using plasmid pRB2663 as a template and M13 forward and reverse primers (as above). Since such deletions are viable, the bub3Δ::ADE2 disruption cassette was introduced directly into the haploid DBY6592 by transformation, producing DBY7829.
To construct double mutants between bim1Δ and the charged-to-alanine tub1 alleles (Richards, 1997 ), DBY7301 was crossed to all of the strains listed in Table Table4.4. Diploids were selected using complementing auxotrophic markers and then were sporulated and dissected. In most cases, the strains carried pRB326, containing the wild-type TUB1 gene; therefore, before analyzing the double-mutant phenotype, cells that had lost the plasmid were selected by growth on 5-fluoroorotic acid.
To construct pairwise combinations of the microtubule cytoskeleton mutants (bim1Δ, bik1Δ, num1Δ, bub3Δ, and tub2–201), the following MATa strains: DBY7303, DBY7835, DBY7836, DBY7837, and DBY7834 were crossed in all combinations to the following MATα strains: DBY7301, DBY7831, DBY7832, DBY7833, and DBY7830. In the cases of the bik1Δ, num1Δ, and bub3Δ strains, the parental strains listed were obtained as haploid segregants from a cross between each of the original disruption strains (i.e., DBY7827, DBY7828, and DBY7829, respectively) and DBY6816.
Strains were mated on YPD medium and then zygotes were micromanipulated, each to a distinct spot on the plate, and allowed to form diploid colonies. The diploids were then sporulated and dissected. In the cases where one or both parental strains contained the TUB1 plasmid pRB326, the plasmid was either eliminated before the cross was made or after the spores germinated. In either case, pRB326 was never present when the phenotypes of the double mutants were scored. Double mutants were identified in any of three ways. In some cases, there were two different auxotrophies marking the two mutants (e.g. num1Δ::URA3 bik1::ADE2 double mutants). In some cases, the phenotypes of the two single mutants were both easily identifiable in the double mutant (e.g., tub2–201 bim1Δ mutants acquired the benomyl resistance conferred by tub2–201 and the temperature sensitivity conferred by bim1Δ). Finally, in cases where the parental phenotypes were similar and the auxotrophic markers which marked the two mutants were identical (e.g., bim1Δ bik1Δ double mutants), the double mutants were identified by segregation analysis. For example, in the case of the bim1Δ × bik1Δ cross, the two Ade+ spores in nonparental ditype tetrads were identified as double mutants.
Expression of the GAL4-TUB1 fusion in strain Y190 (transformed with pRB2514) was confirmed by Western blot analysis (Harlow and Lane, 1988 ) using the anti-hemagglutinin epitope tag antibody 12CA5 (BabCo, Berkeley, CA), diluted 1:1000. Total yeast protein preparation was done as described (Yaffe et al., 1985 ). The λYES cDNA library was amplified as described (Amberg et al., 1995 a).
For the screen itself, strain Y190 containing pRB2514 was transformed with library DNA using the lithium acetate method (Ito et al., 1983 ) and plated on minimal medium (SD) + 10 μg/ml adenine and 50 mM 3-amino-1,2,4-triazole (Sigma, St. Louis, MO). The plates were incubated at 25°C for 10 d. In total, 2.6 × 104 transformants were screened. β-Galactosidase activity was assayed as described (Bai and Elledge, 1996 ). Specificity and reproducibility were tested by cotransformation of strain Y190 with library isolates and pRB2514 or pSE1112 (Bai and Elledge, 1996 ).
Since in preliminary tests TUB2 had been found to bind TUB1 in the two-hybrid system, inserts recovered here were screened for the presence of TUB2 by PCR using one primer corresponding to the vector sequence 2-H1 (Amberg et al., 1995 a) and the other a TUB2 internal primer TUB2–1. Double-stranded dideoxy sequencing was performed with the Sequenase reagent kit (United States Biochemical, Cleveland, OH) using the above vector primer 2-H1.
Strain Y190 was cotransformed with tub1 alanine scan alleles fused to the DNA-binding domain of GAL4 in pRB2510 (made from duplicate PCR constructs, as above) and BIM1 or BIK1 fused to the activation domain of GAL4 as isolated from the cDNA library. Transformants were selected on minimal medium lacking tryptophan and leucine and then patched or spotted onto minimal medium containing 10 μg/ml adenine and 50 mM 3-amino-1,2,4-triazole. Positive interaction was confirmed, as above, using the β-galactosidase assay.
Immunofluorescent staining of yeast was performed using a modification of the methods of Kilmartin and Adams (1984) . Cells were fixed in 3.7% formaldehyde in 0.1 M potassium phosphate buffer (pH 6.5) for 60 min at room temperature and washed in 0.1 M potassium phosphate buffer (pH 6.5) and then in 0.1 M potassium phosphate buffer with 1.2 M sorbitol (pH 6.5) and digested with 0.6 mg/ml Zymolyase 100T (ICN ImmunoBiologicals, Costa Mesa, CA). Cells were applied to the wells of multiwell microscope slides coated with 0.1% polylysine (>400,000 molecular weight, Sigma). Subsequent antibody incubations and washes were performed in phosphate-buffered saline (pH 7.4) with 0.5% bovine serum albumin, 0.5% ovalbumin, and 0.5% Tween 20. YOL1/34 (Accurate Chemical & Scientific Corp., Westbury, NY) diluted 1:20 was used as a primary antitubulin antibody; rhodamine-conjugated rabbit anti-rat IgG (Miles-Yeda LTD) diluted 1:500 was used as a secondary antibody.
For immunofluorescence microscopy, imaging and analysis were performed using the DeltaVision Deconvolution System (Applied Precision Incorporated, Issaquah, WA) attached to an Olympus microscope with a Photometrix PXL Cooled CCD Camera. For visualizing the green fluorescent protein (GFP) fusions, exponentially growing cells were immobilized in low-melt agarose as described previously (Doyle and Botstein, 1996 ).
For 4′,6′-diamino-2-phenylindole (DAPI) staining of intracellular DNA, cells were poisoned by the addition of 1:30 volume of 1 M sodium azide to the medium and stained in PBS with 1 μg/ml DAPI for 10 min. Cells were observed and quantitated using a Zeiss axioscope.
For assessment of karyogamy, matings were performed by mixing together 108 cells of each parent grown to exponential phase in YPD. Mixtures were pelleted, resuspended in 0.2 ml of YPD, spread on small YPD plates (35 × 10 mm, Falcon, Lincoln Park, NJ), and incubated for 3 to 4 h at 30°C. Subsequently, cells were washed off the plates and fixed in 3.7% formaldehyde or poisoned with sodium azide for immunofluorescent staining.
Strain DBY6654 transformed with pRB2652 or pTS161 was grown to exponential phase in minimal medium lacking uracil and containing 2% raffinose instead of glucose. Cells were centrifuged and resuspended in the same medium (control) or with 2% galactose instead of the raffinose. Cell density and viability were monitored; aliquots for immunofluorescence were removed at 10.5 h after induction.
The two-hybrid system for detecting protein interactions in vivo (Fields and Sternglanz, 1994 ; Bai and Elledge, 1996 ) depends on the ability of proteins fused to the two essential domains of the GAL4 transcription activator to interact well enough to restore GAL4 function. Interaction is detected by observing the Gal4p-dependent expression of reporter genes (in our case HIS3 and Escherichia coli LacZ) driven by the galactose promoter in the cell.
A two-hybrid screen was performed using the entire TUB1 coding sequence (including the TUB1 intron) fused to the DNA-binding domain of the GAL4 gene; as described in detail above, this construct was made on a CEN plasmid carrying the TRP1 gene. This plasmid (pRB2514) was introduced into a haploid strain Y190, selecting Trp+ transformants. The bait plasmid-bearing strain was then transformed with a cDNA library fused to the GAL4 activation domain (kindly provided by S. Elledge; cf. Amberg et al., 1995 a) carried on a 2-μm plasmid containing the LEU2 gene. Among about 26,000 Leu+Trp+ transformants, 100 were found to be resistant to 50 mM aminotriazole (indicating HIS3 function). Of these, 16 also expressed substantial levels of β-galactosidase. Among these, seven were eliminated by controls (most could activate the LacZ reporter gene independently of the bait plasmid).
The remaining library isolates were screened for the presence of TUB2 as an insert by PCR, using one primer from the gene and one from the vector; two were detected. The remaining candidates were subjected to restriction mapping, and one representative of each restriction pattern was sequenced. Overall, we found 2 fusions to TUB2 (encoding β-tubulin), 1 fusion to BIK1, 5 fusions in-frame to BIM1 (YER016w), and 1 fusion in-frame to the last 20 residues of ADH1 (this was not followed up further). The 5 BIM1 candidates represented 4 instances of a fusion missing the first 70 residues and one instance of a fusion carrying essentially the entire coding sequence, missing only the first 9 residues.
To test whether TUB3 will substitute for TUB1 in binding BIM1 in the two-hybrid system, a bait plasmid (pRB2639) was constructed in which the only difference from pRB2514 was the substitution of the tubulin gene sequences. Plasmid pRB2639 was tested for interaction by cotransformation as above with the longer BIM1 clone. The resulting transformants both grew in the presence of 50 mM aminotriazole and produced β-galactosidase as well as control cotransformants using pRB2514. Thus, Bim1p appears to be able to interact with either yeast α-tubulin.
Sequence alignment, using FASTA (Pearson and Lipman, 1988 ), of the predicted amino acid sequence of BIM1 to several of its close homologues is shown in Figure Figure1. 1. The protein is clearly well conserved over evolutionary time, with good homology (ca. 33–36% identity and ca. 56–61% similarity) to human, mouse and S. pombe (fission yeast). Blast searches of the EST databases (at NCBI, http://www.ncbi.nlm.nih.gov, July 6, 1997 release) revealed high-scoring homologues in Drosophila, Caenorhabditis elegans, zebra fish, and chicken (not shown). The best-characterized human homologue (EB1) was previously isolated in a two-hybrid screen using APC as bait (Su et al., 1995 ). APC has been shown to bind microtubules (Munemitsu et al., 1994 ; Smith et al., 1994 ; Nathke et al., 1996 ); the finding that an APC-binding molecule is homologous to a tubulin-binding molecule of yeast establishes a second connection between APC and microtubules.
The GFP of the jellyfish Aequorea victoria provides a convenient way to study subcellular localization of proteins in living cells (cf. Stearns, 1995 ). A variety of fusions of the BIM1 coding sequence to GFP coding sequence were constructed (see MATERIALS AND METHODS). One of these, in which a mutant with increased fluorescence, S65T, (Heim et al., 1995 ) of GFP is fused at the N terminus of Bim1p and driven by actin promoter, was introduced into wild-type cells and localization observed in a deconvolution fluorescence microscope system (DeltaVision). As the examples in Figure Figure22 illustrate, the GFP-Bim1p fusion protein colocalizes with both intranuclear and extranuclear (note especially panels I and J) microtubules. As shown below, this GFP-Bim1p fusion complements all of the growth phenotypes of bim1Δ mutants. It is possible that these conditions represent a mild overproduction if the ACT1 promoter is significantly stronger than the BIM1 promoter. Nevertheless, this experiment verifies directly that Bim1p localizes to the microtubule cytoskeleton, presumably by binding sites on α-tubulin.
A bim1 mutation that deletes the entire coding sequence was constructed by double-fusion PCR and introduced into a diploid strain. Upon tetrad dissection at 30°C, it was determined that haploid strains bearing the bim1::URA3 allele are viable. Thereafter, the bim1 deletion construct was introduced directly into haploid strains. Though viable, haploid strains with this mutation grow poorly at temperatures below 14°C and fail to grow at 37°C even though the parental strain grows up to 38°C. Significantly, bim1Δ strains fail to grow in concentrations of the antimicrotubule drug benomyl (i.e., 20 μg/ml) to which normal yeast are entirely resistant. These growth phenotypes are illustrated in Figure Figure3.3. Also shown in Figure Figure33 is the result that each of these growth phenotypes is completely complemented by the presence, on a low-copy plasmid, of the aforementioned GFP-Bim1p fusion driven by the ACT1 promoter.
We examined haploid bim1Δ strains stained with antitubulin antibodies and DAPI using immunofluorescence microscopy in a DeltaVision deconvolution system. The images shown in Figure Figure44 are projections of several focal planes, included to show all of the staining, just as one would see if one focused up and down in a conventional fluorescence microscope. The examples of large-budded cells observed in the cultures shown in Figure Figure44 indicate that the spindles in bim1Δ mutants are short and/or misoriented even at permissive temperature (panels C and D show a cell in which the nucleus is dividing within the mother cell body). At 38°C (panels G-L; note the multibudded cell in panels I and J) nuclei appear to be undivided and the spindles are aberrant, being short and asymmetrically located between mother and daughter, as can be seen by comparing to the images of wild-type large budded cells (panels A, B, E, and F). No abnormal phenotype was observed in unbudded or small budded cells.
Quantitation of the nuclear migration and division defects is shown in Table Table5.5. At nominally permissive temperature (i.e., 30°C), we found a significant increase (relative to wild type) in the frequency of improper nuclear migration (second and fifth columns). At low and high temperature the data are similar. There are also defects in nuclear division (columns 4 and 5) and in the frequency of binucleate mothers (column 2) at all temperatures. Despite the high frequency of these defects, viability of bim1Δ mutants after two generation times even at nonpermissive conditions is nearly normal; although after 24 h at the nonpermissive temperature (38°C), viability is decreased significantly (about sevenfold). Thus, we cannot account for the differential viability at extreme temperatures simply by the morphological changes we see.
Failure of nuclear migration is characteristic of failure of extranuclear microtubule function (Huffaker et al., 1988 ). Extranuclear microtubules are also implicated in karyogamy, the fusion of nuclei after mating. Many tubulin mutations exhibit a bilateral (i.e., both parents mutant) karyogamy failure, as do bik1 mutants (Huffaker et al., 1988 ; Berlin et al., 1990 ; Richards, 1997 ). In preliminary tests we found that cells arising from micromanipulated bim1Δ × bim1Δ zygotes were rarely diploid, whereas from normal crosses such cells are regularly diploid. To quantitate the putative karyogamy defect, we carried out a mating experiment in which zygotes (which were equally abundant in all crosses we did) stained with DAPI were examined in the fluorescence microscope (Table (Table6).6). The data in Table Table66 clearly show that both bim1 and bik1 (included as a control) mutants have bilateral karyogamy defects. Indeed, the bim1 defect is even more striking than the bik1 defect by this assay: whereas after 4 h, 80% bik1 × bik1 zygotes had unfused nuclei and 99% of bim1 × bim1 zygotes had unfused nuclei. The data also clearly show that Bim1p function in only one of the parents suffices to allow karyogamy.
Figure Figure55 shows immunofluorescence (DAPI and antitubulin staining) of zygotes that illustrates karyogamy failure at the level of the extranuclear microtubules, as in the tub2 mutants and bik1 mutants (Huffaker et al., 1988 ; Berlin et al., 1990 ). Unlike the extranuclear microtubules extending between the two nuclei in wild-type zygotes, in bim1 × bim1 zygotes the extranuclear microtubules are oriented at random.
To see whether Bim1p might interact stoichiometrically with tubulin (or another component of the microtubule cytoskeleton), we arranged to overexpress BIM1 from the GAL1 promoter on a CEN plasmid; a similar plasmid with no insert was used as a control. Cells growing exponentially in raffinose medium were shifted into galactose medium. Whereas the control cells increased in number more than 10-fold (both cell numbers and viable cells) after 10 h in galactose, the cells overexpressing Bim1p grew only modestly (ca. twofold) in numbers and died, leaving less than 5% viable after 10 h. Despite the limitations imposed by this method (i.e., long induction times and use of a plasmid), the results are clearly indicative of microtubule function.
Figure Figure66 shows the cell cycle distribution in these cultures. It is clear that many of the cells overproducing BIM1 have accumulated with a large bud. Table Table77 shows quantitation documenting that after 10.5 h, virtually all of the large budded cells (ca. half the culture in the strain overproducing BIM1; Figure Figure6)6) are aberrant, in ways suggesting complete failure of nuclear division. These data also reveal impaired nuclear migration in cells overproducing BIM1 similar to that seen in the bim1Δ mutants.
When the cells from this experiment were labeled with antitubulin antibodies and DAPI and examined in the DeltaVision fluorescence microscope, it emerged that the great majority of the large budded cells had arrested growth with an undivided nucleus, no spindle, and occasional long extranuclear microtubules (Figure (Figure7).7). These results, though clearly more extreme, are reminiscent of the results found with bim1Δ mutants and indicate failure of the mitotic spindle.
To carry out the various cellular functions in which microtubules are implicated, the tubulins must have a number of different ligands to which they can bind. It is likely that not all of these ligands bind tubulin in the same way. To study further the interactions between Bim1p and α-tubulin, we carried out two kinds of experiments. In one, we sought to find out whether there is a subset of tub1 mutations that affect Bim1p binding in the two-hybrid system, essentially as done previously for ligands of actin (Amberg et al., 1995 a). In the other, we sought to discover overlapping of functions with individual tub1 mutations by studying patterns of synthetic lethality and/or synthetic phenotype, as done previously for actin ligands (Holtzman et al., 1994 ).
For both these purposes, we used a set of charged-to-alanine scanning mutations made in the TUB1 gene (Richards, 1997 ). In the case of the two-hybrid differential interaction experiments, we replaced TUB1 in the bait plasmid used to isolate BIM1 with 51 mutant alleles (by PCR, in duplicate as described in MATERIALS AND METHODS). These plasmids were each examined to see whether the mutant Tub1p could interact with Bim1p and Bik1p to identify tub1 alleles that showed differential interaction with either of the two ligands. Interaction was scored, as before, by assessment of growth on minimal medium plates supplemented with 50 mM 3-amino-1,2,4-triazole and confirmed by the β-galactosidase assay (an example of the growth data is given in Figure Figure8).8). The results (Table (Table8)8) were that only a limited number of tub1 alleles failed to interact with Bik1p. The in vivo phenotypes of all of these tub1 alleles are very severe; they are either recessive lethal or very growth impaired, suggesting that lack of interaction may be due to gross structural defects in Tub1p.
The results with Bim1p are much more interesting. They recapitulate the differential interactions with Bik1p, but in addition, there are 11 tub1 alleles that fail to interact with Bim1p but do interact with Bik1p. More encouragingly, six of the alleles that fail to interact with Bim1p are in a contiguous stretch on the primary sequence of the Tub1p (residues 393–432); of these, only one fails to interact with Bik1p. Absent a structure for tubulin, this is as good an indication as one could expect for a legitimate ligand-binding site.
In the case of the genetic differential interaction experiments, double mutants were made by crossing a bim1Δ mutant (DBY7301) to each of the viable tub1 mutants. After tetrad dissection, the double mutants were identified by the segregation of auxotrophic markers: one (ADE2) marking the bim1Δ gene and the other (LEU2) tightly linked to the tub1 mutation. Phenotypes were examined after loss of a wild-type TUB1 plasmid which had been maintained to avoid any potential sporulation and/or germination defects resulting from the tub1 mutation. The results (Table (Table8)8) show that only five tub1 alleles are synthetically lethal with bim1Δ. Most significant is the observation that two of the five are among the six contiguous alleles that showed differential interactions in the two-hybrid assay. Furthermore, some of the other alleles in the contiguous stretch show a severe synthetic phenotype which is nearly, but not quite, lethal. These genetic and two-hybrid interaction results strongly reinforce each other, implicating this particular region of Tub1p in binding Bim1p.
To explore further the role of Bim1p in the function of the microtubule cytoskeleton, genetic interactions between BIM1 and genes encoding other components of the microtubule cytoskeleton were examined. A bim1Δ mutant was crossed to a variety of mutants, including tub2–201, num1Δ, bub3Δ, and bik1Δ. These mutants were chosen because of their demonstrated genetic interactions with TUB1, implicating them either directly or indirectly in microtubule function. TUB2 encodes β-tubulin, which is a component of the tubulin heterodimer (Neff et al., 1983 ). Num1p localizes to the mother cell cortex, and num1 mutants interact genetically with tub1 and tub2 mutants (Farkasovsky and Kuntzel, 1995 ). In addition, num1 mutants have defects in nuclear migration (Kormanec et al., 1991 ). Based on the localization of Num1p and the phenotype of num1 mutants, NUM1 is necessary for the proper function of cytoplasmic microtubules. BUB3 encodes a protein which functions in the mitotic spindle assembly checkpoint (Hoyt et al., 1991 ). Finally, as mentioned previously, the product of the BIK1 gene, because of its phenotypes and localization to microtubules, is likely a structural component of the microtubule cytoskeleton (Berlin et al., 1990 ).
Double mutants were made by crossing each of the mutants listed above to a bim1Δ mutant. Two diploids were made for each combination: one using a MATa bim1Δ parent (DBY7303) and one using a MATα bim1Δ parent (DBY7301). In a combined total of 20 tetrads dissected from both crosses, double mutants containing bim1Δ and either num1Δ, bik1Δ, or bub3Δ were never observed. This lack of double mutants is statistically significant in all three cases (Table (Table9). 9). On the other hand, tub2–201 bim1Δ double mutants are viable. Therefore, bim1Δ is synthetically lethal with num1Δ, bik1Δ, and bub3Δ, indicating possible redundancy of functions between Bim1p and these gene products.
All other pairwise combinations of these mutants were also tested for viability in a similar manner. The only other cross from which double mutants failed to be recovered was bik1Δ × bub3Δ; the statistical significance of this lack of viable double mutants is also shown in Table Table9.9. The synthetic lethal interactions among these microtubule cytoskeleton components are diagrammed in Figure Figure9.9. It is notable that while bim1Δ is synthetically lethal with num1Δ, bik1Δ shows no such effects; indeed, the growth of the bik1Δ num1Δ strain is not more impaired under any condition (including growth on benomyl) relative to either of the two parental strains.
All of the data presented above support the identification of the BIM1 gene (YER016w) as a structural component of the microtubule cytoskeleton of S. cerevisiae. Most of the observations closely resemble similar published data on the genes encoding the tubulin subunits themselves as well as bona fide microtubule-associated genes in yeast (Botstein et al., 1997 ). The essential observations include binding in the two-hybrid system, colocalization with microtubules, characteristic phenotypes of the viable null bim1 mutants, overexpression lethality resulting in complete spindle failure, and specific genetic interactions (synthetic lethality) not only with a subset of tub1 alleles but also with a variety of other genes associated with the microtubule cytoskeleton.
Even though the data for a physical interaction between Bim1p and Tub1p is very strong, we cannot absolutely rule out the possibility that the interaction involves other proteins, although there is little precedent for the kind of allele-specific effects we observe in truly indirect interactions. Final proof will no doubt require an in vitro system capable of assessing assembly and/or function of microtubules.
The original two-hybrid screen yielded, in addition to BIM1, two other genes, BIK1 and TUB2. Both of these genes have been studied quite extensively, and mutants in each have features resembling those of bim1 mutants. Notable among these are the failures in karyogamy and nuclear migration, both associated with functional failure or absence of extranuclear microtubules.
The many similarities and few differences between BIM1 and BIK1 are instructive. Among the similarities are viable null phenotype, hypersensitivity to benomyl, karyogamy failure, nuclear migration, and spindle defects, and, most important, overexpression lethality resulting in spindle failure. The last observation, in particular, is significant in that it suggests, for Bim1p as it did for Bik1p, that it is “required stoichiometrically for the formation or stabilization of microtubules” (Berlin et al., 1990 ); see also Rose and Fink, 1987 ). Indeed, the observation of synthetic lethality between bim1 and bik1 null mutants supports their role in a similar or shared function.
The differences between the bim1 and bik1 null phenotypes may hold some clues as to some differentiation in function. First, the karyogamy failure of bim1 mutants in our hands is more severe than contemporaneous bik1 controls. Second, the BIM1 overexpression effects on nuclear migration are more severe than those reported for BIK1 (Berlin et al., 1990 ). Since nuclear migration and karyogamy both are effected by extranuclear microtubules, it is tempting to speculate that the two genes have largely overlapping functions, but that BIM1 is more critical for extranuclear microtubule function. In support of this idea is the pattern of synthetic lethality: bim1 null mutations show synthetic lethality with num1 null mutations whereas the bik1 null mutation does not. The NUM1 function is clearly associated with extranuclear and not intranuclear microtubule function (Farkasovsky and Kuntzel, 1995 ). Nevertheless, BIM1 seems also to be involved with spindle function, given the decreased number of long spindles in the bim1Δ mutant and the synthetic lethality between bim1Δ and bub3Δ, a gene that functions in the spindle assembly checkpoint.
We describe an attempt to define, within the Tub1p sequence, the regions important for interactions between α-tubulin and Bim1p and Bik1p. Although in the case of Bik1p, we could conclude little, as we found few specific interactions, the case of Bim1p was much more encouraging. Indeed our results predict that a region near the C terminus of Tub1p is the locus of interaction between it and Bim1p. We look forward to the time that molecular structures of microtubules become available so that this kind of prediction can be validated, as was the case for similar experiments with yeast actin (Amberg et al., 1995 a).
The interpretation of the observation that alanine-scanning mutations in similar regions of Tub1p display both synthetic lethality and differential interaction in the two-hybrid system is not straightforward. Using the reasoning of Holtzman et al. (1994) , mutations that fail to interact with a ligand should not be exacerbated by total loss of the otherwise dispensable ligand. We are obliged, therefore, to propose instead that the differential interactions are not a sign of complete loss of binding affinity. Instead, we imagine that each of the many mutations that we can detect as different in the assay provide only a fraction of the binding energy, so that only an ensemble of many alanine-scanning alleles would result in complete failure of binding in vivo. Furthermore, synthetic lethality between these tub1 alleles and bim1Δ implies redundancy of function; one possible explanation is that both Bim1p and this subset of tub1 alleles cooperate to recruit another essential ligand to microtubules.
Finally, we return to the homology between Bim1p and its mammalian homologues. Human EB1 protein, the nearest homologue to Bim1p, was recovered in a two-hybrid screen as a ligand of APC, the gene responsible for a hereditary predisposition to a form of colon cancer (Groden et al., 1991 ; Su et al., 1995 ). It is APC itself (which has no close homologue in the yeast genome), and not EB1, that has been shown to bind microtubules (Munemitsu et al., 1994 ; Smith et al., 1994 ). EB1 has not, to our knowledge, yet been tested for microtubule binding. Our results provide a second line of evidence linking APC and microtubules, since our data suggest that EB1 itself may well bind to microtubules. The fact that there is no close APC homologue in yeast suggests that it is EB1/Bim1p, and not APC, that is the highly conserved structural component common to all eukaryotic genera (supported by the widespread occurrence of Bim1p homologues in a variety of diverse organisms).
To conclude, we have found that yeast BIM1 encodes a protein likely to be a structural component of the yeast microtubule cytoskeleton. The similarity to human EB1, along with the additional microtubule association via APC, suggests that EB1 is likewise a structural component of the mammalian microtubule cytoskeleton, thus providing a framework for the understanding of the functions of microtubules in both systems.
We thank Steve Elledge for λYES cDNA library; Tim Stearns for useful discussions and advice; Koustubh Ranade, Craig Cummings, and Tracy Ferea for critical reading of the manuscript; and Susan Palmieri and Chris Kenfield for assistance with DeltaVision deconvolution system. This work was supported by a grant from the National Institutes of Health (GM-46406).