A Centromere DNA-binding Protein from Fission Yeast Affects Chromosome Segregation and Has Homology to Human CENP-B

The Escherichia coli strain used in this study is DH5α (recA1; GIBCO BRL, Gaithersburg, MD). Procedures for E. coli transformations, cloning techniques, and plasmid isolations were as previously described (Maniatis et al., 1982). The S. pombe strains used in this study were Sp223 (h⁻ ade6.216 leu1.32 ura4.294, a gift from D. Beach, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY); AL91 (h⁹⁰ ura4.D18 swi6::ura4⁺ ade6.M216; Lorentz et al., 1994), EG388 (h⁹⁰ rik1-304 leu1 ura4.D18; Egel et al., 1989), KE108 (h⁹⁰ clr4.55 ura4 ade6.M216; Ekwall and Ruusala, 1994), SBP120390 (h⁻ ade6.704 leu1.32 ura4.294), SBP32590/pSp(cen1)-7Lsup3E (h⁻ ade6.704 leu1.32 ura4.294/ura4⁺ sup3E URA3 TRP1), SBP070196-F5B (h⁺ ade6.210 leu1.32 ura4.D18), SBP070196-A7C (h⁻ ade6.216 leu1.32 ura4.D18), SBP070196-F8B (h⁺ ade6.210 leu1.32 ura4.D18 abp1::ura4⁺), SBP070196-F8C, -F5C (h⁻ ade6.216 leu1.32 ura4.D18 abp1::ura4⁺), SBP082996/pSp(cen1)-7L-sup3E (ade6.704 leu1.32 ura4.294 abp1::ura4⁺/ura4⁺ sup3E URA3 TRP1), and SBP051596 (h⁺/h⁻ ade6.210/ ade6.216 leu1.32/leu1.32 ura4.D18/ura4.D18). The diploid strain SBP051596 was generated in this laboratory by crossing strains ED666 (h⁻ ade6.210 leu1.32 ura4.D18) and ED667 (h⁺ ade6.216 leu1.32 ura4.D18, both gifts from R. Allshire, Medical Research Council Human Genetics Unit, Edinburgh, Scotland, UK). Growth media and conditions were as described (Gutz et al., 1974; Moreno et al., 1991).

DNA transformations and fragment-mediated transplacements were performed with an alkali–cation yeast transformation kit (BIO 101, Inc., Vista, CA). The sup3E-marked cen1 linear minichromosome pSp(cen1)- 7L-sup3E was created by fragment-mediated transplacement of sup3E into pSp(cen1)-7L (Hahnenberger et al., 1989) adjacent to ura4.

For disruption of the abp1 gene, a 1.8-kb HindIII–BamHI fragment containing the entire abp1 coding region (Fig. 1 A) was subcloned into pBluescript KS-. The 1.3-kb BglII–EcoRI region was dropped out of this fragment, leaving 95 and 165 bp of the NH₂- and COOH-terminal coding regions of abp1, respectively. In its place, a 1.8-kb fragment containing the S. pombe ura4 gene was end-filled with the Klenow fragment of DNA polymerase I and inserted by blunt end ligation. Double digestion of this plasmid with HindIII and BamHI produced a 2.3-kb fragment containing the S. pombe ura4 gene and flanking sequences homologous to the abp1 region of the genome. This linear fragment was used for fragment-mediated transplacement (Rothstein, 1983) in the S. pombe diploid strain SBP051596. Stable Ura⁺ transformants were analyzed by Southern blot hybridization for heterozygous disruption of abp1. Diploids heterozygous for abp1 were sporulated and dissected to separate meiotic spores using a dissecting microscope (MSM System, series 200; Singer Instruments, Somerset, UK).

The pSp200 vector, a pBR322-based plasmid, contains S. pombe ars1 on a 1.2-kb EcoRI–EcoRI fragment and S. cerevisiae LEU2 on a 2.2-kb XhoI– SalI fragment cloned into the EcoRI and SalI sites, respectively, of pBR322. S. pombe strain Sp223 was transformed with pSp200 and grown under leucine selection. Genomic DNA was prepared from six Leu⁺ transformants, digested with SalI to linearize pSp200, fractionated by FIGE (field inversion gel electrophoresis), and subjected to Southern blot analysis as previously described (Steiner and Clarke, 1994), using a ³²P-labeled ars1 probe. Exposed film was scanned with a laser densitometer (model LKB 2222-010 UltroScan XL; Bromma, Sweden). On average, the plasmid band was observed to be ∼20 times more intense than the genomic ars1 band.

Total S. pombe DNA was isolated by the method of Beach and Klar (1984) and manipulated with cutoff micropipette tips. Approximately 30 μg of isolated DNA was partially digested with 1 U of Sau3AI for 10 min at 37°C to obtain an average size of 10 kb. Sucrose gradient size–selected DNA was cloned into the BamHI site of pSp200 at a 2:1 weight ratio of insert to vector, respectively, and transformed into E. coli. Of 18 random transformants tested, all had inserts with an average size of 6.8 kb and a range of 1 to 18 kb. Overall, 25,240 transformants representing ∼12 S. pombe genome equivalents were obtained.

The S. pombe library was transformed into SBP120390/pSp(cen1)-7Lsup3E, and Leu⁺ transformants were selected on SD (0.67% yeast nitrogen base without amino acids, 2% glucose) + adenine (10 mg/liter) + uracil (50 mg/liter). Under these conditions, loss of the cen1 minichromosome results in red colony color or red sectoring in a white background (Baum et al., 1994). Of the ∼32,500 colonies obtained, 562 colonies exhibited a sectored phenotype. These isolates were subjected to successive rounds of restreaking on colony color indicator media. Plasmid DNAs from 26 isolates that sectored reproducibly were rescued into E. coli and retransformed into S. pombe. Of these, 25 continued to produce sectoring above background levels and were grouped into classes based on the level of sectoring observed.

The mitotic stability of the pSp(cen1)-7L-sup3E minichromosome was determined in wild-type cells transformed with vector pSp200 or pSp200- abp1 by colony color assay or replica plating, respectively, and measured in abp1 null cells by colony color assay (Clarke and Baum, 1990; Baum et al., 1994).

DNA fragments examined for bandshifts were small (<400 bp) restriction fragments from cc2, as indicated in Fig. 3 A. The gel mobility shift assay was performed as previously described (Baum et al., 1994). End-labeled fragments (5 fmol) were mixed with protein extract and incubated for 20 min at room temperature in binding buffer containing 10 mM Hepes (pH 8.0), 6 mM MgCl₂, 2 mM NaF, 100 mM KCl, 0.5 mM DTT, 10% glycerol, and 1–2 μg poly(dI-dC) (∼2,000-fold excess [wt/wt]; Amersham/USB, Cleveland, OH). Binding specificity was determined by adding 125 fmol (25-fold excess) of the unlabeled DNA fragment as competitor in the binding reaction.

Supershift assays were performed with polyclonal antibody to Abp1p (kindly provided by J.-K. Lee, J.P. Sanchez, and J. Hurwitz, Memorial Sloan-Kettering Cancer Center, New York, NY; Murakami et al., 1996). A 0.4-M KCl chromatin extract or an affinity-purified protein fraction was preincubated in binding buffer with antibody for 20 min at room temperature. Labeled DNA fragment was added, and the mix was incubated at room temperature for an additional 20 min, followed by electrophoresis.

A DNA fragment containing a 57-bp region of fragment c2-10 was amplified by PCR using the following primers: 5′(BglII)-gaagatcTAACTCGACTTTATAAAATTT-3′ and 5′(BamHI)-gaggatcCAAATTAAACATTGCTAATAA-3′. Digestion with BamHI and BglII yielded a 61-bp fragment that was concatamerized at 15°C with T4 DNA ligase, redigested with BamHI and BglII, and cloned into the BamHI site of pBluescript KNS- (Baum et al., 1994). [c2-10₆₁]₄, a clone containing four tandem copies of the insert, was identified and sequenced (Sequenase kit; Amersham/ USB). The [c2-10₆₁]₄ construct was excised as a 266-bp fragment with BamHI and EcoRI. The fragment was biotinylated at the EcoRI end with biotin-16-dUTP (Sigma Chemical Co., St. Louis, MO) and coupled to streptavidin–agarose (GIBCO BRL) as described previously (Lechner and Carbon, 1991). A small aliquot of biotinylated DNA that was ³²P end- labeled at the BamHI site was included in the reaction to determine the coupling efficiency to the column. Approximately 75% of the DNA fragments remained bound to the column. The columns were prepared, run, and stored as described by Lechner and Carbon (1991).

The following buffers were used: B (50 mM KPO₄ [pH 7.0], 100 mM β-glycerophosphate, 10 mM NaF, 10 mM EDTA, 10 mM EGTA, 0.5 M DTT, 1 mM PMSF, 2.5 μg/ml leupeptin, and 2.5 μg/ml pepstatin A); D (10 mM Hepes [pH 8.0], 20% glycerol, 6 mM MgCl₂, 2 mM NaF, 1 mM DTT, 0.2 mM PMSF, 0.5 μg/ml leupeptin, and 0.5 μg/ml pepstatin A); E (10 mM Hepes [pH 8.0], 20% glycerol, 2 mM NaF, 1 mM DTT, and 0.5 M KCl). 4 liters of the overexpressing strain carrying pSp200-abp1 were grown to mid-log phase (A₆₀₀ ∼0.6) in SD + A + U. Freshly harvested cell paste (13 g) was washed with water, and the washed pellet was packed into a 60-ml syringe and extruded into a beaker of liquid nitrogen. Cells were mechanically disrupted in liquid nitrogen (Sorger et al., 1989) using a blender at high speed for 20 min. All subsequent steps were performed at 4°C and as previously described (Lechner and Carbon, 1991) with the following modifications. The freeze-dried powder was resuspended in 24 ml of buffer B and stirred gently for 30 min. After centrifugation (14,000 g for 30 min), the supernatant (low salt extract) was flash frozen in small aliquots and stored at −70°C, and the pellet (chromatin) was further extracted. The pellet was resuspended in 20 ml of buffer B, KCl was added to 400 mM, and the mixture was incubated for 30 min. After centrifugation (34,000 g for 20 min), the soluble material (0.4 M KCl chromatin extract) was further clarified (100,000 g for 1 h). 17 ml of the supernatant was mixed with 28 ml 2× buffer D and 0.15 mg/ml poly(dI-dC) in a final volume of 56 ml to yield 120 mM KCl. Insoluble material was removed by centrifugation (20,000 g for 30 min followed by 100,000 g for 30 min), and the supernatant was passed through a [c2-10₆₁]₄ DNA affinity column (1 ml, described above, equilibrated in buffer D, 120 mM KCl). The column was washed with 15 ml buffer D, 150 mM KCl, and eluted with six 0.5-ml aliquots of buffer E. Elution fractions were flash frozen and stored at −70°C.

S. pombe crude extract or affinity-purified fractions were electrophoresed on a 10% SDS–polyacrylamide gel in 25 mM Tris, 190 mM glycine, and 0.1% SDS (pH 8.3) at 12.5 V/cm for 3 h. Proteins were electroblotted onto nitrocellulose in 25 mM Tris (pH 8.3), 192 mM glycine, and 20% methanol. Membranes were blocked in 2% milk/TBS-T (20 mM Tris, 137 mM NaCl [pH 7.6], and 0.1% Tween 20) overnight at 4°C followed by two rinses with TBS-T. Polyclonal antibody to Abp1p (1:1,000) was incubated with the blocked membrane in 5% milk/TBS-T at room temperature for 1 h followed by three 10 min washes with 2% milk/TBS-T. Goat anti–rabbit IgG–alkaline phosphatase secondary antibody (1:1,000; Bio-Rad Laboratories, Richmond, CA) was incubated with the membrane in 5% milk/ TBS-T at room temperature for 1 h followed by three 15-min washes with TBS-T. Bands were detected with NBT/BCIP (Promega Corp., Madison, WI) (Harlow and Lane, 1988).

A number of multicomponent structures, including the centromere/kinetochore, the DNA replication complex, the mitotic chromosomes, and the mitotic spindle, are likely to be made up of interacting proteins that must be present in precise stoichiometry to yield a functional complex. Overexpression of one component may disturb the stoichiometric relationship and cause chromosome instability because of improper replication or segregation of chromosomal DNA (Meeks-Wagner and Hartwell, 1986). To identify gene products with potential roles in chromosome segregation, we screened an S. pombe genomic library, constructed in the multicopy vector pSp200 (Fig. 1,A), for genes that in elevated dosage cause high loss rates of a stable cen1 linear minichromosome. The library was used to transform strain SBP32590/pSp(cen1)-7L-sup3E. Because of an opal mutation in the ade6 gene, S. pombe strain SBP32590 accumulates a red pigment when grown on media containing low concentrations of adenine and the colony color is red. When the strain also carries the in vitro–constructed minichromosome pSp(cen1)-7L-sup3E, containing a tRNA nonsense suppressor of ade6 (Fig. 1,B), colony color is white. Chromosome loss events, due to either missegregation or replication defects, result in red-sectored colonies (Hofer et al., 1979; Kohli et al., 1989). Approximately 32,500 transformant colonies representing 16 genome equivalents were screened for colonies containing readily discernible red sectors. Roughly 0.07% of the transformed colonies reproducibly displayed increased red sectoring and were grouped into three classes: 9 isolates exhibited very high levels of sectoring in every progeny colony (class III; Fig. 1 C); 5 others showed high levels of sectoring in almost every progeny colony (class II); and 11 isolates exhibited sectoring above background levels, although >75% of progeny colonies were white (class I).

In this screen, elevated numbers of centromeric DNA binding sites for critical proteins might also cause minichromosome instability by sequestering a centromere protein from the kinetochore complex. Thus, plasmid DNAs from each class were first screened by Southern blot hybridization for the presence of DNA sequences derived from centromeric central core and K-repeat DNA elements; however, no significant hybridization was detected (data not shown).

One class III sequence, isolated three times, was localized to a 3.1-kb BamHI fragment (Fig. 1,A). A BamHI and HindIII double digest was used to separate this region into two fragments encompassing 1.3 and 1.8 kb. The presence of the 1.8-kb fragment in the pSp200 vector caused a high rate of minichromosome loss in SBP32590/pSp(cen1)-7Lsup3E, although not as severe as that caused by the original 3.1-kb fragment. On the other hand, the 1.3-kb fragment had a negligible effect on minichromosome stability. The nucleotide sequence of the entire 3.1-kb fragment was determined. In the 1.8-kb region, analysis of the sequence revealed a 1.6-kb open reading frame (Fig. 1,A, shaded box) specifying a protein with significant homology to the mammalian centromere DNA-binding protein CENP-B (see below) and identical to S. pombe Abp1p (Murakami et al., 1996). Further characterization of abp1 was carried out on cells transformed with pSp200-abp1, the pSp200 vector containing the entire 3.1-kb BamHI fragment. In cells newly transformed with pSp200-abp1, the loss frequency of the pSp(cen1)-7L-sup3E minichromosome was 200-fold higher than in transformed cells carrying the vector alone (Table I).

A different class III destabilizing DNA segment was localized to a 2.4-kb BamHI fragment, sequenced, and found to encode S. pombe Nim1p, a negative regulator of the S. pombe mitotic inhibitor Wee1p (Russell and Nurse, 1987a,b). Increased expression of nim1 induces premature mitotic entry, disrupting the temporal ordering of mitotic events and resulting in elevated chromosome loss.

To determine if Abp1p is essential for cell viability, one entire chromosomal copy of the abp1 gene was disrupted by ura4 gene replacement in S. pombe diploid strain SBP051596 (Materials and Methods; Rothstein, 1983). Subsequent sporulation of the heterozygous diploid yielded mostly asci with four viable spores, indicating that Abp1p is not essential for mitotic growth. However, Δabp1 cells grow more slowly than wild-type cells, and this growth difference is accentuated at low temperatures (Fig. 2 A). Similarly, cells containing pSp200-abp1 grow more slowly than cells with vector pSp200, suggesting that alterations in the normal level of abp1 expression affect mitotic growth.

Flow cytometry profiles of Δabp1 cells grown at 30 and 17°C were indistinguishable from wild-type profiles and never revealed an increased population of cells with a DNA content less than 2N, indicating that the slow growth phenotype was not caused by a replication defect (data not shown). In contrast, wild-type cells transformed with pSp200-abp1 and Δabp1 cells lost the cen1 minichromosome at a much higher rate than did wild-type cells, suggesting that slowed cellular growth might result from a defect in chromosome segregation (Table I). Because diploid S. pombe strains are unstable, it is difficult to measure directly the loss frequencies of native chromosomes. However, if native chromosomes are lost at a frequency similar to that of the cen1 minichromosome, this loss would easily account for slower growth rates observed in abp1 null and overexpressing cultures at 30°C. Because aneuploid cells fail to divide, the effect observed on a population of cells stained with DAPI would be minimal, and the population as a whole might look relatively normal. Consistent with this, DAPI staining patterns of Δabp1 cells and pSp200- abp1–transformed cells grown at 30°C are not markedly different from those of wild-type cells grown under similar conditions.

DAPI stains of wild-type cells and pSp200 vector-transformed cells grown at 17°C revealed a majority of normal interphase cells with a single nucleus and a lesser number of dividing cells with two nuclei and a septum (Fig. 2 B and data not shown). In contrast, Δabp1 cells and pSp200- abp1–transformed wild-type cells grown at 17°C displayed a number of morphological abnormalities. Both Δabp1 cells and pSp200-abp1 transformants are generally more rounded and irregularly shaped than wild-type and pSp200-transformed cells. A majority of Δabp1 cells and pSp200-abp1–transformed cells had a single, centrally located DNA mass.

Some additional unusual phenotypes are observed in Δabp1 cells. Approximately 1% of Δabp1 cells grown at 17°C were highly elongated (Fig. 2, C and D). Most lengthened Δabp1 cells had a single mass of DNA stretched to varying degrees across the center of the cell (Fig. 2,C, arrows); others displayed two chromatin masses positioned at opposite ends of the cell plus additional, apparently unsegregated, DNA at more central positions (Fig. 2,D, arrows). Generally, even in smaller cells, the DAPI-staining DNA mass is less compact in Δabp1 cells than in wild-type cells (Fig. 2, C and D). These phenotypes suggest that mitosis occurs abnormally in some Δabp1 and pSp200-abp1– transformed cells grown at low temperature.

To ascertain whether Abp1p is essential for meiosis, several +/Δabp1 heterozygous diploids were sporulated and the asci dissected. Haploid progeny were tested for mating type and abp1 disruption by Southern blotting. Wild-type (Abp1⁺) cells of opposite mating types were crossed and appeared to undergo meiosis normally, forming large numbers of asci (Fig. 2,E, inset). In contrast, when Δabp1 sister spores of opposite mating types are induced to undergo conjugation and sporulation, relatively normal asci are extremely rare; 0–0.7% of cells were observed to form asci out of 2,000–6,000 cells examined for each of 10 separate matings. In addition, aberrant asci-like forms were observed that contained unusually shaped spores (Fig. 2 E, arrows). The meiotic defect is recessive, because these same Δabp1 cells were able to conjugate and sporulate normally in crosses with wild-type cells, although fewer asci were observed than in crosses between wild-type cells. Out of 17 asci dissected from a Δabp1 homozygous mating, one ascus gave rise to four viable spores, and several others gave rise to one to three viable spores. Of the spores dissected, most that appeared normal were viable, whereas those with abnormal morphology failed to produce colonies. Therefore, Abp1p appears to play an essential role in the meiotic process, although rare competent spores are occasionally observed.

In a separate approach to identifying proteins with roles in centromere function, we pursued purification of centromere DNA-binding proteins using sequence-specific affinity chromatography. We demonstrated previously that half of the cen2 central core (cc2), along with K-type repeat sequences, are sufficient to establish a functional, although somewhat compromised, S. pombe centromere on a circular minichromosome (Baum et al., 1994). Such a construct is ∼30-fold less stable in mitosis than a circular minichromosome carrying the entire cen2 region and segregates to two of the four meiotic products as expected, but it exhibits precocious sister chromatid separation in meiosis I. Using convenient restriction sites, this half of cc2 was cleaved into fragments <400 bp in length (Fig. 3,A), and each fragment was tested individually for sequence-specific protein binding activity in gel mobility shift assays (Garner and Revzin, 1981). A 0.4-M KCl extract of chromatin from the wild-type S. pombe strain SBP120390 was found to contain specific binding activity for five fragments from the left half of cc2: c2-1, c2-2, c2-8, c2-10, and c2-11 (Fig. 3, A and B, and data not shown). Based on total binding units, fragment c2-10 binding activity was found in equal proportions in the low salt and 0.4-M KCl extracts, which represent non–chromatin- and chromatin-associated proteins, respectively (Fig. 3,B, lanes 2 and 4, and Table II). Because fewer proteins are in the chromatin fraction, the specific activity was enriched >10-fold in the chromatin extract compared with the low-salt extract.

In an effort to identify centromere-binding proteins associated with those cc2 fragments that exhibit specific gel mobility shifts, we screened extracts prepared from strains containing minichromosome-destabilizing plasmids identified in the genetic screen reported above and from strains containing mutations that derepress expression of a reporter gene embedded in the heterochromatin-like centromeric DNA or that cause an elevated rate of chromosome loss (Allshire et al., 1995). To date, three proteins, Swi6p, Rik1p, and Clr4p, appear to play a role at the centromere (Allshire et al., 1995; Ekwall et al., 1995) as well as at the silent mating type and other loci in fission yeast (Egel et al., 1989; Ekwall and Ruusula, 1994; Lorentz et al., 1994). Mutations in rik1 and clr4 have been shown to alleviate repression of a reporter gene inserted within the centromeric repeated DNA but do not affect repression of genes inserted within the central core DNA (Allshire et al., 1995). Gel shift assays performed with cc2 fragments and extracts prepared from a swi6 null strain and from strains with mutations in rik1 and clr4 did not exhibit any alteration of the gel mobility shift profile compared with extract prepared from a wild-type strain (data not shown). On the other hand, for three of the cc2 fragments, c2-8, c2-10, and c2-11, specific DNA-binding activity was elevated in an Abp1poverexpressing strain (Fig. 3,B, lanes 3 and 5). An increase in binding activity was apparent in both the low-salt and 0.4-M KCl chromatin extract from whole cells. The majority of the overproduced protein was not chromatin associated in vivo, although excess Abp1p appears to favor increased formation of the DNA–protein complex (Fig. 3,B, lanes 3, and Table II). Compared with a wild-type strain, however, c2-10 binding activity that is chromatin associated in vivo was elevated less than twofold in an Abp1p-overexpressing strain, as judged by specific activity (Table II). This increase may be accounted for by saturation of Abp1p binding sites and/or inappropriate binding to lower affinity sites.

Using bandshift assays, the location of a DNA binding site within a given labeled DNA fragment can be approximated by competition with excess unlabeled smaller fragments that step across the region. In the case of fragment c2-10, digestion of the full-length EcoRI–MunI fragment with AflII yields two fragments that when mixed together did not compete with the full-length fragment for binding (Fig. 3,A). In contrast, digestion of the full-length fragment with ApoI yields three fragments, two of which did not compete for binding, but one, a 49-bp ApoI–ApoI fragment containing the AflII site, did compete. Further competition studies suggest that the binding site is located in a 35-bp region between the ApoI and BsrDI sites flanking AflII (Fig. 3,A, bottom). This region is 80% A+T, contains two overlapping short regions of dyad symmetry (Fig. 3,A, arrows), and shares very limited homology (five of nine essential bases; Fig. 3,A, CCC and AA) with the canonical mammalian CENP-B box (see Fig. 3 A legend; Masumoto et al., 1989, 1993). Moreover, AflII cleavage, which inactivates the c2-10 binding site, occurs between these conserved bases, leaving CCC and AA on separate DNA fragments.

Tandem copies of a 61-bp PCR-amplified DNA fragment, c2-10₆₁, centered around the AflII site, were used for DNA affinity chromatography (see Materials and Methods and Table III). A 0.4-M KCl chromatin extract in binding buffer was applied to the DNA affinity column in the presence of excess poly(dI-dC) to reduce nonspecific binding. After a wash, bound protein was eluted with 0.5 M KCl. The peak of c2-10 binding activity contained a single predominant protein with an apparent molecular mass of 62 to 63 kD (Fig. 4 A, elution fraction 3).

Western blot analysis of affinity-purified elution fractions and crude extract indicates that Abp1p is a component of the affinity-purified p62 band. The polyclonal antibody to Abp1p recognized a single 62-kD protein in extracts from wild-type or Abp1p-overproducing strains (Fig. 4,B, lanes 1 and 2) and in the eluted fractions from the affinity chromatography (Fig. 4,B, lanes 3 and 4). In contrast, the cross-reacting protein was not detectable in extract prepared from a Δabp1 strain (null; Fig. 4,B, last lane), indicating that the antibody is specific for Abp1p. If polyclonal antibody to Abp1p was added to gel mobility shift assays with fragment c2-10 and the affinity-purified protein or a 0.4-M KCl extract of chromatin prepared from wild-type or Abp1p-overexpressing strains, a supershift was observed (Fig. 4,C), indicating the presence of Abp1p in the shifted complex. Identical supershifted banding patterns were obtained with crude extracts from a wild-type strain and affinity-purified protein from an Abp1p-overproducing strain. In addition, a bandshift was not observed with extract from an abp1 null strain (Fig. 4 D). Therefore, we conclude that Abp1p is a component of the protein complex assembled on the c2-10 DNA fragment in vitro and that formation of the DNA–protein complex is Abp1p dependent. Furthermore, it is quite likely that the complex contains only Abp1p, because a single protein band is observed in the affinity-purified preparations. Supershift analysis of central core fragments c2-8 and c2-11 indicates that they are also bound by Abp1p (data not shown). Preliminary results from gel filtration chromatography of chromatin extracts indicate that a 175-kD protein complex binds to central core fragment c2-8 and a 125-kD complex binds to c2-10 and c2-11, suggesting that the c2-10 binding complex may represent a homodimer or heterodimer of Abp1p (Baum, M., and L. Clarke, unpublished results; Ngan, V.K., and L. Clarke, manuscript submitted for publication).

Sequence comparisons of Abp1p binding sites in cc2 imply that the binding motif is almost certainly AT rich but are otherwise uninformative. Bandshift competition assays in which a ³²P-labeled c2-10 fragment was competed with a 25-fold molar excess of an unlabeled native or mutationally altered sequence (Fig. 3,A, AAA substitutes for CCC) suggest that the binding sequence does not consist entirely of A and T. Competition with excess cold c2-10 DNA reduced binding to the radiolabeled native c2-10 fragment by 90%, whereas competition with the altered fragment resulted in only a 60% reduction in binding to the radiolabeled native c2-10 fragment. A competitor DNA sequence with a mutation in a critical residue would be expected to reduce binding of Abp1p to the ³²P-labeled c2-10 fragment by <10%. Thus, the intermediate value observed for the CCC to AAA alteration is a reflection of reduced binding affinity, but it also indicates that these residues are not essential for binding. Consistent with this finding, the motif shared between c2-10 and the CENP-B box does not appear to be conserved in other centromeric DNA fragments that bind Abp1p. Sequence comparisons between the 35-bp c2-10 region that contains the Abp1p binding site and other Abp1p-binding DNA fragments from the central cores and centromeric repeats (Fig. 3 A; Ngan, V.K., and L. Clarke, manuscript submitted for publication) suggest that the binding motif encompasses distinct AT-rich sequences interspersed with C and G residues. Identification of a consensus motif, however, has been complicated by the AT richness of the central cores (∼70%) and the length of the DNA fragments implicated (100–400 bp).

A database search using the BLAST program (Atschul et al., 1990) revealed that Abp1p has several domains with significant homology to human CENP-B and several other CENP-B–related proteins. Pairwise alignments were most significant between Abp1p and CENP-B from human and mouse (P values ∼4.3 × 10⁻⁷; Sullivan and Glass, 1991; Earnshaw et al., 1987) and between Abp1p and the CENP-B–related jerky gene product from M. musculus (P value 2.5 × 10⁻¹²; Toth et al., 1995). CLUSTAL V alignments (Higgins et al., 1992) indicate that CENP-B (599 aa) and Abp1p (522 aa) are ∼25% identical and 53% similar over 522 aa (Fig. 5), parameters similar to those reported by Murakami et al. (1996). A CLUSTAL V alignment of Abp1p and Jerky recognizes 20.8% identity and 48.5% similarity over 509 aa (data not shown). Not surprisingly, a similar level of homology has also been reported between CENP-B and Jerky (Toth et al., 1995). The large number of gaps and insertions required to generate the pairwise alignment between Abp1p and CENP-B suggests that the two proteins have similar origins but have diverged significantly from one another (Fig. 5). Interestingly, portions of the NH₂-terminal and central domains of Abp1p, CENP-B, and Jerky have significant homology to regions of the IS630-Tc1 family of transposases, Drosophila pogo transposase, and other pogo-related proteins (Fig. 6; Tudor et al., 1992; Doak et al., 1994; Toth et al., 1995; Smit and Riggs, 1996) including the “D,D35E” motif that is also found in bacterial insertion sequence (IS) transposases and retroviral integrases and is thought to form part of the transposase catalytic site (Kulkosky et al., 1992; Doak et al., 1994; Pollard and Chandler, 1995). When the MACAW 2.0.5. sequence alignment program is used to search for blocks of significant homology between CENP-B and Abp1p in a pairwise fashion, sequences corresponding to block 5 (Fig. 6,B, 32 aa) are recognized as significant, and sequences corresponding to block 3 (Fig. 6,B, 12 aa) are recognized as “maybe” significant (Schuler et al., 1991). The statistical significance of the homology in other regions of the two proteins greatly increases when additional transposase-like sequences are added to the alignment. When compared with typical transposase sequences, the amino acid sequences of CENP-B, Jerky, Abp1p, and the pogo transposase–related yeast transcription factors Rag3p (GenBank accession #X70186) and Pdc2p (Hohmann, 1993) are all somewhat degenerate, indicating that they may lack transposase activity (Fig. 6 B; Smit and Riggs, 1996). Although their exact functions remain to be determined, the homology shared between CENP-B and Abp1p suggests that they may play related roles at the centromere.

We have used simultaneous genetic and biochemical approaches in an effort to identify S. pombe proteins with potential roles in centromere function. Both approaches led to the isolation of Abp1p, a protein with significant homology to the human centromere DNA-binding protein CENP-B. The gene for Abp1p was isolated by screening an S. pombe genomic library, constructed in a multicopy vector, for genes that caused high-frequency mitotic loss of a stable cen1 linear minichromosome. Although Abp1p is not essential, cells overexpressing abp1 as well as abp1 null cells are slow growing and display mitotic defects when grown at low temperature, indicating that alterations in the cellular level of Abp1p affect normal mitotic growth. In addition, Abp1p is required for normal meiotic processes to occur in most cells. Using a separate, biochemical approach, we isolated Abp1p by sequence-specific affinity chromatography using a DNA sequence from the centromeric central core of S. pombe chromosome II. Furthermore, we have shown that Abp1p binds specifically in vitro to at least three sites in the cen2 central core as well as within the K-type (K/K″/dg) repeat element (Fig. 3 A, shaded ovals; Ngan, V.K., and L. Clarke, manuscript submitted for publication), two elements that are required for centromere function (Baum et al., 1994).

Murakami et al. (1996) also independently isolated Abp1p by affinity chromatography using a target sequence containing four tandem repeats of an autonomously replicating sequence (ARS) consensus sequence that was identified in fragments capable of promoting efficient transformation in S. pombe (Maundrell et al., 1988). However, the functional significance of this 11-bp motif [(A/T)TAAATAAA(C/T)(A/T)] is unclear because deletion of this sequence has no apparent effect on the ability of such fragments to mediate autonomous replication (Maundrell et al., 1988), and some centromeric fragments, containing one or more consensus sequences, do not initiate replication at a detectable level in the chromosome (Smith et al., 1995). To detect Abp1p binding activity, Murakami et al. (1996) incubated S. pombe crude cell extracts with dimers or tetramers of a 31-bp oligonucleotide (5′-gaTCCAAATTATATAAATAAATAATTTTTGa-3′) and monitored complex formation in gel mobility shift assays, suggesting that Abp1p binding sites are AT rich but might also contain G and C residues. In support of this, we have found that replacing a CCC motif in fragment c2-10 with AAA causes a minor reduction in Abp1p binding affinity. In addition, S. pombe centromeric central core DNA fragment c2-12 (Fig. 3 A) contains an ARS consensus sequence but lacks appreciable affinity for Abp1p (data not shown). Although the S. pombe centromere contains a 5–10-fold higher density of potential replication origins than does random fission yeast DNA, initiation signals occur at only a small minority of ARS elements present in the K and L repeats, and central core and core-associated repeat sequences lack ARS activity in a chromosomal context (Smith et al., 1995). The absence of ARS activity in central core sequences that have a demonstrated Abp1p binding activity in vitro further suggests that Abp1p has a role in centromere function rather than in replication. Furthermore, FACScan^® analysis of the Δabp1 strain gives no indication of a replication defect. To reflect its centromere-related roles, we suggest that a change in designation from abp1 (Abp1p) to cbp1 (Cbp1p; centromere binding protein) is appropriate.

Amino acid sequence comparisons suggest that Abp1p/ Cbp1p shares appreciable structural homology with human CENP-B. CENP-B consists of at least three structural domains: an NH₂-terminal DNA-binding domain; a central domain of unknown function; and a COOH-terminal dimerization domain. The NH₂-terminal 125 amino acids of CENP-B delimit an experimentally determined DNAbinding domain that shows partial similarity to helix–loop– helix proteins, homeodomain proteins, and a number of transcription factors, although the exact motif responsible for DNA binding remains speculative (Sullivan and Glass, 1991; Yoda et al., 1992; Suzuki et al., 1995). Interestingly, significant homology is also observed between portions of the DNA-binding domain of CENP-B and the NH₂-terminal regions of Abp1p/Cbp1p and other proteins related to the IS630-Tc1 family of transposases, including D. pogo transposase and the two yeast transcription factors, Rag3p and Pdc2p (Fig. 6; Tudor et al., 1992; Hohmann, 1993; Smit and Riggs, 1996). Most likely, all of these proteins bind to DNA via their NH₂-terminal domains and may share a common DNA-binding motif.

In addition to a DNA-binding domain, CENP-B contains a homodimerization domain consisting of 59 COOH-terminal amino acids (Kitagawa et al., 1995). The COOH-terminal region of CENP-B lacks strong sequence conservation with Abp1p/Cbp1p as well as with the IS630-Tc1 family of transposases. However, Murakami et al. (1996) have suggested that the COOH-terminal domain of Abp1p/Cbp1p may function as a multimerization site. In addition, gel filtration data presented here and elsewhere suggest that Abp1p/Cbp1p binds to DNA as a dimer, either alone or as part of a multisubunit protein complex (Ngan, V.K., and L. Clarke, manuscript submitted for publication).

Separating the NH₂-terminal and COOH-terminal domains of CENP-B is a central region of unknown function with homology to the central domains of Abp1p/Cbp1p and the IS630-Tc1 transposase family (Fig. 6; Smit and Riggs, 1996). Although CENP-B and Abp1p/Cbp1p both contain mutations and/or altered spacing in the invariant residues of the D,D35E motif and may lack transposase catalytic activity, it is intriguing that both mammalian and fission yeast centromere DNA-binding proteins share homology to transposases. DNA sequencing has revealed the presence of a large number of transposons contained within the centric heterochromatin of functional Drosophila minichromosome derivatives (Le et al., 1995; Karpen et al., 1996) and within the centromeric region of Neurospora crassa chromosomes (Cambareri, E., R. Aisner, and J. Carbon, unpublished data). Notably, short and long interspersed nucleotide elements (SINEs and LINEs) are clustered at or near mammalian centromeres (Tyler-Smith and Willard, 1993). In addition, the size, structure, and head-to-tail arrangement of tandem repeat elements at the S. pombe centromere resemble known transposon arrangements, although these repeats contain no long open reading frames and no obvious sequence homologies to known transposons (Ngan, V.K., and L. Clarke, manuscript submitted for publication). It has been suggested that transposon repeats, either alone or in combination with satellite DNAs, may play a role in the formation and/or function of centromeric heterochromatin (Carmena and Gonzalez, 1995). Interestingly, fluorescent microscopy has shown that achiasmate chromosomes are held together by homologous regions of centric heterochromatin (Dernberg et al., 1996). These heterochromatic regions apparently contain multiple pairing elements that act in an additive fashion to promote efficient achiasmate meiotic chromosome disjunction in Drosophila, suggesting that cross-multimerization of intrinsic heterochromatic proteins or simultaneous binding to identical sequences on chromosome homologs by “special” proteins or complexes might mediate the pairing process (Karpen et al., 1996). Little is known about meiotic centromere function in S. pombe. One possibility is that multiple centric pairing proteins may maintain sister chromatid cohesion in meiosis I and throughout mitosis in a manner similar to that proposed for the achiasmate segregation elements in Drosophila. Abp1p/Cbp1p binds in vitro to multiple regions in S. pombe centromeric DNA and has apparent roles in both mitosis and meiosis. One potential role for Abp1p/ Cbp1p would be to bind to transposon-like elements along the length of the centromere and facilitate the formation of a higher order cis- and/or trans-DNA structure via homodimerization and/or as part of a multicomponent protein complex.

Although Abp1p/Cbp1p is not absolutely essential for mitotic cell division, alterations in the cellular level of Abp1p/Cbp1p severely disrupt minichromosome segregation and have marked effects on cell growth and native chromosome segregation at low temperatures. The ability of Abp1p/Cbp1p to bind to centromeric DNA and to destabilize a minichromosome when expressed at elevated levels suggests that the stoichiometry of Abp1p/Cbp1p binding at the centromere is important for mitotic chromosome segregation. In addition, the severe meiotic defect observed in cells lacking Abp1p/Cbp1p suggests that Abp1p/Cbp1p performs an important role in the meiotic process. Because Abp1p/Cbp1p apparently binds specifically to both the centromeric central core and K-type repeat elements, and gel filtration experiments suggest that Abp1p/Cbp1p dimerizes in solution, it is intriguing to speculate that Abp1p/Cbp1p may play a role in bringing together K-repeat and central core DNA elements in the formation of a higher order structure at the S. pombe centromere (Clarke et al., 1993; Marschall and Clarke, 1995; Ngan, V.K., and L. Clarke, manuscript submitted for publication), a role similar to that proposed for CENP-B at the mammalian kinetochore. Computer comparisons of sequences that bind Abp1p/Cbp1p do not detect a consensus motif, although they are all AT-rich sequences with some G and C residues. Even though the binding site appears very degenerate, affinity-purified Abp1p/Cbp1p binds to the c2-10 sequence with higher affinity (K_app > 10⁹ M⁻¹; Baum, M., and L. Clarke, unpublished data) than CENP-B to the CENP-B box (6 × 10⁸ M⁻¹; Muro et al., 1992). The ability of Abp1p/Cbp1p to bind to distinct AT-rich sequences in the central core regions, K repeats, and other centromeric repeated DNAs is consistent with the anticipated properties of an S. pombe centromere DNA-binding protein. The lack of appreciable sequence conservation among the S. pombe centromeric central core regions suggests that centromere DNA-binding proteins may recognize small and/or degenerate sequences or a conserved structural motif (Takahashi et al., 1992; Marschall, L., and L. Clarke, unpublished data). Because an abp1/cbp1 deletion is not lethal, Abp1p/Cbp1p may be functionally redundant with one or more other centromere DNA-binding proteins. This is not surprising given the apparent overlapping functions of S. pombe centromeric DNA elements (Baum et al., 1994; Ngan, V.K., and L. Clarke, manuscript submitted for publication) and the potential functional redundancy of CENP-B at the human kinetochore.

Whether Abp1p/Cbp1p plays a role in the formation of higher order DNA structure at the S. pombe centromere and/or plays a role in chromosome pairing remains to be determined. Nevertheless, Abp1p/Cbp1p is the first fission yeast centromere DNA-binding protein to be isolated, and it provides a valuable tool for the identification of other S. pombe kinetochore proteins. Furthermore, the apparent structural and functional homologies shared between Abp1p/ Cbp1p and CENP-B strongly suggest that knowledge gained from studying fission yeast centromeres will translate into an advanced understanding of the structure and functional organization of the higher eukaryotic kinetochore.

We thank Sean George for technical assistance, Vivian Ngan and Ed Cambareri for critical reading of this manuscript, and members of the L. Clarke and J. Carbon laboratories for helpful discussions.

This work was supported by U.S. Public Health Services grant GM33783 from the National Institutes of Health.

Abp1p

autonomously replicating sequence-binding protein

ARS

autonomously replicating sequence

cc2

central core DNA of Schizosaccharomyces pombe chromosome II

DAPI

4′, 6-Diamidino-2-phenylindole

IS

insertion sequence