A Chromatin-associated Kinesin-related Protein Required for Normal Mitotic Chromosome Segregation in Drosophila

The tio gene was localized to polytene subdivision 38B, based on in situ hybridization to the ry⁺ insert in the P element–induced tio¹ allele, consistent with the meiotic map position (2–54.5 map units; no recombinants between tio and pr out of 28 recombination events between b and cn) and deficiency complementation (38A6-B1; 38B6-C1, defined by the distal breakpoint of Df(2L)TW¹⁶¹ and the proximal breakpoint of Df(2L)pr^A20).

Four different tio mutant alleles were studied: tio¹, which encodes a truncated transcript, was found during a search for mitotic mutants among a collection of P(ry⁺) element–induced lethal and semilethal mutations generated by C. Berg (University of Washington, Seattle, WA), L. Cooley (Yale University, New Haven, CT), and A. Spradling (Carnegie Institute, Baltimore, MD) (11), who previously called it sl(2)ry3.  tio^24-O (an apparent transcriptional null) and tio^93-E (a possible hypomorphic allele) were obtained from a collection of P(w⁺) element–induced mutations generated in M. Goldberg's laboratory (Cornell University, Ithaca, NY). Excision of the tio¹ and tio^93-Emarked P elements by introduction of the Δ2-3 source of transposase resulted in reversion to wild-type of all the mutant traits shown by homozygous individuals. The tio^ie1 allele, an apparent transcriptional null, was recovered by imprecise excision of the tio¹Pelement.

All other mutations and chromosomal rearrangements are described in reference 27, except for T(2;3)SM5-TM6B, al² Cy lt^v cn² sp²; ss⁻ bx^34e e Hu Tb, a translocation between the SM5 and TM6B balancers induced by × irradiation by C. González and J. Casal (Centro de Biologia Molecular, Madrid, Spain), which was used in all crosses where homozygous tio larvae had to be selected.

All crosses were made at 25°C in standard Drosophila medium. In all cases, the multiple phenotypic traits associated with tio mutations varied depending on genetic background to the point that, even for transcriptionally null alleles, homozygotes vs. hemizygotes showed quantitatively different phenotypes. To minimize genetic background effects, quantitative differences between alleles were studied in hemizygotes carrying Df(2L)pr^A20, pr cn (38A3-4; 38B6-C1).

To quantify meiotic nondisjunction, 10 females were mated to 5 males in small vials. The parents were transferred to fresh vials every 3 d for a total period of 12 d. Progeny classes were scored every other day from eclosion until the 18th day after parents were introduced into the vial.

Cytological analysis of squashed larval brains stained with aceto-orcein were carried out as in reference 20. In scoring ploidy, the number of sex chromosomes and the number of large autosomes were counted. Second and third chromosomes were not distinguished. Circular mitotic figures (CMFs) were scored as cells with chromosomes or condensed chromatin arranged in a circle around a single apparent pole. Clear cases of CMFs with chromatids (as opposed to chromosomes) oriented toward a single pole were not observed. Mitotic figures were scored as anaphases when they had two or more apparent poles with condensed chromatin oriented toward or clustered around them. In many cases, especially those with overcondensed chromatin, it was not possible to distinguish chromatids versus chromosomes. However, where they could be distinguished, it was chromatids that were oriented toward the anaphase poles.

In situ hybridization to ovaries, embryos, and testes was carried out as in reference 20, with DIG-labeled RNA probes used for embryos and ovaries and DIG-labeled DNA (Boehringer Mannheim Corp., Indianapolis, IN) probes used for hybridization to testes.

The tiovivo locus was cloned using the P element insert in the tio¹ allele. Genomic DNA flanking the P element 3′ inverted repeat from the tio¹ allele was amplified by inverse PCR. The resulting PCR product, which contained 0.1 kb of P element DNA plus about 0.4 kb of flanking genomic DNA, was subcloned by blunt end ligation into the SmaI site in the polylinker of Bluescript KS II (Stratagene, La Jolla, CA) and used to isolate lambda EMBL3 phage carrying wild-type genomic DNA (library from J. Tamkun, UC Santa Cruz) from the region. DNA fragments representing 20 kb of genomic DNA surrounding the tio¹ insertion site were used to screen a 0–4-h embryonic cDNA library in pNB40 (N. Brown, Cambridge University, Cambridge, UK) and an adult testes cDNA library constructed in Lambda Zap II (T. Hazelrigg, Columbia University, New York). A full-length cDNA corresponding to the 3.6-kb KLP38B transcript and a full-length cDNA corresponding to the intronic 1.8-kb transcript were obtained from the 0–4-h embryonic and the testis cDNA libraries, respectively. The two full-length cDNAs were sequenced on both strands using Sequenase. Sequences were processed using the University of Wisconsin Genetics Computer Group package of sequence analysis programs (12). A third cDNA family that hybridized to the rightmost end of the cloned genomic DNA was also identified.

Total RNA was extracted from the appropriate Drosophila stage by the RNAzol method (TM Cinna Scientific Inc., Friendswood, TX). polyA⁺ RNA was affinity purified by passage through an oligo-dT cellulose column. Northern blot analyses were carried out as described in reference 45. DNA probes were either digoxigenin-labeled by the random primed method using the DIG-DNA labeling kit (Boehringer Mannheim Corp.) or radioactively labeled by random priming. Antisense riboprobes were radioactively labeled during in vitro transcription from cDNA clones or subclones as in reference 26.

Antibodies were raised against the leu800 to gln1080 region of the nonconserved tail of KLP38B, expressed as a fusion protein in bacteria. The 3′ PstI restriction fragment (position 2398 to 3240 bp, encoding 281 amino acids) from the full-length cDNA corresponding to the 3.6-kb transcript was cloned into the PstI site of pQE10 (Qiagen, Chatsworth, CA) to produce a 6-His fusion protein (6HKLP38B). The 6HKLP38B fusion protein was prepared following the protocol of the manufacturer for purification of denatured insoluble proteins. Fusion protein was purified by binding to a Ni-NTA resin column (Qiagen) and eluted from the column at pH 5.9. Antiserum was produced by subcutaneous injection of rats with 50 μg of 6HKLP38B fusion protein at 14, 28, 49, and 84 d after the first injection.

Immunostaining of embryos was performed as in reference 20 and of testes as in reference 7. Antiserum raised against KLP38B was used at a dilution of 1:50 for embryos and 1:100 for testes. Secondary antibodies used to reveal tio staining were 1:200 dilutions of FITC- or RITC-conjugated goat anti–rat IgG (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA). Subsequently, tubulin was detected by staining with a 1:200 dilution of YL1/2 antibody (Sera-Lab Inc., Sussex, UK), followed by a 1:400 dilution of anti–rat biotinylated secondary antibody (Amersham International, Buckinghamshire, UK), plus a 1:200 dilution of Lysamil-rhodamine-streptoavidin (Jackson ImmunoResearch Laboratories, Inc.). Alternatively, tubulin was detected simultaneously with KLP38B by staining with a 1:10 dilution of monoclonal antitubulin antibody 3A5 (38) followed by a 1:200 dilution of an anti–mouse RITC-conjugated secondary antibody (Jackson ImmunoResearch Laboratories, Inc.). Immunofluorescence staining of larval brains was performed as in reference 20, using a 1:1,000 dilution of affinity purified anticentrosomin rabbit polyclonal antibody provided by K. Li and T. Kaufman (23) and visualized with a 1:200 dilution of Cy5-conjugated anti–rabbit secondary antibody (Jackson ImmunoResearch Laboratories, Inc.). DNA was stained with propidium iodide after RNase treatment. Immunofluorescence staining was analyzed with a laser scanning confocal microscope (Carl Zeiss, Inc., Thornwood, NY). Dual color overlays and composites of confocal and conventional images were prepared using Adobe Photoshop (San Jose, CA).

Mutations in tio cause abnormal chromosome behavior and segregation in dividing larval neuroblasts (Fig. 1). For hemizygotes and all interallelic combinations of the four different tio alleles studied, squashed preparations of larval brains stained with aceto-orcein showed abnormal mitotic figures, indicating a requirement for tio⁺ activity for normal cell division (Table I). A total of 1,918 mitotic figures from homozygous and hemizygous mutant brains were scored quantitatively, and many more were examined for mitotic phenotypes. Mitotic larval neuroblasts in tio frequently (22% of 1,918 mitoses scored quantitatively) had CMFs, with sex chromosomes and major autosomes arranged in a circle, centromeres pointing inwards, and the small fourth chromosomes toward the center (Fig. 1, C, F, and G). Polyploid figures were common (Fig. 1, E–G): 26% of CMFs and cells in metaphase were aneuploid or polyploid. For many of these, the number of sex chromosomes plus major autosomes was a multiple of the haploid complement of 3. However, in a significant fraction, the number of sex chromosomes plus major autosomes departed by one or two chromosomes from the true polyploid number. For example, of 40 approximately tetraploid cells picked at random from orcein-stained squashed preparations of tio mutant larval brains, 38 were 4N and 2 were 4N-1. Of 419 poly- or quasipolyploid tio cells scored, 47% appeared tetraploid and 4% appeared octoploid, suggesting either failure of chromosome segregation or failure of cytokinesis. However, intermediate levels of ploidy were also observed: 22% of the poly- or quasipolyploid tio cells scored had an approximately triploid number of sex chromosomes plus major autosomes, 5% appeared pentaploid, and 10% appeared hexaploid. 12% had more than 24 large chromosomes. Some cells were clearly aneuploid (Fig. 1 D, for example), indicating missegregation of chromosomes as well as failure of chromosome segregation or cytokinesis. Gross aneuploidy was occasionally observed.

The extent of chromosome condensation in tio mitotic figures usually appeared normal. However, in 14% of metaphase figures scored, chromosomes were overcondensed, indicating metaphase arrest (29). All cells showing overcondensed metaphase chromosomes were highly polyploid, but not all highly polyploid figures had overcondensed chromosomes. The mitotic index and anaphase/ metaphase ratio in tio brains were close to normal (Table I), indicating that most dividing cells progressed through mitosis with relatively normal timing.

Anaphase figures were frequently abnormal (38% of 306 anaphases scored). Anaphase figures were often asymmetric, with chromatids arranged in a circle at one or both poles. Euploid or polyploid bipolar anaphase figures frequently had broad poles (Fig. 1,H). Tripolar and tetrapolar anaphases were found (Fig. 1, I and J), but rarely. Binucleate cells were also occasionally observed in tio brain squashes (data not shown).

Staining of tio/Df larval brains with propidium iodide and anticentrosomin revealed a variety of defects in chromosome and centrosome arrangement and number (Fig. 2), reflecting the range of mitotic defects observed in orcein-stained squashed preparations. Many mitotic figures appeared normal (Fig. 2, A and B). In a random sample of 49 cells in prometaphase, metaphase, or anaphase from tio¹/Df (25 cells) or tio^24-0/Df (24 cells), larval brains stained for immunofluorescence in two different experiments, roughly half (63%) had no apparent defects. However, cells with clearly abnormal numbers and arrangements of chromosomes and centrosomes were common (37%). Defects observed ranged from circular arrays of chromosomes surrounding a single centrosomal mass (6%) (Fig. 2,C), consistent with failure of centrosome separation, to cells with two separated centrosomes but defects in chromosome number (Fig. 2,E) or arrangement (Fig. 2,F) (25%). In some cases (roughly 6% of mitotic figures scored), cells appeared to have some chromosomes aligned at a metaphase plate while other chromosomes in the same cell appeared clustered around only one of the two separated centrosomes (Fig. 2,D), suggesting defects in the ability of chromosomes to form bipolar attachments to the spindle rather than defects in centrosome separation. Some cells (6%) had more than two well-separated centrosomes (Fig. 2,H, 4 centrosomes; Fig. 2,G, 8 centrosomes: 5 are visible, and 3 more were out of the plane of focus). As in quantitating the cells in mitosis, we only counted those with large centrosomes and clear, well-condensed chromosomes, so the frequency of abnormal mitotic figures may have been underestimated. This is especially true for cells with a single centrosomal mass where chromosomes may have been either poorly condensed or too closely massed to be clearly distinguished (compare 6% to the frequency of CMFs in Table I).

tio males had mild meiotic defects. B^SY; tio¹/tio¹ males showed a weak but significant (P < 0.002) increase in sex chromosome nondisjunction (0.42% out of 2,864 progeny scored) compared with control B^SY; tio¹/CyO males (0.19% out of 1,612 progeny scored). Other tio alleles tested were male sterile. Early spermatids in tio males occasionally showed variation in nuclear size, indicating defects in meiotic chromosome disjunction (8), or were binucleate, indicating karyokinesis without cytokinesis (Fig. 1 K, arrows).

Relative viability of tio hemizygotes compared with heterozygous siblings ranged from 0 to as high as 80%, depending on genetic background and culture conditions. tio adults had visible defects often associated with mitotic mutants, including rough eyes and bristle abnormalities. tio adults showed behavioral defects including held up wings, delay in or failure to mate, and retention of fertilized eggs. Females often had one or both ovaries rudimentary. Animals lacking both maternal and zygotic tio function died as mature embryos without strong or consistent pattern defects. Of about 100 mature embryos from tio¹/Df females mated to tio¹/CyO males, only one with defective segmentation and two with general cuticle defects were observed. Head involution defects were more common but less frequent than the expected number of homozygous embryos. Of 300 midstage embryos produced from a similar cross, only two showed gross abnormalities (lack of anterior segments during germ band elongation). The relative viability of homo- or hemizygous mothers did not noticeably affect the phenotype or viability of their offspring. Animals lacking the maternal tio contribution but carrying a paternally contributed wild-type tio allele were viable and appeared normal.

Over 20 kb of wild-type genomic DNA surrounding the insertion site of the tio¹ allele was cloned (Fig. 3,A) (Materials and Methods). Northern blots probed with fragments of the cloned DNA revealed that the tio region encodes two transcripts (Fig. 3): a 3.6-kb transcript abundant in 0–2 h embryos and present in later embryos and adults (Fig. 3,B) and a 1.8-kb transcript, abundant in 4–8-hour embryos and present in later embryos and adults (Fig. 3,B). Full-length cDNAs corresponding to the 3.6- and 1.8-kb transcripts from the region flanking the tio¹ insert were isolated (Materials and Methods) and hybridized to Southern blots of the digested genomic phage clones to determine the arrangement of exons and introns in the region (Fig. 3,A). In addition, the sequences of the cDNAs were compared with a sequence of genomic DNA across the region kindly provided by D. Ruden. The 3.6-kb transcript derived from two exons separated by an 11-kb intron (Fig. 3,A). The oppositely oriented 1.8-kb transcript also derived from two exons, located within the large intron of the 3.6-kb transcript (Fig. 3,A). The P element responsible for the tio¹ mutation was inserted within the large intron of the 3.6-kb transcript and just 5′ of the 1.8-kb transcript (Fig. 3 A). Sequence analysis of a full-length cDNA representing the 1.8-kb transcript showed only short open reading frames and no significant homology with nucleic acid or protein sequences in GenBank.

Analysis of transcripts in tio mutants using cDNA-specific probes indicated that the 3.6-kb transcript corresponds to the tio gene product. The 3.6-kb transcript was truncated to 1.6 kb in tio¹ and returned to normal size upon reversion of the tio¹ mutation to wild type with transposase (Fig. 3,C), while the size of the 1.8-kb transcript was not altered in the tio¹Pelement–induced allele (Fig. 3 D). In situ hybridization to testes showed that message corresponding to the 3.6-kb transcript was reduced in level in testes from tio^93-E and not detected with either a 5′ or 3′ exon probe in testis from tio^24-O or tio^ie1 hemizygotes (data not shown), indicating that the latter are likely to be null alleles.

The 3.6-kb tio transcript encodes a predicted protein belonging to the kinesin heavy chain superfamily (Fig. 4). A gene in polytene interval 38B encoding a kinesin-like protein (KLP38B) had been predicted (13). To confirm that tio encodes KLP38B, we showed that an EMS-induced mutation, kindly provided by D. Ruden and shown by him to cause a stop codon at amino acid residue 760 (43), near the end of the predicted stalk region of KLP38B (see below), failed to complement and showed the characteristic mitotic defects in trans to the P element–induced tio alleles in our study. The predicted protein encoded by the tio open reading frame is 1,121 amino acids long starting from the first in frame methionine. However, the protein could start at a methionine 28, 34, or even 108 amino acid residues downstream (Fig. 4 A). None of these potential start codons occur in a nucleic acid sequence context that strongly matches the consensus for Drosophila translation starts (9).

The predicted KLP38B protein has a characteristic kinesin-like motor domain near but not at its NH₂ terminus (Fig. 4, A and D), suggesting action as a microtubule motor (31). The motor domain is followed by a stalk region (residues 524–775 of KLP38B) containing patches of significant homology with analogous regions of members of the Unc104 subfamily of kinesin-related proteins (Fig. 4, E and F), including KIF1A and KIF1B, kinesin-related proteins associated with synaptic vesicles or mitochondria, respectively (33, 37). KLP38B had the highest overall homology with humorfw, encoded by a cDNA cloned from a rapidly dividing human myeloid cell line (35). The stalk region of KLP38B has two clusters of residues predicted to participate in α-helix and/or coiled coil formation near its beginning and end (Fig. 4, B and C). The stalk region of KLP38B is followed by a COOH-terminal tail domain including a region (Ser996 to Ser1013) with an arrangement of prolines, serines, and basic amino acids resembling the DNA binding motifs of several proteins that bind to AT-rich DNA, including HMG 1 (10).

The KLP38B gene is expressed in mitotically cycling cells. KLP38B mRNA was abundant (Fig. 3,B) and uniformly distributed (data not shown) in preblastoderm embryos, present in the stereotyped premitotic spatio-temporal domains (15) during cycle 14, became restricted to nervous tissue during mid-embryogenesis, and was detected only in the procephalic lobe neuroblasts and nerve cord in late embryos (Fig. 5). KLP38B mRNA was expressed in mitotically active regions in larval brains and imaginal discs (data not shown). In stage 5 and later egg chambers, KLP38B mRNA was detected both in follicle and nurse cells (Fig. 5,C) and generally appeared uniform in the oocyte cytoplasm. In wild-type adult testes, KLP38B transcript was detected only in the mitotically active gonial cells (Fig. 5,D, arrowheads) and mature primary spermatocytes (Fig. 5,D, arrows). A 5′ exon probe but not a 3′ exon probe detected the KLP38B transcript in tio¹ testes, indicating that the truncated message encoded by the tio¹allele (Fig. 3 C) could encode a product containing the motor domain but lacking the COOH-terminal tail.

The KLP38B protein colocalized with condensed chromatin during both mitosis and male meiosis (Fig. 6). Antibodies raised against a 281–amino acid fragment from the nonconserved region of the KLP38B tail (Materials and Methods) stained mitotic chromosomes at metaphase (Fig. 6, A and B) and anaphase and nuclei at telophase (data not shown) in both syncytial blastoderm and cycle 14 embryos. The anti-KLP38B antibodies also stained metaphase (Fig. 6, C and D) and anaphase (Fig. 6, G and H) chromosomes during male meiosis. The antigen did not appear to be localized only to centromeres, but rather colocalized with the mass of condensed chromatin as detected by propidium iodide staining.

The chromosomal staining observed was due to the presence of the tio gene product, as the antigen was not detected on meiotic chromosomes in males homozygous for tio¹ (which presumably encodes a truncated product lacking the epitope against which the antibody was raised) (Fig. 6, E and F, and I and J) or tio^24-0 (a transcriptional null) (data not shown), demonstrating the specificity of the antibody staining detected by immunofluorescence. Immunofluorescence staining of embryos with the preimmune serum showed only background staining (data not shown).

Double-label immunofluorescence staining of syncytial and cellular blastoderm embryos with antitubulin and anti-KLP38B indicated that KLP38B was localized to the positions where the chromosomes lie at metaphase, anaphase, and telophase (Fig. 7), rather than to spindle fibers. KLP38B did not appear to localize to mitotic spindles, with the possible exception of a region where microtubules overlapped chromosomes at microtubule plus ends (for example, Fig. 7, E and F). Anti-KLP38B staining localized to centrosomes or the midbody was never observed.

Detection of KLP38B at the position of chromosomes was cell cycle dependent. In cycle 14 embryos, antibody staining was only detected in the mitotic domains. No localized staining was detected in interphase cells (Fig. 7 G), suggesting that KLP38B could be either dispersed or degraded during interphase.

The tiovivo gene of Drosophila, required for normal chromosome segregation during mitosis, encodes kinesin- related protein KLP38B. Six lines of evidence indicate that mutations in the KLP38B gene rather than the gene encoding the 1.8-kb transcript derived from the large intron of KLP38B are responsible for the mitotic and visible phenotypes associated with tio mutants. (a) The KLP38B message is truncated in the P element–induced tio¹ allele but restored to normal size when the mutation is reverted to wild type, while there was no alteration in size in the 1.8-kb transcript in the same mutant (Fig. 3). (b) Antibodies against the COOH-terminal tail of KLP38B stain meiotic chromatin in wild-type testes, but not in testes mutant for the two different tio alleles tested (Fig. 6). Furthermore, in situ hybridization with a KLP38B probe showed reduced or absent mRNA in testes from at least three different tio alleles. (c) A nonsense mutation in the KLP38B coding region fails to complement the P element–induced alleles used in this study for the characteristic tio mitotic defects. (d) Injection of antibodies against KLP38B into wild-type syncytial blastoderm embryos resulted in formation of circular mitotic figures (43) reminiscent of the mitotic abnormalities observed in tio mutant larval brains in our study. (e) Expression of the KLP38B cDNA under control of the hsp26 promoter but without induction by heat shock rescued the sterile and partially rescued the visible phenotypes associated with the mutants (3). (f) A 10-kb fragment of genomic DNA containing the KLP38B large intron restored normal levels of the 1.8-kb transcript but did not rescue the cytokinesis or adult morphological defects associated with a P element–induced mutation that affected KLP38B expression when introduced into flies by P element–mediated germ line transformation (36).

KLP38B protein is associated with condensed chromatin during mitosis. Although direct binding to DNA has not yet been tested, KLP38B, like other members of the chromokinesin family of kinesin-like proteins proposed by Vernós and Karsenti (48), could associate with condensed chromatin directly via a possible DNA binding motif in its COOH-terminal cargo domain. The possible DNA binding motif of KLP38B contains a predicted cdc2 phosphorylation site (34) at Ser1010, suggesting that cell cycle–dependent association of KLP38B protein with chromatin might be regulated by phosphorylation. Indeed, L. Alphey and co-workers (3) have demonstrated physical interaction between KLP38B and PP187B, a phosphatase involved in mitosis in Drosophila (4). As KLP38B transcripts accumulate in cells about to enter mitosis or meiosis, the protein could be degraded after telophase and synthesized de novo before the next cell division.

Lack of KLP38B function leads to a variety of mitotic defects, including aneuploid and polyploid cells, asymmetric metaphase and anaphase figures with abnormal arrangements of chromatin, and a high frequency of CMFs resembling the array of monooriented chromosomes around monopoles in newt lung cells (5) and several other Drosophila mitotic mutants (18, 19, 22, 46). Cells where chromosomes remain in a circular array around a single centrosomal mass (as in Fig. 2 C) would be unlikely to undergo cytokinesis, contributing to the observed high incidence of true polyploids. Ohkura et al. also observed elevated frequencies of 4N and 8N cells in KLP38B mutants, which they attributed to failure of cytokinesis (36). However, Ohkura et al. did not observe the variety of other mitotic defects found in our study, perhaps because they characterized two hypomorphic alleles, whereas strong loss of function or null alleles were included in our study.

We propose that KLP38B protein bound to condensed chromatin facilitates interactions between chromosome arms and microtubules important for both bipolar attachment of chromosomes to the spindle and bipolar spindle assembly. The CMFs and wide range of mitotic defects characteristic of tio could arise from defects in both of these processes.

KLP38B could act to push chromosome arms away from spindle poles, either by coupling chromosome arms to astral microtubule dynamics or by plus end–directed microtubule motor activity. When a chromosome first captures an astral microtubule in early prometaphase, it moves rapidly toward the corresponding spindle pole (40). Loss of KLP38B function could lead to monooriented chromosomes because of insufficient antipoleward force to move chromosomes out to a position where they encounter and capture microtubules from the opposite pole. If some chromosomes are still monooriented by anaphase onset (as often occurs in cultured newt lung cells [42]), it could account for the aneuploid cells and odd levels of ploidy observed in tio mutants. For example, Fig. 2,D shows a cell in which some chromosomes are aligned at the metaphase plate while several chromosomes appear to remain monooriented toward the same pole. If such a cell were to enter anaphase, the products might be one daughter with a near triploid chromosome constitution and one daughter with fewer than the normal number of chromosomes (as in Fig. 2 E).

KLP38B could also act to capture and stabilize astral microtubules in the vicinity of chromatin, thereby converting a pair of asters to the classic bipolar spindle shape and stabilizing spindle bipolarity by increasing antiparallel interactions between microtubules from opposite poles. A similar role was proposed for the Xenopus chromokinesin XKLP1 based on in vitro studies (49).

Ability of chromokinesins to mediate interactions between chromosomes and spindle microtubules could play an integral role in the mechanism of centrosome separation. When centrosomes begin separation in prophase, each organizes an astral array of microtubules, but in many cases little or no central spindle is evident (6, 30, 39, 51)_. If nuclear envelope breakdown occurs before centrosomes are well separated, stabilization of microtubules extending toward the chromosomes by KLP38B and other chromokinesins would create a V-shaped microtubule array with the vertex at the chromosomes and the tips at the centrosomes, an arrangement commonly observed in Drosophila prometaphase neuroblasts (52). The resulting juxtaposition of microtubules from opposite poles could facilitate antiparallel interactions that help drive the poles apart, perhaps via action of bipolar complexes of other plus end– directed kinesin-related proteins such as the BIMC homologue KLP61F (14, 22, 24, 25, 52). If KLP38B mutants are deficient in ability of chromosome arms to stabilize microtubule arrays, there could be insufficient force to reliably separate spindle poles after nuclear envelope breakdown, leading to the CMFs (Fig. 1) and monopolar chromosome arrays (Fig. 2,C) observed in tio larval brains. Subsequent failure of cytokinesis would then give rise to tetra- and octoploid cells (Table I) and polyploid cells with four or eight centrosomes (Fig. 2, G and H). Whether cells form monopoles leading to true tetraploids or well-separated spindle poles with abnormally oriented chromosomes in tio may depend on cell-to-cell variation in the relative timing of centrosome separation and nuclear envelope breakdown.

The effects of tio null mutations were surprisingly mild, perhaps due in part to functional redundancy with the product(s) of similarly acting gene(s), for example nod. Both tio and nod encode predicted microtubule motors that bind to chromosomes. Although nod function is not essential for mitosis, nod is expressed in somatic cells and a dominant allele results in a phenotype in mitotic cells (17). If KLP38B and Nod play similar roles in mitosis, wild-type nod function might not be essential for viability or mitosis because KLP38B is normally sufficient. Likewise, normally cryptic function of Nod in mitotic cells may partially alleviate the effects of loss of function of KLP38B.

Much of the role played by chromosomes in assembly of a bipolar spindle can be substituted by plasmid DNA-coated beads previously equilibrated in a Xenopus interphase egg extract to allow the DNA to form chromatin (21). Although these beads probably lacked kinetochores, when placed into a mitotic extract they induced local assembly of microtubules, which then resolved into a fine bipolar spindle (21). Chromokinesins like KLP38B, XKLP1, and Nod are good candidates for the chromatin-associated factors that allow plasmid-coated beads to mimic chromosome arms. Thus, the properties of microtubule motors associated with chromatin, together with plus end–directed microtubule motors that cross-link and slide antiparallel microtubules and minus end–directed microtubule motors that bundle microtubule ends, can provide many of the functions essential for bipolar spindle assembly in animal cells.

We thank D. Ruden (University of Kansas), M. Goldberg, and A. Spradling for providing tio alleles, and L. Alphey (University of Manchester), D. Ruden, and D. Glover (University of Dundee) for communicating unpublished results. We thank K. Li and T.C. Kaufman (Indiana Univesity) for anticentrosomin antibodies and M. Scott (Stanford Univeristy, Stanford, CA) for access to a confocal microscope at Stanford. M.T. Fuller thanks L.S.B. Goldstein (University of California, San Diego) for pointing out the extended region of homology between KLP38B and KIF1A and KIF1B. The group in Madrid is indebted to A. Villasante (Madrid) for his continuous help and encouragement, N. Azpiazu, M. Calleja, I. Guerrero, and E. Sanchez-Herrero (all in Madrid) for discussions, and A. Sanchez for technical assistance.

CMF

circular mitotic figure

tio

tiovivo

The sequence data for tio/KLP38B is available from GenBank/EMBL/DDBJ under accession number Y15427.