Mitch – a rapidly evolving component of the Ndc80 kinetochore complex required for correct chromosome segregation in Drosophila

Among the cellular structures most responsible for the accurate distribution of chromosomes during cell division are the centromeres and their associated kinetochores. Centromeres are the sites where sister chromatids are most closely attached during the period between chromosome replication and anaphase onset (reviewed in Chan et al., 2005; Craig and Choo, 2005). Centromeres also elaborate kinetochores: proteinaceous structures containing microtubule (MT) motors and other molecules that connect chromosomes to the spindle (such as CENP-E (Putkey et al., 2002), power chromosomal movements (such as dynein) (Savoian et al., 2000), and disassemble MT plus ends (such as MCAK) (Hunter et al., 2003). Kinetochores are also the seat of the spindle assembly checkpoint (SAC), which prevents the onset of anaphase until the kinetochores are stably attached to the spindle. Several molecular components of the SAC are specifically present on kinetochores that are not attached to the spindle (reviewed by Lew and Burke, 2003).

Many kinetochore proteins were first discovered in genetic screens, particularly in the yeast Saccharomyces cerevisiae. Mutations that disrupted the fidelity of chromosome segregation led to the discovery of kinetochore proteins, such as Ctf19, Mcm21 and Cin8 (Hoyt et al., 1992; Poddar et al., 1999; Spencer et al., 1990). Most of the canonical SAC proteins were originally identified through screens for mutations that prevented cells from arresting mitosis in the presence of spindle poisons (Hoyt et al., 1991; Li and Murray, 1991). Although many yeast kinetochore components have proven to be conserved in higher eukaryotes, others have no identifiable metazoan homologs (reviewed by Meraldi et al., 2006). Because yeast kinetochores are imperfect models for kinetochores in multicellular eukaryotes, we have conducted genetic screens for mutants that disrupt mitotic chromosome segregation in Drosophila melanogaster. The utility of this approach is illustrated by our identification of a multisubunit complex containing the proteins Zw10, Rod and Zwilch that associates with kinetochores and is conserved throughout metazoan, but not yeast, lineages (Scaerou et al., 2001; Williams et al., 2003). Here, we describe a new kinetochore protein, Mitch (mitotic chaos), which is required for correct chromosome alignment and segregation during meiosis and mitosis in Drosophila.

Our laboratory recently screened for mitotic mutants among a large collection of ethylmethanesulfonate (EMS)-mutagenized Drosophila stocks (Koundakjian et al., 2004) in which homozygosity for EMS-treated third chromosomes caused lethality during third instar larval or pupal stages (see Materials and Methods). To find mitotic mutants, brains isolated from dying third instar larvae were treated with colchicine, then fixed and stained with orcein, because these conditions are particularly favorable for visualizing chromosome number and morphology (Gatti and Goldberg, 1991). The brains of larvae homozygous for a particular mutagenized third chromosome, bearing the mutation we now call mitch¹, displayed high levels of aneuploidy (Fig. 1B, compare with wild-type in Fig. 1A; Table 1). Substantial levels of aneuploidy also occurred in the brains of larvae carrying mutations for other alleles and interallelic combinations of mitch that we subsequently obtained (Fig. 1C; Table 1). Our analysis of the chromosome complements of aneuploid cells suggested that all chromosome types were equally subject to missegregation in these mutant strains (data not shown).

To view all the stages of mitosis in mutant brains, we investigated the mitch mutant phenotype in the absence of colchicine, visualizing mitotic figures with orcein (Fig. 1F-I) or by staining fixed brains for DNA and tubulin (Fig. 2). Three major types of aberration were observed. First, many mitch mutant cells contained moderately overcondensed chromosomes (Fig. 1F, Fig. 2D, Table 1), suggesting delays in mitotic progression. Second, the chromosomes were generally unable to congress correctly to the equator of the spindle: in contrast with wild-type (Fig. 2A), few mitch cells contained a well-formed metaphase plate (Fig. 2C). Third, the majority of anaphases in mitch mutant cells displayed lagging chromosomes and the unequal allocation of chromosomes (Fig. 1H,I; Fig. 2D); such defects are not seen in wild type (Fig. 1G, Fig. 2B). Since tubulin staining revealed relatively normal bipolar spindles, the abnormal chromosome distribution seen in fixed mitch neuroblasts is unlikely to be caused by disruptions of the spindle apparatus (Fig. 2).

To better understand the mitch mutant phenotype, we obtained time-lapse DIC recordings of cultured mutant brain cells (Fig. 3; see also supplementary material Movies 1-5). These movies verified that chromosomes in mitch mutants fail to congress correctly during prometaphase or to segregate faithfully at anaphase. Particularly notable was the persistence of mono-oriented chromosomes around the spindle poles (Fig. 3B,C; supplementary material Movies 2 and 3). In all of our recordings of mutant neuroblasts, at least two chromosomes were mono-oriented for periods averaging about 1 hour (range, 25 minutes to >2 hours). The contrast with wildtype, in which full congression and stabilization of chromosomes at the metaphase plate is achieved in less than 4 minutes, is striking.

Mutations of mitch not only cause defects in mitosis, but also in the two meiotic spermatocyte divisions. Bivalent congression during meiosis I and II was clearly disrupted in mitch mutant spermatocytes (Fig. 4B-C,I; compare with wild-type in Fig. 4A,H). Chromosome segregation during anaphase and telophase of both meiotic divisions (wild type shown in Fig. 4D,E,J) was also defective. Approximately 85% of ana/telophase I figures exhibited an unequal distribution of chromosomes, and in one-third of these cases, one daughter cell received the entire chromosome complement while the other received no chromosomes (Fig. 4F,G). A large percentage of ana/telophases in meiosis II were also unequal (Fig. 4K) and, as expected from the asymmetrical first division, many of the resulting cells did not contain any chromosomes at all (Fig. 4L). Our findings are consistent with previous reports in other Drosophila mutant strains that meiosis can occur in the absence of chromosomes (Bucciarelli et al., 2003).

Depletion of Mitch in Drosophila S2 tissue culture cells using RNA interference (RNAi) resulted in mitotic phenotypes similar to those seen in mitch-mutant larval brains. At metaphase, cells treated with mitch double-stranded (dsRNA) exhibited defects in chromosome congression, including apparently mono-oriented chromosomes near the spindle poles (supplementary material Fig. S1). Depletion of Mitch by mitch dsRNA was verified by antibody staining (data not shown). Furthermore, in mitch RNAi cells incubated with colchicine, the spindle checkpoint was defective, with precocious sister chromatid separation (PSCS) occurring in 39% of cells (n=739), in contrast to 0.7% PSCS in control cells receiving no dsRNA (n=520) (supplementary material Fig. S2). Cells incubated with dsRNA from genes with no role in mitosis do not cause these effects (Yu et al., 2004).

Wild-type larval brain cells treated with spindle poisons exhibit a mitotic arrest (c-metaphase) in which sister chromatids remain attached at the centromeres, whereas cyclin B levels remain elevated for extended periods (Gonzalez et al., 1991). The disruption of MTs creates unattached kinetochores, which generate spindle assembly checkpoint (SAC) signals that delay anaphase onset (Basu et al., 1999). SAC components such as Bub3 and Mad2 strongly localize to kinetochores when checkpoint signaling is `on' in wild-type cells with disrupted spindles (Basu et al., 1998; Logarinho et al., 2004; Williams et al., 2003).

The persistence of mono-oriented chromosomes in mitch mutant neuroblasts suggested the presence of unattached kinetochores that maintain SAC signaling and consequently cause mitotic delays or arrest. We found that the SAC is indeed operational in mitch cells. First, as previously mentioned, the majority of chromosomes in mitch cells become overcondensed (Fig. 1F and Fig. 3, Table 1), as occurs when wild-type cells are delayed in metaphase for extended periods by MT poisons (Gatti and Goldberg, 1991; Gonzalez et al., 1991). Second, the ratio of cells in prometaphase/metaphase to those in anaphase/telophase was generally twofold higher in mitch mutant versus wild-type brains (Table 1). Unexpectedly, in mitch mutants the mitotic index was not elevated relative to wild-type cells, and was in fact reduced in the strongest allelic combinations (Table 1). We attribute this low mitotic index to the death of mutant cells that were delayed in division for long periods. In this respect, Acridine Orange staining revealed a substantial increase in the number of apoptotic cells in mitch mutant brains (data not shown). The most direct evidence that mitch mutants contain a functional SAC comes from time-lapse observations of living cells (Fig. 3A-C; Movies 1-3). Of the mitch cells recorded that eventually entered anaphase, the average time between nuclear envelope breakdown and anaphase onset was 39±22 minutes compared with 11±3 minutes in wild type; these numbers underestimate the severity of the defect because several of the mitch mutant cells observed by us never entered anaphase after 1-2 hours of continuous recording.

Given the evidence above that the SAC is operational in untreated mitch neuroblasts, we were surprised to find that spindle poisons fail to activate SAC signaling in mitch mutant cells (as they do in wild-type cells). Instead of arresting in mitosis, colchicine-treated mitch cells prematurely enter anaphase. For example, in contrast to wild-type cells in which colchicine leads to an increased mitotic index (MI) from 1.04 to 2.8 (see Table 1), colchicine does not elevate the MI in mitch mutant cells (0.91 in the absence of colchicine and 1.1 in its presence; Table 1). Additionally, in the presence of colchicine, mitch mutant cells display a high incidence of PSCS, which is a hallmark for exit from mitosis (Fig. 1D,E and Table 1). Neither wild-type (Fig. 5A; Table 1) nor mitch heterozygote brains (data not shown) show significant levels of PSCS. Colchicine-treated mitch cells with PSCS also show other signs of mitotic exit such as low levels of cyclin B (see Fig. 5B and compare with Fig. 5A in which the chromosomes are still attached) and low levels of Bub3 and Mad2 staining at kinetochores (Fig. 5C-E). Most convincingly, we found that in live cell recordings that colchicine-treated mitch mutant neuroblasts exited mitosis after only 18.1±4.5 minutes (Fig. 3E; supplementary material Movie 5), which is roughly the same amount of time that untreated wild-type cells spend in prometaphase/metaphase. By contrast, three of four control live wild-type cells remained blocked in mitosis for more than 3 hours after colchicine exposure, although one cell adapted to the SAC and exited mitosis after 1 hour (Fig. 3D; supplementary material Movie 4). All of these findings are consistent with the idea that colchicine-treated mitch cells prematurely enter anaphase because they inappropriately adapt to the SAC, either because the SAC is nonfunctional or because Mitch is specifically required for SAC function in the absence of spindle MTs.

It has been suggested that various components of the SAC detect different aspects of kinetochore/spindle interactions, with some components monitoring MT occupancy at kinetochores and others sensing tension across attached kinetochores (Logarinho et al., 2004; Waters et al., 1998). One explanation for the presence of a functional SAC in untreated, but not in colchicine-treated, mitch mutant cells is that mitch mutations specifically interfere with the ability of the checkpoint to detect the absence of MT attachments. To test this, we incubated wild-type and mitch-mutant brains with taxol, which stabilizes MTs (Williams and Goldberg, 1994). With this treatment, kinetochore-MT interactions are maintained, but the centromeres are not under tension (Waters et al., 1998). We observed high levels of PSCS in taxol-treated mutants, and also noted that the mitotic index was similar to that in colchicine-treated mitch mutants, yet significantly lower than that in drug-treated wild-type brains (Table 1). These data imply that the SAC in mitch mutants is either non-functional or prematurely overridden not only when MTs are absent, but also when MTs are present but their dynamics compromised.

Finally, to test whether the SAC bypass in mitch mutants is specific for MT poisons or instead a more general response to spindle perturbation, we examined larvae carrying mutations in both mitch and abnormal spindle (asp). In the absence of a functional asp gene, cells exhibit spindle abnormalities that normally delay anaphase onset and mitotic exit, even though many of the kinetochores become attached to the spindle (Gonzalez et al., 1998). We found, however, that in untreated mitch¹ asp¹ double mutant brains the anaphase frequency was appreciably increased with respect to that in asp¹ brains (Table 1). Moreover, a significant level of PSCS was apparent in mitch¹ asp¹ double-mutant brains, but not those of asp¹ single mutants (Table 1). These mitotic phenotypes resemble those previously reported for zw10 asp and rod asp double mutants (Basto et al., 2000); mutations in zw10, rod and zwilch also bypass the spindle checkpoint in the presence of MT poisons (Basto et al., 2000; Williams et al., 2003). However, unlike mitch, these latter genes are also required for a functional SAC in untreated neuroblasts.

By deletion mapping, we delimited the mitch¹ mutation to the region of chromosome 3R between the centromere-proximal breakpoints of Df(3R)ry1608 (87D4-6) and Df(3R)ry614 (87D2-4) (supplementary material Fig. S3) (Leonhardt et al., 1993). We determined the position of these breakpoints at the DNA level by quantitative autoradiography and single-embryo PCR (see Materials and Methods), delimiting mitch to a 30-kb region that contains five predicted genes and three previously characterized lethal complementation groups (supplementary material Fig. S3). A mutation in one of these complementation groups [l(3)87Da²; hereafter mitch²] was found to be allelic to mitch¹. The brains of third instar mitch¹/mitch² larvae exhibited the same mitotic defects previously described (Table 1), and most of these animals subsequently died during pupal stages. A few mitch¹/mitch² escapers survived to adulthood but were sterile; the testes of heteroallelic males displayed meiotic defects similar to those shown in Fig. 4. Consistent with previous observations of other l(3)87Da alleles that are no longer extant (Hilliker and Chovnick, 1981; Hilliker et al., 1980), the large majority of heteroallelic escapers were female (>80%), for reasons that remain unknown.

Sequencing of the exons in the five candidate genes revealed that mitch² has a nonsense mutation at the beginning of the third exon of the gene CG7242 that truncates the original 222-amino-acid-long gene product to 129 residues. To date, we have been unsuccessful in identifying the molecular lesion associated with mitch¹, but we suspect involvement of a mutation affecting CG7242 gene expression because the protein product is no longer detectable by immunofluorescence (see Fig. 6).

To verify the identification of CG7242 as mitch and to obtain additional mutant alleles, we generated deletions from imprecise excisions of a P element that inserted into the short intergenic region between CG7242 and its neighbor granny smith (grsm, CG7340; see Materials and Methods and supplementary material Fig. S1). Although this P element is inserted only 63 nucleotides upstream of the presumptive CG7242 transcription initiation site, it is without obvious phenotypic effect in homozygotes or in hemizygotes bearing the P element and the deficiency Df(3R)ry75. Several of the imprecise excisants are lethal in homozygotes and, moreover, failed to complement mitch¹and mitch²; they did, however, complement alleles of the two other essential genes in the region. Two of these excisions, mitch^19B and mitch²², remove the majority of CG7242 coding sequences, yet do not extend into any other nearby gene (supplementary material Fig. S1). The mitch^19B and mitch²² deletions cause mitotic phenotypes similar to those associated with mitch¹ and mitch² (Table 1). Other excisants that remove the entirety of the neighboring CG7340 gene are viable as homozygotes and exhibit normal mitosis (C.K., unpublished). Considered together, these data equate mitch with CG7242 and l(3)87Da. The COILS2 program (Lupas et al., 1991) predicts with high probability that Mitch contains at least one coiled-coil domain (amino acids 45-99) corresponding to the second exon of CG7242. Searches of databases with Mitch sequences did not uncover any significant matches with other proteins.

To determine the intracellular location of the Mitch protein, we prepared antibodies against both native and denatured forms of Mitch (see Materials and Methods). Affinity-purified antibodies against both forms were used to localize Mitch in larval brain neuroblasts by immunofluorescence. In wild-type colchicine-treated cells, all four of these antibodies specifically stain the kinetochores (Fig. 6A). Mitch is located between CID (the centromeric histone H3 variant) in the inner kinetochore/centromeric region, and Zwilch, which is part of the Zw10-Rod complex in the outer kinetochore (Fig. 7) (Blower and Karpen, 2001; Williams et al., 2003). Mitch staining corresponds very closely to that of Bub3 (Fig. 7) (Basu et al., 1998). In all mitch mutant strains examined, the kinetochores in larval neuroblasts failed to stain with any of the anti-Mitch antibodies (Fig. 6A), so the kinetochore staining in wild-type cells indeed reflects the intracellular location of Mitch. The assignment of Mitch to the kinetochore was verified by the targeting of epitope-tagged Mitch protein to kinetochores in Drosophila tissue culture cells (Fig. 6B; see also Fig. 8A).

The association of Mitch with kinetochores is cell-cycle dependent because Mitch immunofluorescence signals are only detected in dividing cells from prometaphase to late anaphase. This implies that Mitch is a transient kinetochore protein recruited to the centromere/kinetochore only during cell division. Interestingly, unlike most SAC proteins which leave the kinetochore at anaphase (Basu et al., 1999; Basu et al., 1998), Mitch remains at the kinetochores throughout the metaphase-to-anaphase transition. The localization of Mitch in meiosis is similar to that seen in mitosis. Mitch localizes to the kinetochores from prometaphase I through anaphase I, then from prometaphase II through anaphase II (data not shown).

We next asked whether the association of Mitch with kinetochores was affected in animals mutant for genes encoding either other known kinetochore proteins, including zw10, rod, zwilch, bubR1, bub3, polo and cenp-meta (Basu et al., 1999; Basu et al., 1998; Llamazares et al., 1991; Williams et al., 2003; Yucel et al., 2000), or certain spindle components, such as asp and mast/orbit (Gonzalez et al., 1998; Maiato et al., 2002). Mitch was correctly targeted to kinetochores in all of the mutants we examined (data not shown). Mitch localization is dependent, however, upon the presence of CID (the Drosophila orthologue of the centromere-specific H3-like protein CENP-A) and the DNA-binding centromere protein CENP-C. In the tissue culture cells treated with dsRNA corresponding to CID or CENP-C, in which the centromere localization of these proteins was eliminated, Mitch failed to reach the kinetochores (supplementary material Fig. S4).

Similarly, during mitosis in mitch¹ and mitch^19B mutants (whether in untreated, colchicine- or taxol-treated brains) we did not detect any deviations from wild-type in the localization of other centromere/kinetochore proteins, including BubR1 (Basu et al., 1999), Bub3 (Basu et al., 1998), Mad2 (Logarinho et al., 2004), 3F3/2 (Ahonen et al., 2005; Logarinho et al., 2004), Cenp-meta (Maiato et al., 2006; Williams et al., 2003; Yucel et al., 2000), Zwilch (Williams et al., 2003), CID (Blower and Karpen, 2001; Henikoff et al., 2001), dynein heavy chain [DHC (Li et al., 1994; Starr et al., 1998)], Klp10A (Rogers et al., 2004), Polo (Logarinho and Sunkel, 1998) and CENP-C (Heeger et al., 2005) (supplementary material Fig. S5). Thus, the absence of Mitch does not radically alter overall kinetochore structure. It is of further interest that some kinetochore components, such as Bub3 and DHC, localize to mono-oriented chromosomes near the poles in mitch mutants more strongly than to chromosomes that have congressed to the spindle equator (supplementary material Fig. S5B,F indicated by arrows). Because the levels of these proteins normally decrease dramatically upon bipolar attachment (Basu et al., 1999; Basu et al., 1998; Starr et al., 1998), this finding solidifies the previous conclusion that the non-congressed chromosomes are mono-oriented and are engaged in SAC signaling.

One would expect a protein that is crucial for chromatid segregation to be well-conserved during metazoan evolution. However, BLAST and PSI-BLAST searches using the amino acid sequence of Mitch have not found significant homologies outside of the genus Drosophila. Comparison of the mitch sequences of D. melanogaster and its close relative D. simulans (see Materials and Methods) confirms the surprisingly rapid evolution of this essential gene. Nonsynonymous nucleotide divergence (K_a) at the mitch locus between these two species is 3.8%, whereas the average K_a was found to be ∼1% in a survey of 45 random loci (V. DuMont and C.F.A., unpublished data).

One possible explanation for these observations is that positive selection drives the divergence of Mitch. Under this hypothesis, new amino acid variants at some positions in Mitch are favored by natural selection and, thus, tend to fix in a population more rapidly than do neutral or deleterious variants. The centromere-specific histone CID provides one well-known example of such adaptive evolution (Malik and Henikoff, 2001). In order to test this idea, we obtained mitch sequences from thirteen Drosophila species (supplementary material Fig. S6), and applied maximum-likelihood tests for positive selection of the same amino acids in different lineages as described in the Materials and Methods (see also Nielsen and Yang, 1998; Swanson et al., 2003). We were unable to reject the null-hypothesis that recurrent positive selection drives the evolution of a subset of codons in mitch.

We also investigated whether the rapid divergence at the mitch locus might partly be due to selection on different codons in different lineages, by asking whether there has been an excess of amino acid substitutions between D. melanogaster and D. simulans. We sequenced mitch from ten isofemale lines obtained from an African (Zimbabwe) population of D. melanogaster, and applied the McDonald-Kreitman test to these data using D. simulans as an outgroup (Begun and Aquadro, 1994; McDonald and Kreitman, 1991). This allowed us to compare the ratios of nonsynonymous to synonymous substitutions between species to the same ratio within D. melanogaster. We again found no evidence that selection has driven any of the amino acid sequence diversity in Mitch proteins.

To identify proteins that interact with Mitch, we tagged Mitch with both protein A and calmodulin-binding protein (CBP) and expressed this construct in stable transfectants of Drosophila Kc cells. The presence of the tagged protein at the kinetochore was verified with antibodies against the protein A tag (Fig. 8A). Through tandem affinity purification (TAP) techniques (Puig et al., 2001; Veraksa et al., 2005), followed by mass spectrometry of associated proteins, we found that Mitch primarily co-purified with subunits of the Ndc80 complex, in particular the Drosophila Hec1 (Ndc80) and Nuf2 homologues (Bharadwaj et al., 2004; McCleland et al., 2004; Meraldi et al., 2006) (Fig. 8B). In addition, purification of Mitch also pulled down minor amounts of the Drosophila Mis-12 homologue (Meraldi et al., 2006) and CG1558, the latter of which does not exhibit homology to known proteins. The identification of Mitch as a component of the Ndc80 complex is consistent with two-hybrid screen data that revealed associations between Mitch, Hec1 and Nuf2 (Giot et al., 2003), as well as with recent results from two independent affinity-purification studies of the same complex (Przewloka et al., 2007; Schittenhelm et al., 2007).

We describe here the discovery of a new kinetochore protein, Mitch. In the absence of Mitch, chromosomes fail to become correctly aligned during spindle assembly, and persistently mono-oriented chromosomes are often seen in the vicinity of the spindle poles. The aneuploidy-producing aberrations seen in mutants during anaphase and telophase are thus likely to be secondary consequences of earlier problems in kinetochore attachment and chromosome alignment.

The persistently mono-oriented chromosomes in mitch mutants could be of two classes. First, the sister kinetochores could be attached to MTs emanating from the same spindle pole. Normally these `syntelic' connections are corrected by machinery mediated by Aurora kinase that increases the turnover of kinetochore-MT attachments in the absence of tension across the centromere (reviewed in Cimini and Degrassi, 2005). Alternatively, the pole-associated mono-oriented chromosomes in mitch mutants could be `monotelic', with only one sister kinetochore connected to MTs and the other kinetochore being unattached. Our data do not allow us to distinguish between these two possibilities. It is, however, important to note that these two classes of mal-attachments are not mutually exclusive: the correction of syntely requires the detachment of one kinetochore from the spindle, creating a transient monotelic chromosome.

It is also unclear whether the absence of Mitch produces mono-oriented chromosomes directly or whether its absence prevents the correction of mal-attachments normally produced during spindle assembly. A precedent for the former possibility is offered by studies of a protein complex at S. cerevisiae kinetochores that includes Mtw1p, Dsn1p, Nnf1p and Nsl1p (Pinsky et al., 2003). Although the precise mechanism by which this complex operates is not known, it is clear that Aurora kinase produces unattached kinetochores in the absence of Mtw1p. The absence of Mitch might also directly promote mono-orientation simply by impairing MT binding by the kinetochore, particularly because Drosophila kinetochores attach on average to only five MTs in male meiosis (Lin et al., 1981) and to an average of only 11 MTs in S2 cells (Maiato et al., 2006). Such a reduction in MT-binding efficiency is thought to lead to the persistence of mono-oriented chromosomes in mammalian tissue culture cells injected with antibody against CENP-E (McEwen et al., 2001). Alternatively, Mitch might be involved in correcting the mono-orientations that occur early in mitosis through a mechanism whereby kinetochores that are originally unattached to kMTs can eventually nucleate and/or organize their associated kinetochore fibers (Maiato et al., 2004).

The relationship between Mitch and the SAC appears paradoxical, and suggests that the SAC is more complicated than previously envisioned. Mono-oriented chromosomes with unstable kinetochore attachments would be expected to maintain checkpoint activity – and this is what we see in both fixed and living non-drug-treated mitch mutant cells (Fig. 2 and Fig. 3A-C; Table 1). Surprisingly, however, the SAC is not functional in mitch neuroblasts exposed to colchicine or taxol, or when correct bipolar spindle assembly is compromised by lack of the MT-associated protein Asp. Under all of these conditions, mitch mutants prematurely disjoin their chromatids and exit mitosis (Fig. 1D,E; Table 1). That these cells enter anaphase prematurely is further indicated from the observations that cyclin B levels are low (Fig. 5A,B), and that Mad2 and Bub3 are much reduced or absent from their kinetochores (Fig. 5C-E). Indeed, direct live cell observations revealed that colchicine-treated mitch neuroblasts transit through mitosis with the same kinetics as untreated wild-type cells (Fig. 3D,E). A rapid escape from mitosis in the presence of spindle poisons is characteristic of Drosophila cells carrying mutations in genes encoding SAC components (Basu et al., 1999; Logarinho et al., 2004; Williams et al., 2003).

We conclude that Mitch is not directly required for SAC function, but it is required for maintaining a checkpoint signal if the spindle is perturbed. The inference that the SAC is in place but does not correctly signal under such conditions is consistent with our finding that checkpoint components like Bub3, BubR1 and Mad2 are associated with kinetochores in drug-treated mitch mutant cells before – but not after – sister chromatid separation (Fig. 5). One way to view this signaling is that Mitch plays a role in slowing the adaptation of cells to the checkpoint. Normal drug-treated cells eventually exit mitosis into the next G1; our data indicate that such adaptation to the SAC is accelerated when Mitch is absent.

The SAC monitors the accumulation of microtubules at the kinetochore and/or the resultant tension created across the centromere upon the attachment of sister kinetochores; recent results in Drosophila as well as in mammalian systems suggest that some checkpoint components sense MT occupancy whereas others track tension (Logarinho et al., 2004; Morrow et al., 2005). We found that the SAC fails to function in mutant brains treated with either colchicine or taxol. Since colchicine-treated cells are devoid of spindle MTs, kinetochores are by definition unattached and are not subject to tension. Kinetochores in taxol-treated cells are attached to the spindle but are not under tension (Waters et al., 1998). The most obvious common feature of colchicine and taxol treatments is the failure of microtubule subunits to become incorporated into, and removed from, kinetochore MTs. We thus posit that Mitch is required in the SAC only when the dynamic behavior of spindle MTs is compromised. Drug-induced disruptions in kinetochore MT dynamics are either not detected in mitch mutant cells, or the disruptions are detected but downstream events in the SAC signaling pathway are disrupted; in either case, mitch mutant cells would exit mitosis prematurely. Under this model, Mitch would not be required for SAC function in untreated cells because attached kinetochores have dynamic MT connections.

Mitch is not the only kinetochore-associated SAC component to display this paradoxical behavior. Other investigators previously found that in untreated larval brains of animals homozygous for mutations in cenp-meta (which encodes one of two Drosophila proteins related to the mammalian kinesin-like protein CENP-E), mitotic progression is delayed in a prometaphase-like state (i.e. the cells contain an active SAC) (Yucel et al., 2000). Importantly, however, we have shown that the SAC is not functional in cenp-meta mutants treated with colchicine (Williams et al., 2003). These similarities between Mitch and Cenp-meta suggest that the effects of mitch mutants might be mediated through Cenp-meta. In fact, since CENP-E has recently been implicated in a novel mechanism that allows the congression of mono-oriented chromosomes (Kapoor et al., 2006), the persistent mono-orientation of chromosomes in mitch mutants could in theory also result from a failure of Cenp-meta to bind to, or function correctly at, the kinetochores. The former possibility appears to be excluded by our finding that Cenp-meta is targeted to kinetochores in mitch mutants (supplementary material Fig. S5G), but it is not inconceivable that almost all aspects of the mitch mutant phenotype could still be explained by indirect effects on Cenp-meta function.

Our inability to find Mitch homologs outside of the genus Drosophila and our observation that nonsynonymous nucleotide divergence is considerably higher than average at the mitch locus led us to apply tests for positive selection of this locus within the Drosophilids. We found no evidence for positive selection on a subset of codons in Mitch, as a model of sequence evolution that allows for positive selection does not perform significantly better than models that do not allow for positive selection. Similarly, we found no evidence that selection has accelerated rates of amino acid substitution between the two species D. melanogaster and D. simulans. Divergence at this locus may therefore be the consequence of relatively weak functional constraints on the amino acid sequence of the Mitch protein.

We have found that Mitch associates strongly with Hec1 and Nuf2 (components the Ndc80 complex) and to a lesser extent with a component of the Mis12-MIND complex (Mis12). Since Hec1 and Nuf2 in other organisms are associated with the proteins Spc24 and Spc25 (Wei et al., 2006), it appears likely that Mitch corresponds functionally with Spc24 or Spc25 even though such a relationship is not obvious at the amino acid sequence level (see also Schittenhelm et al., 2007). The COILS2 program (Lupas et al., 1991) reveals that Mitch in all analyzed Drosophila species contains a coiled-coil domain in the N-terminal half, a characteristic also seen in Spc24 and Scp25 in other species. It will thus be of interest in the future to establish whether mutations in other components of the fly Ncd80 complex exhibit the same paradoxical behavior of the SAC we observed in mitch mutants.

If Mitch is indeed an ortholog of Spc24 or Spc25, it is surprising that the contradictory behavior of the spindle checkpoint described has not previously been noted. In yeast and vertebrate cells, spc24 and spc25 mutations also result in defective chromosome congression and alignment (Bharadwaj et al., 2004; Janke et al., 2001), presumably due to defects in kinetochore-microtubule attachments (McCleland et al., 2004). However, in budding yeast, spc24 and spc25 mutants are checkpoint defective, with spc24 mutants failing to arrest in the presence or absence of nocodazole (Janke et al., 2001). Spc25-depleted human cells retain the spindle checkpoint, even though the localization of MAD1 is affected (Bharadwaj et al., 2004). Xenopus Spc24 and Spc25 depletion results in delocalization of Mad1 and Mad2 from the kinetochore, and abrogation of the spindle checkpoint in normally progressing cells as well as those treated with MG132, a proteasome inhibitor which normally causes metaphase arrest (McCleland et al., 2004). We find it remarkable that the relationship between these core kinetochore proteins and the spindle checkpoint is so variable in different lineages, perhaps reflecting the rapid evolution in the amino acid sequences of Spc24 and Spc25.

Strains bearing lethal mutations on the third chromosome were balanced over TM6 or TM6B carrying Tubby (Tb) and either Stubble (Sb) or Humeral (Hu). These markers allowed us to distinguish balancer-carrying adults (Sb or Hu) and larvae (Tb) from mutant homozygotes or trans-heterozygotes. The wild-type strain used in these experiments was Oregon-R. Other stocks were obtained from the Bloomington (IN) stock center.

To isolate mitotic mutants, our laboratory screened 1583 strains from the laboratory of Charles Zuker (University of California, San Diego) that carried ethylmethanesulfonate (EMS)-induced mutations on the third chromosome causing lethality during third instar larval or pupal stages (Koundakjian et al., 2004). One of these stocks contained the original mitch¹ allele.

Additional mutant alleles containing small deletions of mitch (CG7242) were derived as imprecise excisions of the homozygous viable P element insertion CK4A, which is inserted between mitch and the adjacent gene granny smith (grsm; CG7340; supplementary material Fig. S3). The CK4A insertion was itself obtained by mobilizing the P-element in l(3)S125006a (Deak et al., 1997) using Δ2-3 transposase. The CK4A P element was then remobilized with Δ2-3 transposase, and the resulting progeny were tested for failure to complement the deficiency Df(3R)ry75 (Fig. 6) and mitch¹.

To delimit the genomic region containing mitch, we first determined the relevant breakpoints of the deletions Df(3R)ry1608 and Df(3R)ry614. This was accomplished by Southern blot analysis as previously described (Williams et al., 2003), and by using a single embryo PCR technique we have detailed elsewhere (Yu et al., 2004). To identify the nucleotides altered in presumptive mitch point mutations, genomic DNA fragments representing the exons of the five candidate genes in the region of interest (supplementary material Fig. S3) were PCR amplified in duplicate and sequenced from homozygous mitch¹ and mitch² animals.

The extent of the small deletions of mitch obtained as imprecise excisants of CK4A (see above; supplementary material Fig. S3) was determined by PCR using primers in CG7242 and CG7340. PCR products spanning deletion breakpoints were identified as new bands not found in amplifications of wild type or CK4A genomic DNA. These PCR products were then sequenced to localize the deletion breakpoints at the nucleotide level.

To amplify the mitch locus from other Drosophila species, we compared the sequences of the corresponding genomic regions in D. melanogaster and D. pseudoobscura to find short, highly conserved regions both internal to mitch and also in the flanking genes CG7340 and CG7242 that could be used as PCR primers. Genomic DNA from D. simulans, D. mauritiana, D. teissieri, D. sechellia, D. erecta, D. yakuba, D. orena, D. eugracilis, and D. lutescens were prepared and amplified using these primers and Taq polymerase (Fermentas, Hanover, MD). Additional, species-specific primers were also required in certain PCR amplifications. The DNA sequences of these primers, as well as PCR reaction parameters, are available upon request.

All primers were ordered from Integrated DNA Technologies (Coralville, IA). Amplified bands were extracted and purified from agarose gels (Qiagen, Valencia, CA), and sequenced by the BioResource Center (Cornell University, Ithaca, NY). Each gene was sequenced on both strands and from multiple, independently performed PCR amplifications. Additional mitch coding sequences were compiled using Seqman (DNAStar, Madison, WI) from the genomic sequences of D. pseudoobscura, D. mojavensis and D. virilis available at NCBI.

We used PAML (Nielsen and Yang, 1998) to compare models of sequence evolution. This analysis included all the species in the subgenus Sophophora shown in supplementary material Fig. S4, but did not include the species in the subgenus Drosophila (that is, D. virilis and D. mojavensis) because of uncertainties in amino acid alignment between more distant species. In each of two comparisons, one model (M8) allows for, but does not require, positive selection on a subset of codons, as indicated by a d_N/d_S>1. The second model, the null model, does not allow the ratio d_N/d_S to exceed one. Two null models, M7 and M8A, were used. In order to infer positive selection, two conditions must be met. First, the sequence data must fit M8 significantly better than the null model. Second, M8 must find evidence for a class of codons with d_N/d_S>1.

We obtained a full-length CG7242 cDNA clone (LD37196; Research Genetics, Athens, GA), and PCR-amplified a fragment containing the entire open reading frame. The PCR primers introduced a recognition site for the restriction enzyme XbaI at the cDNA's 5′ end and a HindIII site at the 3′ end. Restriction enzyme-digested PCR product was then cloned into the vector pMAL-C2 (New England Biolabs, Beverly, MA) digested with the same enzymes. Native and denatured fusion proteins were purified, and antibodies to these proteins were raised in rabbits and guinea pigs as previously described (Williams et al., 2003).

Although all four preparations of anti-Mitch antibodies work well in immunofluorescence experiments (see below), they are at best mediocre reagents as probes for western blotting using standard methodology. The antibodies recognized the Mitch protein when it is overexpressed in E. coli or in Kc Cells, and we occasionally saw a very faint band of the predicted size (26 kDa) in blots of wild-type larvae that disappeared in blots of mitch mutants. A band of approximately the same size was observed in extracts of Drosophila Kc line tissue culture cells that were transfected with constructs that overexpress Mitch under the control of the actin 5C promoter (data not shown). The weakness and inconsistencies in the western blots obtained with anti-Mitch antibodies as probes could be caused either by low levels of native Mitch protein expression, or by the relative inability of this protein to bind to the PVDF membranes (Millipore Corp., Billerica, MA) we employed.

To express V5-epitope tagged Mitch protein in Drosophila tissue culture cells, mitch sequences were amplified from cDNA LD37196 with primers corresponding to the 5′UTR (5′-CGGAATTCTCTGGAAAACGGCGCTTAG-3′) and the end of the coding region (5′-AGATCTAGAGGTGTGGCTCATCGGCGA-3′) using Taq polymerase (Fermentas, Hanover MD). pAC5.1/V5-HisB (Invitrogen, Carlsbad CA) was digested with EcoRV and T-tailed using Taq polymerase in the presence of dTTP, followed by ligation (T4 DNA ligase; Fermentas) with the mitch fragment. The pAC-mitch-v5-his construct was verified by DNA sequencing. Plasmid DNA was prepared by alkaline lysis (Qiagen) and transfected into Kc cells. For each transfection, 3 μg of plasmid DNA was mixed with 10 μl of Cellfectin (Invitrogen) in HyCCQ medium (HyClone, Logan UT) and used to transfect 3 ml of log-phase Kc cells, which were harvested for western blotting and immunofluorescence after 48 hours.

RNAi was performed by treating Kc tissue culture cells with dsRNA corresponding to CID,CENP-C or mitch according to standard procedures (Yu et al., 2004). For CENP-C and mitch, cells were incubated with dsRNA for 4 days before fixation, whereas CID depletion occurred effectively only after 7 days of exposure to dsRNA, with additional dsRNA added after 3 days. Cells were processed for immunofluorescence as described below for brains, or for chromosome spreads according to (Yu et al., 2004).

We studied the mitch phenotype of the mutant alleles by staining third-instar larval neuroblasts with orcein and by immunolocalization with the anti-Mitch antibodies, using previously described methods (Williams and Goldberg, 1994). Brains were fixed, squashed and immunostained using the following antibodies: mouse anti-alpha-tubulin (Amersham Corp., Arlington Heights, IL), rabbit anti-Zw10 (Williams et al., 1992), rabbit anti-cyclin B and CENP-C (gifts of Christian Lehner, University of Bayreuth, Germany), chicken anti-CID (Blower and Karpen, 2001), anti-Klp10A (Rogers et al., 2004), rabbit anti-dynein heavy chain (Starr et al., 1998), rabbit anti-SCC1 (Warren et al., 2000), rabbit anti-BubR1 and anti-Bub3 (Logarinho et al., 2004), and anti-Mad2 (Logarinho et al., 2004). Antibodies were raised in rabbits against Cenp-meta (Cocalico Biologicals, Reamstown PA) using protein produced in E. coli from pQE-cmet (Yucel et al., 2000). Drosophila Kc tissue culture cells were fixed by a procedure identical to that used for larval brains (Williams and Goldberg, 1994), and stained with mouse anti-V5 antibody (Invitrogen). Secondary antibodies (all from Jackson Laboratories, West Grove, PA) were TRITC (tetramethylrhodamine isothiocyanate)-conjugated anti-rabbit IgG, TRITC anti-chicken IgY, Cy2 anti-guinea pig IgG, and TRITC anti-mouse IgG, all at 0.1 mg/ml final concentration and incubated overnight at 4°C. DNA was stained with Hoechst 33248 dye at a final concentration of 0.5 μg/ml. Samples were examined by epifluorescence, and images were obtained by using a Pentamax CCD camera (Roper Scientific, Tuscon, AZ) attached to an Olympus BX50 microscope.

Homozygous phenotypes were studied by dissecting Tb⁺ larvae for all mutant alleles except for mitch², for which Tb⁺ larvae could not be obtained. We assume the chromosome bearing this allele carries an additional lethal mutation from the original EMS mutagenesis. To study the mutant phenotype of mitch², we instead dissected the brains of mitch²/Df(3R)ry614 larvae.

For analysis of mitotic progression in live Drosophila neuroblasts, brains from third instar larva were dissected and processed as described previously (Fleming and Rieder, 2003; Savoian and Rieder, 2002) with the following modifications. The dissected brain was put on a coverslip with a small drop of neuroblast culture medium supplemented with 20% FBS. A small square piece from an agarose layer was gently placed on top of the brain monolayer. This coverslip was then flipped onto a microscope glass slide containing two coverslip fragments that act as spacers. A gradient of flatness was then created by allowing the medium covered by the agar overlay to slip underneath one of the spacers and leaving the other dry. This enables the experimenter to select a degree of flatness ideal for high-resolution LM analyses that does not inhibit progression through mitosis or cytokinesis. The edges of the coverslip were then sealed with vasoline:lanolin:paraffin (VALAP; 1:1:1) to prevent evaporation. All brains used were still viable at the moment the recordings were stopped, in some cases after more than 6 hours. Three brains from the wild-type strain Oregon R were used as controls to record six neuroblasts, and six brains from mitch mutants were used to record ten neuroblasts. Images of neuroblasts were acquired at room temperature using a Deltavision Restoration Microscopy System mounted on an Olympus IX70 DIC inverted light microscope. Single image planes were acquired every 30 seconds using a Roper CM350 CCD camera.

The entire coding sequence of mitch was cloned into pMK33-NTAP and pMK33-CTAP (Veraksa et al., 2005) and the resulting constructs were stably transfected into Drosophila Kc tissue cells using Cellfectin (Invitrogen). TAP-Mitch was assayed on western blots using HRP-conjugated anti-protein A antibody (Rockland, Gilbertsville PA) and by immunofluorescence of fixed cells using goat anti-protein A antibody followed by TRITC-conjugated anti-goat antibody (Jackson Laboratories). Protein complexes from one liter of TAP-Mitch stable line cells were isolated following (Puig et al., 2001), using a lysis buffer for making Drosophila extracts (Veraksa et al., 2005). After purification using IgG-Sepharose and Calmodulin-Sepharose beads (Invitrogen), the final eluate was precipitated with trichloroacetic acid (TCA), resolubilized in Laemmli sample buffer (Bio-Rad, Hercules CA) and subjected to SDS-PAGE. Bands were excised, trypsinized and analyzed by MALDI (Cornell Bioresource Center.)

We thank Charles Zuker and Edmund Koundakjian for sharing their collection of mutagenized Drosophila stocks. Claudio Sunkel, Paula Sampaio, Gary Karpen, Christian Lehner, Gary Gorbsky, Don Cleveland, Alexey Veraksa and Greg Rogers generously provided us with antibodies, fly stocks, and other reagents. We acknowledge the participation of Leah Colvin and Ruth Ann Riel in the generation of deletions via P element mobilization. This work was supported by NIH grants GM48430 to M.L.G. and GM40198 to C.L.R.; work in the laboratory of C.K. was supported by funding from the Viking Children's Fund and the Northland Affiliate of the American Heart Association. H.M. was a recipient of a postdoctoral fellowship SFRH/BPD/11592/2002 from Fundação para a Ciência e a Tecnologia of Portugal. The equipment used for the time-lapse photography in this study is part of the Wadsworth Center's Video-Enhanced LM core facility.