Chromosome lagging at anaphase and migration of both sister chromatids to the same pole, i.e. nondisjunction, are two chromosome-segregation errors producing aneuploid cell progeny. Here, we developed an assay for the simultaneous detection of both chromosome-segregation errors in the marsupial PtK1 cell line by using multiplex fluorescence in situ hybridization with specific painting probes obtained by chromosome flow sorting. No differential susceptibility of the six PtK1 chromosomes to undergo nondisjunction and/or chromosome loss was observed in ana-telophase cells recovering from a nocodazole- or a monastrol-induced mitotic arrest, suggesting that the recurrent presence of specific chromosomes in several cancer types reflects selection effects rather than differential propensities of specific chromosomes to undergo missegregation. Experiments prolonging metaphase duration during drug recovery and inhibiting Aurora-B kinase activity on metaphase-aligned chromosomes provided evidence that some type of merotelic orientations was involved in the origin of both chromosome-segregation errors. Visualization of mero-syntelic kinetochore-microtubule attachments (a merotelic kinetochore in which the thicker microtubule bundle is attached to the same pole to which the sister kinetochore is connected) identified a peculiar malorientation that might participate in the generation of nondisjunction. Our findings imply random missegregation of chromosomes as the initial event in the generation of aneuploidy in mammalian somatic cells.
Missegregation of duplicated chromosomes within a bipolar mitosis produces aneuploid cells, i.e. cells showing a gain or loss of entire chromosomes. One well-known segregation error is chromosome lagging, occurring when a single chromatid remains at the spindle equator after chromosome migration because its kinetochore is connected to both spindle poles through a merotelic attachment (Cimini et al., 2001). Another segregation error is nondisjunction, i.e. the migration of both sister chromatids to the same pole. The term nondisjunction, originally referring to genetically identified meiotic missegregation, has been commonly used to imply a failed separation of homologous chromosomes or sister chromatids at the centromere. However, this idea has been brought into question by fluorescence in situ hybridization (FISH) data showing that co-migration of sister chromatids to the same pole occurs in the vast majority of the cases after centromere separation in somatic cultured cells (Cimini et al., 1999). Furthermore, precocious sister-chromatid separation due to cohesion defects has been shown to end up as sister-chromatid co-migration in yeast (Tanaka et al., 2000; Kitajima et al., 2004) and premature centromere splitting generates cell aneuploidy in the mosaic variegated aneuploid (MVA) syndrome (Kajii et al., 2001), all implying that there are mechanisms different from retained cohesion in the origin of nondisjunction. In relation to genetic consequences, nondisjunction produces one trisomic (three copies of a specific chromosome) daughter cell and a monosomic (only one copy of a specific chromosome) one; most lagging chromosomes will form a micronucleus (a chromatin body separated from the main nucleus) producing a monosomic daughter cell with or without a micronucleus in half of the cases (Cimini and Degrassi, 2005). Finally, micronuclei can be incorporated in one of the daughter nuclei at the following mitosis producing further unbalanced karyotypes (Rizzoni et al., 1989). Monosomic and trisomic cells might constitute different threats to cell populations and individuals. All monosomic and most trisomic gametes produce nonviable embryos and are, therefore, an important cause of spontaneous miscarriages and stillbirths in humans. However, chromosome 13, 18 or 21 trisomies are compatible with life and are the cause of severe genetic syndromes (Hassold and Hunt, 2001). In addition, aneuploidy is a nearly ubiquitous feature of human cancers and it has been suggested to play a crucial role both in tumour development and progression (Rajagopalan and Lengauer, 2004; Weaver and Cleveland, 2006). Most cancer cells display a complex pattern of chromosomal abnormalities, showing both chromosome losses and gains, and concomitant structural aberrations (http://cgap.nci.nih.gov/Chromosomes/Mitelman). Interestingly, recurrent gains or losses of specific chromosomes or chromosome regions (Holloway et al., 1999; Streblow et al., 2007; Vasuri et al., 2008) are frequently observed in tumours, although the biological significance of these aneusomies in cancer is hard to establish (Sen, 2000; Albertson et al., 2003). In fact, it is still unknown whether specific chromosomes have different propensities to undergo errors in chromosome segregation or whether their presence in cancer cells is the result of cellular selection of cells bearing random aneuploidy.
In mammalian cell cultures, chromosome loss is observed at low frequencies and its frequency is enhanced in cells recovering from a mitotic arrest induced by microtubule-interacting drugs (Cimini et al., 2001). Much less is known about nondisjunction. Unequivocal detection of nondisjunction requires a rather complicated assay, in that segregation of individual chromosomes has to be followed by in situ hybridization on ana-telophase cells. So, very few studies have investigated nondisjunction in human cells; however, those that have nevertheless suggest that nondisjunction occurs at frequencies higher than chromosome loss (Cimini at al., 1999; Thompson and Compton, 2008). Merotelic kinetochores have been suggested to produce nondisjunction when more microtubules on the merotelic kinetochore are connected to the same pole to which its sister is attached (Salmon et al., 2005). However, due to technical difficulties in visualizing merotelic attachments, this hypothesis has remained unproven.
To address these issues, we developed an assay for the simultaneous detection of chromosome loss and nondisjunction for all chromosome pairs in the marsupial PtK1 cell line. This cell line was chosen because PtK1 cells have been the model system for studies on chromosome loss at mitosis and possess a very low chromosome number (2n=12), making easier the analysis of chromosome segregation for individual chromosomes on anaphase cells. Segregation of individual chromosomes was followed using `multiplex FISH' (M-FISH) (Speicher et al., 1996), a special kind of chromosome painting that has been applied to unambiguously differentiate chromosomes in several marsupial species (Rens et al., 2003a; Rens et al., 2006).
Results and Discussion
For the simultaneous detection of chromosome loss (CL) and nondisjunction (ND) PtK1 chromosome-specific painting probes were developed from flow-sorted chromosomes. PtK1 chromosome preparations produced six different peaks in the flow karyotype, corresponding to the six chromosome pairs (Fig. 1A). PtK1 chromosome-specific paints with different colours were generated through high purity sorting and degenerate oligonucleotide primed (DOP)-PCR amplification incorporating fluorescence tagged dUTPs. These paints were applied in M-FISH experiments on metaphase cells to identify each of the six marsupial chromosome pairs in a single metaphase spread by a different colour (Fig. 1B). Furthermore, M-FISH analysis on metaphase cells showed that the small chromosome 5 was present in just one copy in this PtK1 cell line, as visible in the reconstructed karyotype (Fig. 1B).
Missegregation events do not occur at a different rate for specific chromosomes
Chromosome segregation was then evaluated by M-FISH analysis on ana-telophase cells. The high specificity of the chromosome paints rendered it possible to visualize the position occupied by each chromosome in the anaphase cell, with the largest chromosome, chromosome 1, protruding towards the periphery of the group of migrated chromosomes and the smallest chromosome, chromosome 5, in a central position surrounded by the other chromosomes (Fig. 2A). Chromosome misdistribution was also easily identified using M-FISH. The lagging of chromosomes 1 and 3 between the two groups of chromosomes (Fig. 2B, arrows) as well as the presence of three copies of chromosome 4 at one pole (Fig. 2B, arrowhead), i.e. an event of ND, could be appreciated by analyzing the images from the individual filters. With this approach we set out to investigate chromosome-specific differences in missegregation rates. ND and CL are rare events in untreated cells; therefore, no clear differences among chromosomes were observed in untreated cultures, although the extremely low number of events did not allow to draw any statistically sound conclusion (data not shown). A treatment with the anti-mitotic agents nocodazole (NOC; Fig. 2C) or monastrol (MON; Fig. 2D) was used to increase the frequencies of ND and CL in ana-telophases harvested after drug release. To assess whether specific chromosomes were involved at different rates, we first analyzed the deviation from the expected value of ND or CL calculated by assuming that all chromosomes responded equivalently (specific chromosome ND = total ND/6). The number of ND and CL events observed for each chromosome pair was then compared with the expected value in a χ2 test. This analysis showed no significant deviations from the expected values for both drugs, indicating that no chromosome was preferentially involved in ND or CL. To further investigate this, we performed a different data analysis comparing missegregation rates between each chromosome and every other one. This analysis revealed that chromosomes 1 and 5, which represent the largest and smallest chromosome in PtK1 cells, respectively, showed lower ND frequencies compared with those observed for all other chromosomes after NOC treatment (chromosome 1: P<0.05; chromosome 5: P<0.01). This low representation of the two chromosomes was not observed for CL: in this case, chromosomes 1 and 5 did not show any significant difference from the other chromosomes. Analysis of chromosome susceptibility to missegregate after MON treatment showed that chromosome missegregation rates were homogeneous for all of the chromosomes, the only exception being that the CL frequency for chromosome 4 was significantly lower than that observed for the other chromosomes (P<0.01). In conclusion, statistically significant differences between chromosomes were limited and did not permit the identification of any chromosome-specific tendency because they were not homogeneous neither for the chromosome involved, nor for the type of missegregation analyzed (ND vs CL), nor were they related to a specific inducing drug. Collectively, these data indicate that the six PtK1 chromosomes undergo ND and/or CL to the same extent. Furthermore, they suggest that correction of kinetochore malorientations, which involves the Aurora-B-dependent phosphorylation of the depolymerizing kinesin MCAK (Kline-Smith et al., 2004; Lan et al., 2004; Knowlton et al., 2006; Andrews et al., 2004) together with the activity of the Kif2b kinesin (Bakhoum et al., 2008), and microtubule flux (Ganem et al., 2005), is equally efficient on the different chromosomes, producing random chromosome missegregation at mitosis.
Among the many factors that have been proposed to influence the susceptibility of individual chromosomes to missegregate, chromosome size and chromosome position within the mitotic spindle are regarded as crucial. Our data on anaphase cells (Fig. 2A; supplementary material Fig. S1) as well as previous live imaging data on the same cell line (Cimini et al., 2004) provide evidence that larger chromosomes tend to position at the metaphase-plate periphery, possibly owing to a steric hindrance of the large chromosome arms. In this area of the spindle, chromosome oscillations tend to be shorter and less frequent (Khodjakov and Rieder, 1996; Cimini et al., 2004), which might be due to reduced directional instability mediated by accumulation of the Kif18a kinesin on peripheral kinetochores (Stumpff et al., 2008). Reduced chromosome oscillations could theoretically decrease the correction of erroneous kinetochore-microtubule attachments, which requires the release of microtubules from kinetochores and their successive interaction at a different angle (Lampson et al., 2004). Our data indicate that chromosome oscillations at the metaphase plate do not affect missegregation rates of individual chromosomes but rather suggest a similar correction efficiency among the different kinetochores.
It is well known that random chromosome missegregation occurs in meiosis. Aneuploidies for nearly all chromosomes are represented among spontaneous abortions at very early stages of development, but only trisomies for a few chromosomes are vital at birth, demonstrating that meiotic aneuploidies are subjected to prenatal selection (Hassold and Hunt, 2001). Our data suggest a similar scenario for mitotic cells, in which the recurrent presence of specific chromosomes in several cancer types is not the result of a different propensity of specific chromosomes to undergo missegregation but is the outcome of a selection process that depends on the identity of the extra chromosome. In line with this, a recent study has shown that aneuploid murine embryonic fibroblasts (MEFs) containing an extra copy of only one chromosome exhibit decreased proliferation rates and decreased cell fitness (Williams et al., 2008). However, the timing of cell immortalization, one of the early steps in cell transformation in culture, depended on which additional chromosome was present in the different cell lines (Williams et al., 2008). Thus, it can be hypothesized that, although the presence of an extra chromosome is generally detrimental, a limited number of specific unbalances of gene products caused by individual chromosome aneuploidy, possibly in conjunction with additional mutations in oncogenes and/or tumour suppressors, might promote tumorigenesis.
Merotelic attachments and nondisjunction
Having established that no preferential involvement in chromosome missegregation occurs for specific chromosomes, we then wished to obtain new insights into the mechanism(s) underlying ND. For correct segregation to occur, sister kinetochores must attach to microtubules connected to opposite poles at metaphase. This so-called amphitelic orientation satisfies the mitotic-checkpoint requirements and allows the segregation of the two sisters to opposite poles. Other orientations are also possible and are transiently present during prometaphase, including monotelic (a chromosome with only one kinetochore attached to microtubules), syntelic (both sister kinetochores connected to the same pole) and merotelic orientations, the latter being the cause for chromosome lagging when persisting at anaphase (Cimini and Degrassi, 2005). To gain insight into the role of these different orientations in the generation of ND, we took advantage of the different mechanisms of action of NOC and MON. NOC is a well-known spindle poison capable of disassembling the mitotic spindle. Following NOC removal, the mitotic spindle reassembles through both microtubule nucleation from centrosomes and incorporation into the centrosomal-based microtubule arrays of microtubules nucleated at kinetochores (Tulu et al., 2006; Torosantucci et al., 2008). During this process, several extra-centrosomal foci of α-tubulin are present at very short times after NOC removal (Cimini et al., 2003; Tulu et al., 2006) (supplementary material Fig. S2); these foci favour the formation of high numbers of incorrect kinetochore-microtubule interactions, in large part merotelic attachments. MON is a small-molecule inhibitor of the mitotic kinesin Eg5 that is able to arrest cells in mitosis with monoastral spindles; chromosomes in MON-treated cells frequently exhibit sister kinetochores attached to microtubules in a syntelic orientation (Kapoor et al., 2000). Upon MON removal, syntelic orientations are transformed into amphitelic ones through shortage and severing of the syntelic K-fibres and successive attachment of other microtubules at the kinetochores (Lampson et al., 2004) (supplementary material Fig. S2). During this process, several types of mis-attachment are produced.
Missegregation for all PtK1 chromosomes was analyzed in untreated cells and in cells undergoing ana-telophase 45 or 60 minutes after NOC or MON release (Fig. 3). Approximately 6% of untreated PtK1 ana-telophase cells underwent chromosome missegregation (35 events/569 cells), showing 4.0% ND and 2.1% CL when controls from MON and NOC experiments were pooled together. This suggests that ND occurs at higher rates in comparison with CL in somatic mammalian cells, as previously observed in human primary fibroblasts (Cimini et al., 1999). ND and CL were efficiently induced over control values in cells harvested 45 minutes after NOC release (Fig. 3A) (P<0.005) with an approximate twofold induction of ND in comparison with CL (28.1% ND vs 12.6% CL). Ana-telophases recovering from MON treatment exhibited similar rates of missegregation (19.6% ND vs 13.1% CL), demonstrating that MON also efficiently induces chromosome missegregation (Fig. 3B) (P<0.001). Cells that were harvested 60 minutes after drug release displayed lower frequencies of CL and ND, both after NOC (Fig. 3A) and after MON (Fig. 3B) treatment, suggesting that the correction of faulty attachments during metaphase decreases the frequencies of CL and ND for both drugs. Merotelic kinetochore attachments occur frequently in early mitosis and are partially corrected before anaphase onset when metaphase is prolonged using the proteasome inhibitor MG132 (Cimini et al., 2003). We also delayed anaphase onset by MG132 treatment to determine whether this had an effect on ND. The frequencies of ND were efficiently decreased by incubation in MG132 medium during recovery from either drug (P<0.001 and P<0.05 at 45 and 60 minutes, respectively), although they never reached control levels (Fig. 3A,B). CL frequencies were also decreased (Fig. 3A,B), although to a lesser extent (P<0.05 when data from both times were pooled together). Altogether, these data indicate that a fraction of the faulty attachments responsible for ND is corrected before anaphase, as already observed for lagging chromosomes (Cimini et al., 2003).
Correction of merotelic attachments before anaphase requires Aurora-B kinase activity as part of an inter-kinetochore tension-sensing mechanism (Kline-Smith et al., 2004; Lan et al., 2004; Knowlton et al., 2006; Andrews et al., 2004; Cimini et al., 2006). We therefore investigated whether Aurora-B activity was responsible for the observed time-dependent correction of ND by using hesperadin, a known Aurora-B inhibitor (Hauf et al., 2003). The inhibitor was used at a dose (100 nM) that only partially suppressed Aurora-B activity, as determined by histone H3 (Ser10) phosphorylation (Fig. 4A,B), because partial inhibition of the kinase activity had already been shown to increase merotelic orientations and lagging chromosomes at anaphase without influencing other aspects of mitosis such as checkpoint efficiency or cytokinesis (Cimini et al., 2006). To investigate whether merotelic attachments are involved in ND, cells were released from a NOC-induced mitotic arrest in MG132 medium and hesperadin was added 1 hour later. In such a way, Aurora-B activity was inhibited when chromosomes had already reached the metaphase configuration in the vast majority of the cells (supplementary material Fig. S3), a moment when syntelic attachments, which maintain chromosomes close to the poles (Cimini et al., 2002; Cimini and Degrassi, 2005), were already corrected. Cells were kept in MG132 and hesperadin for a further 1 hour and then allowed to progress to anaphase by medium change. Inhibition of Aurora B in this time window very efficiently increased CL frequency above the value observed with MG132 alone (Fig. 4C), as already reported (Cimini et al., 2006). Interestingly, correction of improper attachments generating ND was also inhibited, as shown by the increased frequencies of ND above the MG132 value (Fig. 4C) (P=0.126). These data indicate that improper kinetochore-microtubule attachments generating ND are present on aligned chromosomes and that Aurora-B activity is required for their correction.
So, which type of kinetochore-microtubule interactions produces ND? Anaphase onset in the presence of syntelic attachments would produce chromosome ND, because both sister kinetochores are connected to the same pole. However, several lines of evidence converge to demonstrate that syntelic attachments activate the mitotic checkpoint (Pinsky and Biggins, 2005; Porter et al., 2007), rendering it less likely that syntelic chromosomes will undergo ND in checkpoint-proficient cells. In our data, the high ND frequencies obtained after inducing large amounts of merotelic orientations (NOC treatment) (Fig. 3A) as compared with preferentially inducing syntelic orientations (MON treatment) (Fig. 3B), as well as the effect of Aurora-B inhibition on aligned chromosomes (Fig. 4C), suggest that kinetochore-microtubule interactions that are different from syntelic attachments, which localize chromosomes close to the spindle poles, also contribute to ND. Monotelic orientations are very unlikely to generate ND. Even if monotelic chromosomes were able to reach the metaphase plate by sliding on already existing K-fibres (Kapoor et al., 2006), the unattached kinetochore would strongly activate the mitotic checkpoint, blocking anaphase onset. Merotelic chromosomes congress to the metaphase plate (Cimini et al., 2002). They have been shown to maintain connection to the two microtubule bundles during anaphase and lag at the cell equator when the two bundles are of similar size (balanced merotelic) or move to the pole connected to the thicker microtubule bundle (Cimini et al., 2004). In principle, a merotelically oriented kinetochore, in which the thicker microtubule bundle is connected to the same pole to which the sister is attached, would segregate to the same pole as the sister. These so-called `mero-syntelic' attachments have been shown to be responsible for ND in live crane-fly meiosis (Janicke at al., 2007), a favourable system to detect kinetochore-microtubule interactions, owing to the large dimension of the cells. To identify these different merotelic attachments in PtK1 cells, we analyzed Ca2+-resistant kinetochore-microtubule interactions after z-stack acquisition and image deconvolution of prometa-metaphase cells, which were imaged to visualize the centromere, the outer kinetochore region and microtubules using CREST, Hec1 and tubulin antibodies, respectively (Fig. 4D). In cultured mammalian cells, clear identification of kinetochore-microtubule interactions is restricted to the peripheral area of the spindle because of the high density of microtubules into the central area. Taking this into account, we were nevertheless able to show that, in addition to syntelic and amphitelic attachments (not shown), both balanced merotelic attachments, mero-amphitelics (a merotelic kinetochore with the thicker MT bundle connected to the pole opposite to that of the sister kinetochore) and mero-syntelics (the thicker microtubule bundle towards the pole to which the sister kinetochore is connected) are present during recovery from NOC (Fig. 4D). In our microscopic analysis, about 10% of aberrant orientations were mero-syntelics (4/45). So, by visualizing mero-syntelic attachments we identified a peculiar kinetochore-microtubule interaction that might be involved in the origin of ND in mammalian somatic cells. However, further high-resolution live-cell studies or electron-microscopy evidence are required to unequivocally identify mero-syntelics as being responsible for ND.
In conclusion, our M-FISH studies in conjunction with the analysis of kinetochore-microtubule interactions provide evidence indicating that ND and CL are generated by different types of merotelic attachments. These might originate during the extremely dynamic process of centrosome separation and microtubule growth that occurs during recovery from drug treatment. The position of centromeres in relation to centrosomes during the initial stages of this process seems to be a stronger determinant for the formation of incorrect interactions with respect to structural differences at kinetochores among the different chromosome pairs, as we did not observe a preferential involvement of specific chromosomes in segregation errors. Finally, our findings imply random missegregation of chromosomes as the initial event in the generation of aneuploidy and in the acquisition or loss of specific chromosomes in cancer.
Materials and Methods
Cell culture and treatments
PtK1 cells were grown as described (Rens et al., 1999). NOC, MON and MG132 (Sigma-Aldrich) were dissolved in DMSO. To promote chromosome missegregation, cells were incubated in 2 μM NOC for 3 hours or 100 μM MON for 4 hours. Thereafter, cells were washed three times in pre-warmed medium to release the mitotic block, re-incubated in fresh medium and fixed at the indicated times after drug release. In MG132 experiments, 15-25 minutes after NOC or MON release, cells were incubated in 7 μM MG132 for 2 hours, then washed three times for 5 minutes in pre-warmed medium and harvested after a further 45 or 60 minutes. 100 nM of the Aurora-B kinase inhibitor hesperadin (a generous gift of Tarun Kapoor, The Rockefeller University, NY) was supplied to some cultures for the last hour of MG132 treatment. Medium was then exchanged to remove hesperadin and MG132, and cells were fixed after 45 or 60 minutes.
PtK1 chromosome preparation and flow sorting were performed as previously described (Rens et al., 1999). Flow-sorted chromosomes were used as templates for DNA amplification by DOP-PCR (Telenius et al., 1992) and primary DOP-PCR products were used to incorporate biotin-16–dUTP (Roche), FITC-dUTP, Cy5-dUTP, Cy3-dUTP (Amersham), Texas-Red–dUTP (Molecular Probes) and DEAC-dUTP (NEN Life Science Products) by PCR. FISH was performed as previously described (Rens et al., 2003b), except that chromosome paints were denatured at 70°C for 10 minutes and slides were denatured at 55°C for 10 seconds. The biotin-labelled probe was visualized by Cy5.5-avidin (Amersham). After detection, slides were counterstained in 0.08 μg/ml DAPI in 2×SSC for 2 minutes and mounted in ProLong Antifade (Molecular Probes).
For the analysis of Ca2+-resistant kinetochore-microtubule interactions, slides were fixed as described (Kapoor et al., 2000) except that they were post-fixed in methanol. Primary antibodies were: CREST serum (Antibodies Incorporated, 1:50), anti-α-tubulin FITC-conjugate (Sigma-Aldrich, 1:100) and anti-Hec1 (Abcam, 1:500). For phospho-H3 staining, cells were fixed in ice-cold methanol for 5 minutes, permeabilized 5 minutes in PBS/0.5% Triton-X and then incubated with anti-Ser10-phospho-H3 antibody (Upstate Biotechnology, 1:100). DNA was counterstained in 0.1 μg/ml DAPI.
Fluorescence microscopy and image acquisition
M-FISH images were captured using Leica QFISH software (Leica Microsystems) and a cooled Sensys CCD camera (Photometrics) mounted on a Leica DMRXA microscope equipped with six specific filter sets and a 63× (1.3 NA) objective. Lagging chromosomes and ND were scored on anaphases with fully migrated chromosomes (mid-anaphase), late anaphases and telophases. Chromosomes were considered lagging if they were located inside the central 25% of the pole distance after all chromosomes had migrated. Between 133-413 ana-telophases were scored for each experimental point and data were analyzed using the χ2 test. Immunofluorescence analysis of kinetochore orientation was carried out under an Olympus Vanox microscope equipped with a 100× (1.35 NA) objective and a SPOT CCD camera (Diagnostic Instruments). Colour-encoded images were acquired using ISO 2000 software (Deltasistemi). Three dimensional deconvolution and reconstruction was performed on 0.3-μm-interval z-stacks of optical sections using AutoDeblur 9.3 (AutoQuant Imaging) software.
We thank Tarun Kapoor for generously providing hesperadin, and Daniela Cimini for many valuable suggestions and comments on this work. This work was partially supported by grants from CNR (RTL 2007) and from AIRC.