Generation of trisomies in cancer cells by multipolar mitosis and incomplete cytokinesis

Abstract

One extra chromosome copy (i.e., trisomy) is the most common type of chromosome aberration in cancer cells. The mechanisms behind the generation of trisomies in tumor cells are largely unknown, although it has been suggested that dysfunction of the spindle assembly checkpoint (SAC) leads to an accumulation of trisomies through failure to correctly segregate sister chromatids in successive cell divisions. By using Wilms tumor as a model for cancers with trisomies, we now show that trisomic cells can form even in the presence of a functional SAC through tripolar cell divisions in which sister chromatid separation proceeds in a regular fashion, but cytokinesis failure nevertheless leads to an asymmetrical segregation of chromosomes into two daughter cells. A model for the generation of trisomies by such asymmetrical cell division accurately predicted several features of clones having extra chromosomes in vivo, including the ratio between trisomies and tetrasomies and the observation that different trisomies found in the same tumor occupy identical proportions of cells and colocalize in tumor tissue. Our findings provide an experimentally validated model explaining how multiple trisomies can occur in tumor cells that still maintain accurate sister chromatid separation at metaphase–anaphase transition and thereby physiologically satisfy the SAC.

Whole chromosome gains (typically trisomies and tetrasomies) are the most common type of chromosome aberration in cancer cells (Mitelman Database of Chromosome Aberrations in Cancer, 2010; http://cgap.nci.nih.gov/Chromosomes/Mitelman). It is well established that chromosomal alterations in cancer can arise as a consequence of abnormal segregation of chromosomes at mitosis, but it remains to be shown precisely how extra copies of whole chromosomes are gained. It has been suggested that deficiency of the spindle assembly checkpoint (SAC) or other key mechanisms controlling sister chromatid separation could promote the generation of trisomies in cancer cells through a continuously elevated rate of concurrent chromosome gain and loss (i.e., nondisjunction) at metaphase–anaphase transition (1–4). The SAC deficiency model has been challenged by the fact that mutations in mitotic checkpoint genes have been found only in a minority of human cancers (3–7), but the absence of such mutations could still be explained by epigenetic modifications of mitotic control genes or by mutations in SAC genes that are not yet characterized. Therefore, it has remained difficult to experimentally validate the association between SAC deficiency and trisomies. This problem could be circumvented by estimating directly the rate of sister chromatid separation failure at mitosis. To do this, we used FISH to monitor the segregation of individual chromosomes in ana-telophase cells. This method was then applied to Wilms tumor (WT), which is a prototypical model for cancers with whole chromosome gains, showing polysomies in the majority of cases with abnormal karyotypes, of which 62% have two or more coexisting trisomies and 16% have tetrasomies (Mitelman Database of Chromosome Aberrations in Cancer, 2010). In contrast to previous assumptions, we find that continuous generation of trisomies through SAC deficiency is unlikely to explain the generation of multiple whole chromosomes gains in these tumors. Instead, our data indicate that a previously uncharacterized mechanism consisting of combined spindle multipolarity and cytokinesis failure could explain trisomy generation in WT.

Results

First we determined the baseline rate of chromosome missegregation in short-term cultures from normal dermal fibroblast samples (Table S1) and found that the median rate was 4.0 × 10⁻⁴ (range, 3.3–4.1 × 10⁻⁴) per chromosome per mitosis, equivalent to one missegregation in approximately 50 cell divisions (Fig. 1 A and B and Fig. S1A). As a positive control for SAC deficiency, we then used cells from patients with mosaic variegated aneuploidy (MVA) syndrome, a rare autosomal-recessive condition associated with a high rate of constitutional mosaic aneuploidies, predominantly trisomies and monosomies. A subgroup of patients with MVA exhibit biallelic mutations of the SAC key gene BUB1B (8). We analyzed fibroblasts from three MVA cases, all of which showed SAC deficiency by failure to arrest normally at metaphase during nocodazole exposure; two with and one without biallelic BUB1B mutations (8). All three MVA cases exhibited rates of chromosome missegregation that were more than 10-fold higher compared with normal fibroblasts (Table S1 and Fig. 1C). Ana-telophase cells in which missegregation was detected showed a bipolar orientation and had only two centrosomes, as shown by combined FISH and immunofluorescence in one of the MVA cases (Fig. S1B). Elevated rates of missegregation in bipolar mitoses were also found in the colorectal carcinoma cell lines SW480 and LoVo, known to exhibit SAC deficiency (1). In contrast, the rate of missegregation in bipolar mitoses was similar to fibroblasts in the SAC-competent colorectal cancer cell line DLD1. Notably, all three colorectal cancer cell lines also showed multipolar ana-telophase cells coordinated by multiple centrosomes. None of these cell divisions produced daughter cells with the same chromosome copy number as that of the mother cell, resulting in a high missegregation frequency. Thus, SW480 and LoVo exhibited at least two types of chromosomal instability, one caused by SAC dysfunction and one by centrosomal disturbances, consistent with previous studies (1, 9).

Fig. 1.

Chromosome missegregation in bipolar and multipolar mitoses. (A) FISH with centromeric probes for chromosomes 7 (red), 12 (green), and 18 (violet) shows amphitelic chromosome segregations at anaphase in an F-N fibroblast. Homologous chromosomes equidistant from the cellular equator have been classified in probable sister chromosome pairs (broken white lines). (B and C) Probes for chromosomes 13 (green), 18 (violet), and 21 (red) shows 3–1 missegregation at telophase of chromosome 21 (arrows) in an F-N fibroblast (B) and an MVA28 fibroblast (C). (D) Time-lapse series (t is time in minutes) showing a tripolar metaphase (green broken lines; t = 0) followed by tripolar ana-telophase (t = 14–36 min). Cytokinesis is initiated along two cleavage furrows in this cell division (t = 70 min), but one of the cleavage furrows regresses and only two daughter cells are formed (t = 306 min), of which the larger is binucleate as evidenced by two clusters of nucleoli. The larger cell (t = 3,192 min) enters mitosis (red broken lines), forming a single mitotic plate and divides into two daughters (t = 3,586 min). Daughter cells from both cell divisions remained without evidence of cell death or degeneration throughout the observation time (139-h total time lapse; Movies S1 and S2). (E and F) Immunofluorescence staining for β-tubulin (green) and MAD2L1 (red) in WiT49 cells shows retention of MAD2L1 foci in a complex tetrapolar anaphase cell with lagging chromosomes (F, arrows), whereas no MAD2L1 foci are present in a tripolar anaphase cell (E); note the absence of a β-tubulin-positive midbody between the two upper poles in E. (G) Immunofluorescence stain655555ing for β-tubulin (red) combined with FISH for the centromeres of chromosomes 4 (green), 7 (violet), and 9 (yellow) in a postmitotic HEK293D cell shows 3–1 segregation of chromosome 7, resulting in trisomy 7 in the binucleated daughter cell (nuclei a and b) and monosomy in the mononucleated daughter cell (c). (H) β-Tubulin staining combined with FISH for the centromeres of chromosomes 3 (yellow), 11 (violet), and 15 (green) in a telophase HEK293D cell shows concurrent formation of trisomies for chromosomes 3 and 11 in the binucleated daughter cell (nuclei a and b) and monosomies for these chromosomes in the mononucleated daughter cell (c); note the absence of midbody between a and b. (I) Time-lapse fluorescence/phase contrast microscopy in HEK293D H2B-GFP cells shows a tripolar anaphase (t = 0) resulting in one binucleated (a, b) and one mononucleated (c) daughter cell (t = 1,555 min; Movie S4).

We then screened cells from five primary WT and the WiT49 WT cell line, all of which had hyperdiploid-triploid karyotypes with whole chromosome gains, typically trisomies. Even though aneuploid cells were specifically selected for analysis, to avoid scoring contaminating nonneoplastic cells in primary tumors, a significantly elevated rate of missegregation compared with normal fibroblasts was not observed in bipolar ana-telophase configurations in WT cells (P = 0.39; Mann-Whitney U test). WTs exhibited rates of missegregation at bipolar mitosis that were, on average, 11 fold lower than those of SAC-deficient MVA fibroblasts and colorectal cancer cell lines (P = 0.0078). However, all six WTs showed ana-telophase cells in multipolar configurations (0.8–4.2% of anaphase cells, compared with none in >1,500 fibroblasts scored; Fig. S1C). These cell divisions resulted in unequal copy numbers in sister nuclei at rates that were 16 to 128 times higher than those resulting from missegregation at bipolar mitosis.

Multipolar cell divisions have been observed in many human tumor types but their role in tumorigenesis has remained disputed, primarily because clonogenic survival of daughter cells from such mitoses appears to be significantly reduced (10, 11). To assess whether the observed multipolar cell divisions might still contribute to clonal evolution in WT, we performed holographic time-lapse imaging of WiT49 cells. By using a low-intensity laser to provide enhanced contrast imaging, this method allows periods of continuous observations of growing cells for more than 1 wk without the need of transfection with fluorescent markers, which might otherwise induce alterations in cellular phenotype. Because more than 80% of the multipolar anaphase cells observed in the primary tumors were tripolar, we analyzed only multipolar mitoses that divided toward three anaphase poles. Surprisingly, only a minority (two of 18) of these divisions resulted in three daughter cells (Fig. 1D and Fig. S2 A and B). The majority resulted in multinucleate single daughter cells (seven of 18) or, more commonly (nine of 18), underwent cytokinesis with complete ingression of the cleavage furrow along one plane only, with another furrow that was typically initiated but failed to show complete ingression (Fig. S2C). Because chromosomes nevertheless segregated toward three poles, the latter mitoses resulted in the formation of two daughter cells, one binucleated and one mononucleated. Of the 18 daughter cells resulting from cytokinesis along only one cleavage plane, three of the binucleated cells during the period of observation again underwent mitosis, in which they showed an intermingling of prometaphase chromosomes, giving rise to a single mitotic plate before cell division (Movies S1 and S2), demonstrating that cells having undergone this type of cell division may proliferate further. Immunofluorescence on fixed WiT49 cells corroborated that approximately 50% of the tripolar telophase configurations exhibited cleavage along one plane only, as evidenced by a single midbody by beta tubulin staining (Fig. 1E). Furthermore, 66% of cell divisions of this type showed a complete absence of kinetochore MAD2L1 staining, indicating that metaphase–anaphase transition had occurred through satisfaction of the SAC (Fig. 1 E and F and Fig. S1F). Accordingly, immunofluorescence combined with FISH showed a segregation pattern consistent with amphitelic sister chromatid separation in greater than 80% of such ana-telophase configurations in WiT49, including the formation of trisomies in the binucleated daughter cells (Fig. S1 D and E).

WTs originate from embryonic renal progenitor cells that have undergone maturation arrest (12). To validate our findings in an independent system, we therefore turned to a human embryonic kidney cell line (HEK293D) that was obtained by transforming primary human embryonic kidney cells with sheared human adenovirus type 5 (13), leading to deregulation of the centrosome cycle and multipolar mitoses (14). These cells had acquired copy number alterations for some chromosomes, but many were still retained as disomies in the stem line (15). To evaluate further whether trisomies could be acquired also in these cells through tripolar mitosis and incomplete cytokinesis, we performed transfection with an H2B-GFP construct, allowing concurrent phase contrast and fluorescent imaging in which chromatin and cytokinesis could be studied in real time. Of bipolar HEK293D cell divisions, the vast majority (95%) underwent complete cytokinesis, resulting in two daughter cells (Figs. S3A and S4A and Movie S3). In contrast, the majority (80%) of tripolar anaphase configurations showed failure of cytokinesis, typically resulting in one binucleated and one mononucleated daughter cell (Figs. S3B and S4B and Movies S4 and S5); only a minority (20%) underwent complete cytokinesis (Fig. S3C and Movie S6). Combined β-tubulin immunofluorescence and FISH for centromeres of chromosomes 3, 4, 7, 9, 11, and 15 (disomic in the HEK293D H2B-GFP stem line) was then performed on telo-interphase cells with single cleavage planes and a 2:1 nuclear distribution. Of these, the majority exhibited unbalanced segregation between the two daughter cells, typically with a 3:1 chromosome distribution, resulting in a trisomy in the binucleated daughter cell (Fig. 1 G–I and Fig. S4 C and D). Thus, trisomic cells formed through tripolar division coupled to incomplete cytokinesis also in the HEK293D model system, corroborating the data from WiT49.

A diploid cell undergoing tripolar nuclear division with amphitelic sister chromatid separation followed by cytokinesis along only one furrow will result in one hyperdiploid and one hypodiploid daughter cell, with a difference in chromosome number determined by the number of chromosomes present on the metaphase axis along which the cleavage furrow fails to undergo complete ingression (Fig. 2A). Presence of several chromosomes on this axis will result in the simultaneous generation of multiple trisomies and/or tetrasomies in the hyperdiploid daughter cell. This type of cell division is therefore an attractive model for the generation of several whole chromosome gains through a single event. However, the frequency of missegregation in bipolar cell divisions was greater than zero in WT cells (Table S1), and the possibility that multiple trisomies and/or tetrasomies were generated by sequential gain of chromosomes (Fig. 2B) could therefore not be completely excluded. To test which of the two alternatives better predicted the pattern of whole chromosome gains in published karyotypes from primary WTs, we analyzed the frequency of tetrasomies in 152 cases with two or more whole chromosome gains (Mitelman Database of Chromosome Aberrations in Cancer, 2010). The sequential model predicts the highest probability for the acquisition of tetrasomies, as chromosomes already in a trisomic state would have a higher probability than disomic chromosomes for being involved in any subsequent missegregation (Fig. 2C). The tetrasomy frequencies found in the 152 WTs were clearly distinct from those expected from the sequential generation model (Fig. 2B), whereas they closely mirrored the ratios predicted by the tripolar mitosis-incomplete cytokinesis model (Fig. 2A).

Fig. 2.

Models for the generation of trisomies and tetrasomies. (A) A tripolar nuclear division with amphitelic sister chromatid separation and segregation, followed by incomplete cytokinesis, will generate tetrasomies in one daughter cell (blue membrane, Right) for chromosomes (blue) of which both homologues are located on the metaphase axis (blue line, Left), along which the cleavage furrow fails to ingress completely (red arrow, Right), whereas trisomies will be generated in the same daughter cell for homologues (red) located on this axis and on either of the other axes (green lines, Left); disomies will be retained when both homologues (green) are present on the axes (green) along which cytokinesis is complete. (B) A bipolar mitosis with missegregation of one chromosome (red) will generate one trisomic and one monosomic daughter cell. Another missegregation event in the trisomic cell population involving the same (red) chromosome will result in tetrasomic and disomic daughter cells (Lower Left), whereas missegregation involving another chromosome will result in two trisomies (Lower Right). (C) The frequency of tetrasomic tumors (blue plot) in 152 WTs with at least two whole chromosome gains is well in accordance with the model in A (green plot) but differs significantly (χ² test) from the distribution predicted from the model in B (red plot). The proportion of cells carrying specific chromosomal imbalances in primary WT biopsies WT-F (D), WT-G (E), and WT-H (F), estimated from B-allele frequencies at SNP-array analysis. Trisomies (+) are present in an equal proportion of cells (D and E; red demarcations), whereas segmental imbalances are typically present in clones of different sizes (E and F). Abnormalities present at similar proportions are signified by identical colors (red, green, or blue bars), whereas abnormalities present in significantly different proportions of cells are signified by different colors. Gray bars indicate populations for which the proportion confidence intervals overlapped with that of at least one other population. del, hemizygous deletion; dup, duplication; trp, triplication; upd, uniparental disomy.

To compare the two models further, we performed SNP-based array comparative genomic hybridization analysis of 15 primary WTs, all of which had multiple trisomies and/or multiple segmental imbalances. By calculations based on allele frequencies obtained by the BAF segmentation algorithm (16), we assessed the proportion of cells containing each specific genomic alteration found in every tumor biopsy specimen (Fig. 2 D–F and Fig. S5). Sequential acquisition of chromosome alterations would most likely produce genetically distinct populations during tumor development, with the largest clone containing the abnormality acquired first, the second largest the abnormality acquired next, and so on. Such a clonal hierarchy was indeed observed for segmental/structural aberrations in the majority of cases (11 of 13) with multiple structural imbalances, well in accordance with a previous study showing a sequential acquisition of structural changes in WTs through chromosomal breakage-fusion-bridge cycles (17). In contrast, of the 24 trisomies detected, all but one (+13 in WT-B) were present at frequencies that were identical to the other trisomies present in the same case (P < 0.001; trisomies compared with structural changes by Fisher exact test), indicating that the trisomies in each case were acquired at the same time point in tumor development. To corroborate this pattern with an independent method, we then analyzed chromosome copy numbers by interphase FISH in foci of 30 to 100 cells in sections from two primary WTs, one of which exhibited trisomy 7 and 12, and the other trisomy 8 and 12. By comparing the number of centromeric FISH signals for these chromosomes to a known disomic reference chromosome, the spatial distribution of trisomic clones could be traced in each tumor (Fig. 3 and Fig. S6). In neither case could any focus exhibiting only one of the trisomies be found, as would be expected from the sequential model. Instead, well delimited tumor regions that were trisomic for both the assayed chromosomes were observed to border directly on areas of disomic tumor cells, again consistent with a concurrent generation of trisomies through as single abnormal mitotic event, as predicted by the tripolar mitosis-incomplete cytokinesis model.

Fig. 3.

Spatial distribution of trisomic cells in tumor tissue. (A) Tissue section from WT-G with trisomies 7 and 12 in a subpopulation of tumor cells, previously detected by SNP-based array comparative genomic hybridization. Most nuclei (stained by diaminophenylindol; blue) are sectioned, leading to a reduced number of probe signals. Therefore, the distribution of trisomic cells was mapped by calculating copy number ratios in foci of 30 to 100 cells by FISH with centromeric probes for chromosomes 7 (A, II) and 12 (A, III), using centromere 16 as a reference for disomy (Fig. S6E shows signal number ratios). This allowed demarcation of areas containing cells with trisomies (red borders in A, I) and disomies (blue borders) in the corresponding H&E section. Blue and red filled circles in A, II, and A, III, indicate cell populations classified as disomic and trisomic, respectively. The adjacent normal kidney (green borders) contains only disomic areas. The asterisk represents disomic stromal tissue surrounding a blood vessel. (B) Representative FISH image of normal kidney tubules (area B in A, I; rectangular area in B, I, is shown at higher magnification in B, II) with disomy for chromosomes 7 (red), 12 (green), and 16 (violet; arrow in B, II). (C and D) Trisomic cells were detected in epithelial (C in A, I) and stromal (D in A, I) tumor elements, as exemplified by cells showing three signals for each of chromosomes 7 and 12 (arrows in high-power images C, II, and D, II).

Discussion

Taken together, our data show that an elevated frequency of nondisjunction at bipolar mitosis mediated by a defective SAC is unlikely to explain the occurrence of multiple whole chromosome gains in WT. Instead, a combination of spindle multipolarity and failed cytokinesis appeared to be a strong candidate mechanism. Daughter cells from such asymmetrical cell divisions were capable of again entering mitosis and forming novel clones. Our data did not provide information regarding the long-term clonogenic in vitro survival of cells that had acquired trisomies through this mechanism, but the finding that clones harboring double trisomies could be mapped to confined topographical regions in tumor tissue indicated that at least some daughter cells that had acquired concurrent trisomies in WT can expand clonally and contribute to tumor development. Considering the relatively high frequency of baseline chromosome missegregation found in bipolar mitoses (1:50 cell divisions), our model does not exclude that additional whole chromosome gains and losses occur during the expansion of clones with multiple trisomies, leading to an even more complex panorama of chromosome aberrations. Neither does it contradict previous studies showing that polysomic chromosomes are prone to undergo structural rearrangements (18). However, our findings that the multipolar anaphase cells giving rise to trisomies occurred at a very low frequency in each of the studied tumors, and that their daughters did not always survive to proliferate further, indicate that the acquisition of trisomies through such mitoses is probably a rare, possibly once-only, phenomenon during the development of a tumor. This could explain why tumor cells only exhibiting trisomies are typically quite stable cytogenetically, showing subclones in only a minority of cases. Furthermore, our data indicate that clonal evolution of polysomies occurs through discrete steps in which multiple trisomies may arise, rather than as a continuous process. This is in stark contrast to previously suggested models based on defective control of sister-chromatid separation (1–5), in which gains and losses of chromosomes are more likely to be acquired continuously during tumor cell proliferation. Our suggested model may be of importance for the generation of trisomies in other tumors that frequently show multiple trisomies and is consistent with studies of allele dosages in pediatric high hyperdiploid acute lymphoblastic leukemias, showing that hyperdiploidy most probably originates in a single aberrant mitosis (19). The present study adds to several other arguing for the importance of centrosome dysregulation for the generation of aneuploidy (9, 10, 20, 21), but it is unique in that it suggests an empirically based mechanism directly linking supernumerary centrosomes and spindle multipolarity to trisomy formation. Furthermore, it shows that multiple chromosome copy number alterations can occur through mitoses that physiologically satisfy the SAC.

Materials and Methods

Short-term cultures established from fibroblasts and primary tumors were subcultured no more than five times before analysis. MVA12 exhibited the biallelic BUB1B mutations 2211–2212insGTTA _S738fsX753 and c.2441G > A_p.R814H (8), whereas MVA41C contained mutation c.2144–2A > G and c.464A > G_ p.Y155C. Cell culture, fixation, FISH, and immunofluorescence were performed as described previously (14). Chromosome segregation was scored by tricolor FISH in ana-telophase cells using centromeric and single-copy probes (Abbott). For holographic time lapse microscopy, the HoloMonitor M2 (Phase Holographic Imaging) was used to digitally capture holograms every 2 to 5 min as described by Mölder et al. (22). For time-lapse phase-contrast/fluorescence microscopy, NIS Elements Br software (Nikon Instruments) was used to acquire images at 5-min intervals. Detection of genomic imbalances by SNP array was performed by HumanCNV370-Duo/Quad Genotyping BeadChips (Illumina) according to the manufacturer's specifications. A detailed description of the experimental methods can be found in SI Materials and Methods and Dataset S1.