Truncating APC mutations have dominant effects on proliferation, spindle checkpoint control, survival and chromosome stability

Germline mutations in the adenomatous polyposis coli (APC) gene are responsible for familial adenomatous polyposis (FAP). In addition, approximately 80% of all sporadic colorectal tumours have mutations in APC (Polakis, 1997). Consistent with Knudson's `two-hit' hypothesis, tumour cells from FAP patients and sporadic cases harbour mutations in both APC alleles, indicating that APC is a classic tumour suppressor. The tumour suppressor function of APC is now well established (Bienz and Clevers, 2000; van Es et al., 2001). In the absence of Wnt signals, APC facilitates phosphorylation of β-catenin by GSK3β, thus targeting β-catenin for ubiquitin-mediated degradation. Following activation of the Wnt pathway, APC/GSK3β-mediated phosphorylation of β-catenin is down regulated allowing β-catenin to accumulate, which in turn results in transcriptional activation of a number of proliferative genes (Fearnhead et al., 2001; van Es et al., 2001). Upon loss of APC function, β-catenin accumulates in the absence of Wnt signals thereby promoting deregulated proliferation (Sieber et al., 2000).

Although both alleles are mutated in APC-defective colorectal cancer cells, these cells do not completely lack APC protein. Rather, while one allele acquires a truncating mutation, the second allele undergoes either loss of heterozygosity (LOH) or a second truncating mutation (Kinzler and Vogelstein, 1996). Thus, APC mutations typically result in the expression of N-terminal fragments. Analysis of tumours from FAP patients has revealed that there appears to be an interdependence of the `two hits' on APC in that there is a correlation between the initial germline APC mutation and the nature of the second somatic mutation (Lamlum et al., 1999). Specifically, germline mutations around codon 1300 are associated with allelic loss of the remaining wild-type APC allele while mutations 3′ or 5′ of this region predominantly show a second truncating mutation. A similar relationship between APC mutations in sporadic colorectal cancer has also been observed (Rowan et al., 2000). Thus, it has been argued that different N-terminal APC fragments may have different dominant negative or gain-of-function properties (Lamlum et al., 1999; Rowan et al., 2000). As such, APC mutation may also have oncogenic properties.

While it is not yet clear how APC mutation might exert oncogenic properties, it is becoming increasingly apparent that APC is a multifunctional protein (van Es et al., 2001). In particular, APC plays several roles in regulating the cytoskeleton. By stimulating Asef, a Rac-specific guanine nucleotide exchange factor, APC regulates the actin cytoskeletal network and cell morphology (Kawasaki et al., 2000). APC also binds and stabilises microtubules (Zumbrunn et al., 2001). By localising to the plus ends of microtubules, APC clusters at membrane protrusions in areas of the cell that are actively migrating (Mimori-Kiyosue et al., 2000a; Nathke et al., 1996). The C terminus of APC binds EB1 (Su et al., 1995), a protein that localises to the distal tips of microtubules during both interphase and mitosis (Juwana et al., 1999; Mimori-Kiyosue et al., 2000b). EB1 localises to the centrosome (Louie et al., 2004), and in several systems EB1-related proteins have been shown to be required for assembly and positioning of the mitotic spindle, indicating a role for EB1 in accurate chromosome segregation (Beinhauer et al., 1997; Muhua et al., 1998; Rogers et al., 2002). Recent evidence suggests that APC may also play a role in chromosome segregation. Drosophila APC homologues have been implicated in the orientation and positioning of the mitotic spindle in germline stem cells (Yamashita et al., 2003). In mammalian cells, APC also localises to centrosomes during interphase (Tighe, 2001; Louie et al., 2004). In addition, APC localises to kinetochores during mitosis in a microtubule-dependent manner (Fodde et al., 2001; Kaplan et al., 2001) and has been shown to interact with the spindle checkpoint proteins Bub1 and BubR1 (Kaplan et al., 2001). Bub1 and BubR1 can also phosphorylate APC in vitro, a reaction that is enhanced following phosphorylation by GSK3β (Kaplan et al., 2001). Significantly, mouse embryonic stem (ES) cells harbouring homozygous truncating APC mutations exhibit near-tetraploid karyotypes (Fodde et al., 2001; Kaplan et al., 2001). More recently, it has been shown that spindles formed in mitotic Xenopus egg extracts depleted of APC protein exhibited changes in microtubule density, resulting in the formation of weaker spindles (Dikovskaya et al., 2004). Aberrant spindle structures and weakened kinetochore-microtubule attachments have also been reported in human cells harbouring APC mutations (Fodde et al., 2001; Green and Kaplan, 2003).

These latter observations are of particular interest as the majority of human colorectal tumours (85%) are highly aneuploid. Although it has been appreciated for many years that human tumours are frequently aneuploid, it was only recently shown that this is due to an underlying chromosome instability (CIN) phenotype (Lengauer et al., 1997). Despite recent advances in our understanding of the mechanisms responsible for ensuring accurate chromosome segregation in human cells (Jallepalli and Lengauer, 2001; Musacchio and Hardwick, 2002; Taylor et al., 2004), the mutations that give rise to CIN remain obscure. The observations that mouse ES cells with homozygous APC mutations are aneuploid, that kinetochore-microtubule interactions appear weak in CIN cells, and that spindles are aberrant in APC-depleted extracts raises the possibility that APC mutation may cause CIN. Indeed, both APC mutation and the acquisition of CIN occur very early during tumourigenesis (Powell et al., 1992; Shih et al., 2001). Furthermore, there is a correlation, albeit an imperfect one, between APC status and CIN: colon cancer cells with APC mutations usually exhibit CIN, whereas near-diploid colon cancer cells frequently have wild-type APC (Cahill et al., 1999; Fodde et al., 2001; Green and Kaplan, 2003; Rowan et al., 2000).

These observations provide interesting but circumstantial evidence to suggest that the APC mutation may cause CIN in human colon cancer (Fodde et al., 2001; Green and Kaplan, 2003; Kaplan et al., 2001). Consistent with this notion, ectopic expression of the N-terminal 750 amino acids of APC in a non-CIN colon cancer line with wild-type APC reduced the accumulation of cells in mitosis following spindle damage (Tighe et al., 2001). This phenotype is typical of cells with a compromised spindle checkpoint (Taylor and McKeon, 1997) raising the possibility that APC mutations may induce CIN in a dominant manner. In this study, we set out to directly test this possibility. We show that truncating APC mutations do have dominant effects, which can result in chromosome instability. Thus, these observations raise the possibility that the oncogenic properties of APC mutation may be due to the initiation of CIN, which drives the accumulation of other mutations, including the loss of the second APC allele.

All cell lines were cultured as described previously (Tighe et al., 2001). Nocodazole (Sigma; 5 mg/ml in DMSO) was used at a final concentration of 0.2 μg/ml unless otherwise stated. Stable cell lines were created using the Flp-In™ System (Invitrogen) according to the manufacturers instructions. Briefly, a parental host cell line, HCT-116 LacZeo, was created by transfecting pFRT/lacZeo into HCT-116 cells by electroporation. Transfected cells were selected in 10 μg/ml Zeocin™ (Invitrogen), and single colonies were picked and expanded. β-Galactosidase expression was analysed using the β-Gal Assay Kit (Invitrogen) and single copy integrants identified by Southern blot analysis. pcDNA5/FRT (Invitrogen) was modified to include the 5′ untranslated sequence from human lamin A cDNA and an N-terminal Myc epitope tag (Taylor and McKeon, 1997). N-terminal fragments of APC (N-APC) encoding amino acids 2-750, 2-1309 and 2-1807 were amplified with Pfu polymerase (Stratagene), cloned into pcDNA5/FRT/Myc then sequenced. Plasmids were purified using ion exchange chromatography (Qiagen) then co-transfected with pOG44, a plasmid expressing the Flp recombinase (Invitrogen), into HCT-116 LacZeo cells using Lipofectamine Plus™ (Invitrogen) according to the manufacturer's instructions. Following selection in 200 μg/ml hygromycin B (Roche), colonies were pooled and expanded. The clonal line pLP-N750 was created by cloning a cDNA fragment encoding amino acids 2-750 of APC into pLPCX Myc (Hussein and Taylor, 2002). This was then electroporated into HCT-116 cells and transfectants selected in 0.5 μg/ml puromycin (Clontech). Single colonies were picked and expanded. Tetracycline-inducible 293-based cell lines expressing the N-APC fragments were created in a similar manner. Flp-In™ T-REx™-293 cells (Invitrogen) were co-transfected with pcDNA5/FRT/TO/Myc vectors and pOG44, selected in 15 μg/ml blasticidin (Invitrogen) and 150 μg/ml hygromycin B (Roche). Protein expression was induced by the addition of 0.01 μg/ml tetracycline (Sigma) for 24 hours.

Cells grown on coverslips were fixed in ice-cold 100% methanol for 10 minutes, washed in PBS plus 0.1% Triton X-100 (PBST), blocked in 5% non-fat milk and then incubated at room temperature for 30 minutes with combinations of the following primary antibodies: 9E10 (mouse monoclonal anti-Myc antibody, 1:250); ACA (human anti-centromere, Antibodies Inc, 1:100); SB1.3 (sheep anti-Bub1, 1:1000); SBR1.1 (sheep anti-BubR1, 1:1000), TAT-1 (mouse anti-tubulin, 1:100); RAA.1 (rabbit anti-Aurora A, 1:10000), Ab-3 (mouse anti-APC, Oncogene, 1:20); rabbit anti-C-Nap1 (a gift from Andrew Fry, 1:100). To analyse localisation of checkpoint proteins and spindle morphology, cells were permeabilised for 90 seconds in microtubule stabilising buffer (100 mM Pipes pH 6.8, 1 mM MgCl, 0.1 mM CaCl₂, 0.1% Triton X-100), fixed for 10 minutes in 4% formaldehyde, then blocked and incubated with the appropriate primary antibodies as described above. Following washes with PBST, cells were stained for 30 minutes at room temperature with appropriate Cy2-, Cy3- or Cy5-conjugated secondary antibodies (Jackson Immuno-Research Laboratories), all diluted 1:500. Following washes, cells were stained with Hoechst 33358 at 1 μg/ml in PBST then mounted in 90% glycerol plus 20 mM Tris-HCl pH 8.0. Fluorescence microscopy was performed using a Zeiss Axiovert 200 equipped with epifluorescence and a Photometrics Cool Snap HQ CCD camera driven by Metamorph software (Universal Imaging). Deconvolution microscopy was performed using a wide field optical sectioning microscope (Deltavision, Applied Precision) as previously described (Taylor et al., 2001). Briefly, for each cell, a Z-series of images at 0.2 μm intervals was captured at each wavelength and then processed using constrained iterative deconvolution. Deconvolved image stacks were projected and fluorescence signal intensities quantified using SoftWoRx (Applied Precision). Acquired images were then imported into Photoshop (Adobe) for printing. To quantify the kinetochore bound protein, the average pixel intensities from at least 20 kinetochores from four or more cells was measured and background pixel intensities subtracted. Statistical analyses were performed using InStat® v3.0 (GraphPad Software Inc).

Soluble proteins were extracted in buffer (50 mM Tris pH7.5, 100 mM NaCl, 5 mM EDTA, 5 mM EGTA, 0.5% NP-40, 40 mM β-glycerol phosphate, 50 mM NaF, 5 mM sodium orthovanadate and 0.5 μg/ml each of leupeptin, pepstatin, chymostatin) then resolved on 3-8% Tris-acetate gradient gels (Invitrogen) according to the manufacturer's recommendations. Proteins were transferred onto nitrocellulose in 50 mM Tris, 380 mM glycine, 5% methanol and 0.02% SDS at 25 V for 20 hours, then blocked in TBST (50 mM Tris pH7.6, 150 mM NaCl, 0.1% Tween 20) plus 5% non fat dried milk for at least 1 hour at room temperature. Blots were then incubated overnight at 4°C with either 9E10 (mouse monoclonal anti-Myc antibody, 1:100) or Ali 12.28 [mouse monoclonal anti-APC antibody (Efstathiou et al., 1998), 1:1000]. After washing in TBST, bound primary antibodies were labelled with horseradish-peroxidase-conjugated goat anti-mouse antibodies (Zymed) diluted 1:2000 in 5% non fat dried milk/TBST. After washing in TBST, bound secondary antibodies were detected using the SuperSignal chemiluminescence system (Pierce) and imaged on Biomax MR film (Kodak).

Mitotic index measurements were determined as described previously (Tighe et al., 2001). Briefly, loosely attached and adherent cells were harvested then centrifuged onto microscope slides, fixed in 3% formaldehyde in PBS, stained with Hoechst and mounted as described above. The state of chromosome condensation was used to score the cells as either mitotic or interphase. To determine relative cell number we used crystal violet staining as a readout of cell mass. Briefly, 1×10⁴ cells were plated into 12-well dishes then, at the times indicated, the cells were fixed in 4% formaldehyde essentially as described previously (Hussein and Taylor, 2002). Fixed cells were stained with 0.1% crystal violet, the bound dye then extracted with 10% acetic acid and the optical density determined at 590 nm. To analyse cell survival after prolonged exposure to nocodazole, cells were plated in 6-well dishes, then incubated for 48 hours in the presence of 0, 0.02 or 0.2 μg/ml nocodazole. The nocodazole was washed away and fresh medium added. At the time points indicated, cells were fixed and stained with 4% formaldehyde plus 0.1% crystal violet, then processed as described above.

Cells were treated for up to 6 hours with 0.2 μg/ml nocodazole, then incubated in 0.45% hypotonic buffer (32 mM KCl, 16 mM Hepes, 0.5 mM EDTA pH 7.4) at 37°C for 20 minutes. They were then fixed in methanol:acetic acid (3:1) and stored at –20°C overnight (Taylor et al., 1996). Fixed cells were dropped onto acetic-acid-coated slides and air-dried. Chromosomes were stained with Hoechst, mounted and imaged as described above.

To determine whether truncating APC mutations have dominant effects on spindle checkpoint function and/or chromosome stability, we generated a panel of cell lines expressing either full length APC or three N-terminal fragments (Fig. 1A). These mutants, N750, N1309 and N1807 were all expressed as Myc-tagged fusion proteins and encode the first 750, 1309 and 1807 amino acids of APC respectively. To verify that these mutants expressed proteins of the expected size, stable 293-based cell lines expressing the Myc-tagged N-APC mutants under tetracycline control were generated. Immunoblotting with both anti-Myc antibodies and anti-APC antibodies (Fig. 1B) detected bands of the expected sizes, confirming the integrity of the expression cassettes.

A previous study using 293 cells showed that expression of an N-APC mutant interferes with microtubule plus-end attachments (Green and Kaplan, 2003). Because we wanted to determine whether N-APC mutants could induce CIN, we were unable to make use of the 293-based lines described above simply because 293 cells are already hypotriploid and have a severely compromised spindle checkpoint (not shown). Therefore, we chose HCT-116 as the parental line. These cells are derived from a human epithelial colon cancer (Brattain et al., 1981). However, they are diploid and chromosomally stable (Lengauer et al., 1997), they have a robust spindle checkpoint (Tighe et al., 2001) and they have two wild-type APC genes (Rowan et al., 2000). Therefore these cells represent a tractable and powerful model system to identify genetic lesions that may promote CIN during the evolution of human colon cancer (Cahill et al., 1998; Jallepalli et al., 2001; Michel et al., 2001).

To facilitate a direct comparison of the different APC mutants, we chose a transfection strategy that allows expression cassettes to be integrated into a pre-defined site in the genome (O'Gorman et al., 1991). This strategy eliminates differences in expression levels that may arise because of different sites of integration. Briefly, an FRT recombination site was integrated into the HCT-116 genome using a plasmid that encodes a Zeocin-lacZ fusion protein. Zeocin-resistant cell lines were first screened for β-galactosidase activity (Fig. 1C) then screened by Southern blotting to ensure the presence of only a single FRT site (not shown).

One resultant line, HCT-116 LacZeo, was then chosen for further manipulation. cDNA fragments encoding a green fluorescent protein (GFP) and the APC open reading frames (ORFs) were then transfected into HCT-116 LacZeo cells on a plasmid containing an FRT site. These vectors were cotransfected along with a plasmid encoding the FLP recombinase. As a negative control for the functional experiments, we also transfected a plasmid that only encoded the Myc tag. Hygromycin-resistant colonies were then pooled, expanded and analysed. Consistent with appropriate FLP-mediated recombination between the two FRT sites, the resulting hygromycin-resistant cells were largely Zeocin sensitive and showed a dramatic reduction in β-galactosidase expression (Fig. 1C). FACS analysis of pools of colonies following transfection of the GFP plasmid showed that at least 85% of the hygromycin-resistant cells expressed GFP (not shown). Furthermore, microscopy analysis of individual colonies showed that the vast majority of cells in each colony expressed GFP (not shown). Immunofluorescence analysis of the cells transfected with full length APC and the N-APC mutants showed Myc staining in the cytoplasm and nucleus (Fig. 1D). In contrast, cells transfected with the empty vector showed only background staining. Immunoblotting confirmed expression of the N750 mutant (Fig. 1E) and quantitative analysis indicated that this mutant was expressed at approximately 20-40% of the endogenous APC. Similar analysis indicated that the other N-APC mutants were expressed at no more than 10% of the endogenous protein (not shown).

Previously we showed that expression of the N-terminal 750 amino acids of APC reduced the accumulation of mitotic cells following exposure to nocodazole (Tighe et al., 2001), a spindle toxin that prevents microtubule polymerisation. To determine whether the lines generated here behaved in a similar manner, we treated the panel of cell lines with 0.2 μg/ml nocodazole for 18 hours then determined the mitotic index (Fig. 2A). Whereas the mitotic index of the populations expressing full length APC (66%) and the N1309 mutant (69%) did not differ significantly from the Myc control cells (75%) (P>0.05), the mitotic index of the N750 (54%) and N1807 (55%) mutants were significantly different (P<0.05).

To confirm this observation, we determined the mitotic index of the Myc, N750, N1309 and N1807 populations over a 48 hour time course in the presence of 0.2 μg/ml nocodazole (Fig. 2B). Consistent with previous observations (Tighe et al., 2001) the mitotic index of all cultures was initially below 5%, increased ∼10-fold during the first 12 hours and peaked at 18 hours. After that time, the mitotic index of each culture fell, consistent with the cells adapting and exiting mitosis (Taylor and McKeon, 1997), such that by 42 hours the mitotic index was less than 10%. The mitotic indexes of both the N750 and N1807 populations were lower than the control from 12 to 42 hours. In contrast, but consistent with the data in Fig. 2A, the N1309 line behaved more like the control cells. Thus, consistent with previous results (Tighe et al., 2001), expression of the N-terminal 750 and the N-terminal 1807 amino acids of APC can reduce the accumulation of mitotic cells following exposure to nocodazole.

The observation that expression of the N750 and N1807 mutants reduced the accumulation of mitotic cells in response to spindle damage is consistent with the notion that these cells have a compromised spindle checkpoint. However, a reduced mitotic index following spindle disruption could also be explained if these cells had a reduced population doubling time (Hussein and Taylor, 2002). Therefore, to rule out this possibility we determined the relative growth rates of the control and N-APC cultures. Briefly, equal numbers of cells were plated on day 0 and the relative cell number determined every day over a 7-day time course. To our surprise, the N750 cells grew markedly slower than the controls and the other N-APC lines (Fig. 3A). Specifically, whereas the control cells reached confluency by day 6, the N750 cells were less than 50% confluent.

Because expression of the N750 mutant reduces the population doubling time (Fig. 3A), the reduction in mitotic index as shown in Fig. 2 cannot simply be interpreted as the cells having a compromised spindle checkpoint. Therefore, we used a separate assay to determine whether expression of the N-APC mutants compromised the spindle checkpoint. Rather than monitoring the ability of cells to accumulate in mitosis in the presence of nocodazole, we measured the ability of cells to exit mitosis following release from a nocodazole block (Ditchfield et al., 2003). Briefly, the various cell lines were synchronised in mitosis by exposure to 0.2 μg/ml nocodazole for 13 hours. Mitotic cells were isolated by selective detachment then replated either in the continued presence of 0.2 μg/ml nocodazole or in drug-free medium. The mitotic index was then determined every hour (Fig. 4A). In the absence of nocodazole, the mitotic index of the control line fell to ∼50% within 4 hours, consistent with many of the cells exiting mitosis. Cells expressing either full length APC or the N1807 mutant behaved in a similar manner: after 4 hours, 44% and 40% of the cells remained in mitosis. In contrast, the mitotic index of the N750 and N1309 cultures fell faster than the controls: four hours after release, only 30% and 22%, respectively, remained in mitosis. In the continued presence of 0.2 μg/ml nocodazole, the mitotic index of the Myc control cells remained high (>90%), consistent with a sustained mitotic arrest. In addition, cells expressing full length APC, N1309, N1807 also remained arrested for the duration of the experiment. However, the mitotic index of the N750 culture fell to 72% despite the presence of 0.2 μg/ml nocodazole. Thus, while expression of the N1309 mutant can accelerate mitotic exit in the absence of nocodazole, N750 can accelerate mitotic exit in both the presence and absence of nocodazole. Although the N1807 mutant reduced the accumulation of mitotic cells when cultured in nocodazole (Fig. 2) it did not appear to accelerate mitotic exit following release from a nocodazole block (Fig. 4). While the reasons for this are not clear, the data in Figs 2 and 4 are consistent with the notion that the N750 and N1309 mutations can partially compromise the spindle checkpoint.

It has been shown previously that, following exposure to spindle toxins, cells with an compromised spindle checkpoint endoreduplicate (Michel et al., 2001; Taylor and McKeon, 1997) indicating that they by-pass the post-mitotic checkpoint which normally induces cell cycle arrest and/or apoptosis after an aberrant mitosis (Margolis et al., 2003). Therefore, to determine whether expression of N-APC mutants enhanced endoreduplication following prolonged mitotic arrest, the panel of cell lines was treated with 0.2 μg/ml nocodazole for up to 72 hours then analysed by flow cytometry to determine DNA content. We did not observe any increase in endoreduplication in the N-APC lines relative to the controls (not shown). Rather, the majority of cells appeared to undergo apoptosis. While this observation suggests that the spindle checkpoint and post-mitotic checkpoint are largely intact in the lines expressing the N-APC mutants, this assay is relatively insensitive in that a large number of cells would have to endoreduplicate in order to be detected by the FACS analysis. Therefore, we devised an alternative assay to determine whether expression of N-APC mutants might enhance survival of a small number of cells following prolonged mitotic arrest. Briefly, the control and N750 lines were treated with 0.2 μg/ml or 0.02 μg/ml nocodazole for 48 hours. The drug was then washed away and the cells replated on day 0. The cells were then analysed every day over a 7-day time course using phase-contrast microscopy and crystal violet staining (Fig. 5).

On day 0, the density of cells in the control cultures was higher than in the N750 culture, as judged from the phase contrast images (Fig. 5A) and crystal violet staining (Fig. 5B). However, analysis of the cells 7 days later revealed a more striking difference. The control culture contained more `floaters' and cellular debris relative to the N750 culture. In addition, the remaining adherent control cells appeared larger with many cytoplasmic vacuoles (Fig. 5A). In contrast the N750 cells appeared morphologically normal. The crystal violet staining showed that although there were significant amounts of cellular material attached to the dish in the control cultures, it was not organised into individual colonies and was concentrated in the centre of the dish (Fig. 5B). This suggests that in the control cultures there was little cellular proliferation following removal of nocodazole. Consistently, quantification of the bound crystal violet stain showed that there was very little increase in relative cell number during the 7-day time course (Fig. 5C). In contrast, in the N750 culture, large numbers of colonies were visible and spread evenly throughout the dish (Fig. 5B). Notably, in the culture initially treated with 0.02 μg/ml nocodazole these colonies appeared quite large, suggesting that they were proliferating. Indeed, quantification of the bound crystal violet stain showed that after day 4, the relative cell number in the N750 population increased several fold (Fig. 5C) consistent with initiating exponential growth. In contrast, the HCT-116 lines transfected with either full length or the other N-APC mutants behaved more like the control cells (not shown).

Like the parental HCT-116 cells, the Myc control and N750 lines are near diploid (not shown). To determine whether the cells that survived the 48 hour 0.02 μg/ml nocodazole exposure were near diploid, we examined metaphase spreads at various times during the 7-day time course. At day 0, the majority of cells in both the control and N750 populations were near diploid although both contained cells which were near tetraploid [22% of control and 16% of N750 cells contained >79 chromosomes (Fig. 6A)]. By day 6, both populations contained less tetraploid cells (11% in the control and 12% in N750) and the near diploid cells had a tighter distribution. Specifically, in the control cells, the proportion of cells containing 44-49 chromosomes increased from 31% at day 0 to 56% at day 6. Similarly, in the N750 culture, the proportion of cells containing 44-49 chromosomes increased from 26% at day 0 to 41% at day 6. However, while there appeared to be no major differences in ploidy between the control and N750 populations, it was striking that in the N750 culture we frequently saw chromosomes with two constrictions (Fig. 6B) suggesting that they were dicentric. Quantification revealed that such aberrant chromosomes were approximately 10 times more frequent in the N750 population compared to the controls (Fig. 6B).

Dicentric chromosomes can initiate breakage-bridge-fusion cycles leading to extensive genome rearrangements (Koshland et al., 1987; McClintock, 1940). Indeed, in the N750 line, we saw an increase in the number of aberrant anaphases, including bridges and lagging chromosomes (Fig. 6C). We were therefore intrigued to determine the longer term fate of the N750 survivors. The surviving cells in both the control and N750 cultures following the 48-hour exposure to 0.02 μg/ml nocodazole were cultured for a further 30 days. Note that although the number of survivors in the control culture was low (Fig. 5C), it was possible to isolate sufficient cells for further expansion, subculturing and analysis. Analysis of metaphase spreads revealed that the control culture was basically diploid with 96% of the cells containing 44-49 chromosomes (Fig. 6A). In stark contrast, the cells in the N750 culture displayed a wide range of karyotypes. Not only is the distribution bimodal, but the distribution around each peak is wide, with only 20% and 30% of the cells containing the near diploid (44-49 chromosomes) and near tetraploid chromosome number (>89% chromosomes), respectively (Fig. 6A,D). We did not observe chromosomes with two constrictions in the N750 line, raising the possibility that the dicentrics observed earlier had been resolved, possibly leading to the highly abnormal genome observed after 30 days (Fig. 6D).

All the observations described above were obtained using the lines created by pooling hygromycin-resistant colonies generated following transfection of the FRT based plasmid into the HCT-116 LacZeo line. To confirm the N750 observations, we generated a pair of clonal, HCT-116 based lines (see Materials and Methods) using a different vector, a different transfection system and a different selection regime. One line was created to express the Myc-tagged N750 mutant (pLP-N750), while the other expressed just the Myc tag (pLP-Myc). These lines were analysed using the same assays as described above. Consistently, the pLP-N750 line accumulated less mitotic cells than the control when cultured in the presence of 0.2 μg/ml nocodazole for 18 hours (Fig. 7A). Accordingly, over a 48-hour time course, the mitotic index of the pLP-N750 culture was markedly reduced compared to the control culture (Fig. 7B). In addition, cells expressing the N750 mutant exited mitosis faster than the controls following release from a nocodazole block (Fig. 7C). Furthermore, 7 days following a 48-hour exposure to 0.02 μg/ml, more colonies were apparent in the pLP-N750 culture than in the control (not shown). And finally, in contrast to the control cells, the survivors in the pLP-N750 culture contained rearranged chromosomes (Fig. 7D).

Thus, using two different vectors, two different transfection systems and different antibiotics to select stable integrants, our observations appear to be consistent. When compared to `empty vector' control cells, cells expressing the N-terminal 750 amino acids of APC showed a reduced accumulation in mitosis following exposure to nocodazole (Fig. 2), exhibited a proliferation defect (Fig. 3) and exited mitosis faster following release from a nocodazole block (Fig. 4). In addition, following a 48-hour exposure to 0.02 μg/ml nocodazole, cells expressing the N-terminal 750 amino acids of APC generated more survivors (Fig. 5) and accumulated more chromosomal aberrations (Fig. 6B) than the controls. Finally, following prolonged culture, the N750 survivors become highly aneuploid (Fig. 6D). Taken together, these observations are consistent with the notion that N-terminal APC mutants can have dominant effects on cellular proliferation, mitotic progression, cell survival and chromosome stability.

We have shown that expression of N750 reduces the accumulation of mitotic cells in the presence of nocodazole (Tighe et al., 2001) (Fig. 2). Several colon cancer cell lines, which display CIN, exhibit a similar phenotype (Tighe et al., 2001). We therefore asked whether the CIN cells, in comparison to non-CIN colon cancer cell lines, exhibited any of the other phenotypes induced by N750. To our surprise, the CIN lines all grew markedly slower than the non-CIN cells (Fig. 3B). In addition, three of the CIN lines, LoVo, SW480 and SW837, all exited mitosis rapidly in the absence of nocodazole such that less than 20% of the cells were in mitosis after 4 hours. Note, however, that in this assay the correlation is not perfect: >50% of the HT29 (CIN) cells remained in mitosis after 4 hours. Consistently, the HT29 line achieves the highest mitotic index following exposure to spindle toxins (Tighe et al., 2001). In addition, DLD-1 cells, which are classified as non-CIN, exited faster than both HCT-116 and RKO. This is interesting as DLD-1 cells have an APC mutation (Ilyas et al., 1997) and do have a tendency to become tetraploid (Fodde et al., 2001). Thus, although the correlation is not perfect, it is nevertheless intriguing that in three assays, expression of N750 in a non-CIN colon cancer line induces phenotypes that appear to be more typical of the CIN lines. Although we have not determined how the CIN lines behave in the survival assay, they are of course already highly aneuploid (Lengauer et al., 1997), as are the long term N750 survivors following the 48-hour nocodazole exposure (Fig. 6A,D and Fig. 7D).

To gain insight into how the N-APC mutants might exert dominant effects on spindle checkpoint function and chromosome stability, we set out to analyse the localisation of the N-APC mutants in the HCT-116-derived lines. It has been shown previously that both full length and N-APC mutants localise to centrosomes during interphase (Louie et al., 2004; Tighe et al., 2001; Yamashita et al., 2003). However, because of technical reasons we were unable to detect centrosome localisation of the exogenous Myc-tagged N-APC mutants. Therefore we analysed a panel of colon cancer cell lines harbouring endogenous N-APC mutants. The only line in which we could detect N-APC at the centrosome was DLD-1. Consistent with our previous observations (Tighe et al., 2001), the endogenous N-APC mutant in these cells localised to the centrosome in interphase (Fig. 8A). Interestingly, projections and 3D models of deconvolved image stacks showed that in interphase, the N-APC mutant localised asymmetrically, that is, to a domain between the two centrosomes (Fig. 8A). This is consistent with recent observations showing that full length endogenous APC localises close to the mother centriole (Louie et al., 2004). The localisation of full length APC or N-APC mutants during mitosis in human cells has not been reported. Significantly therefore, in DLD-1 cells, the endogenous N-APC mutant localises to centrosomes during the early stages of mitosis, i.e. prophase and prometaphase (Fig. 8B). However, it was virtually undetectable at centrosomes in metaphase and anaphase. Whether this reflects epitope masking or dissociation from the centrosome remains to be seen. Similarly, whether full length APC behaves this way during mitosis remains to be seen.

It has recently been shown that the spindle assembly is aberrant in Xenopus egg extracts depleted of APC and colorectal cell lines that harbour APC mutations (Dikovskaya et al., 2004; Green and Kaplan, 2003). Furthermore, expression of an N-APC mutant in 293 cells recapitulated this phenotype (Green and Kaplan, 2003). Therefore we analysed spindle morphology in the HCT-116-derived lines expressing N-APC mutants (Fig. 9A). In comparison to the controls, N-APC-expressing cells had significantly shorter spindles. As determined by the pole-pole distance measurements (Fig. 9B and Table 1), the Myc control cells had an overall spindle length of ∼8.6 μm, compared with ∼7.3 μm in the N750 cells. Values in the N1309 and N1807 cells were ∼7.0 μm and ∼7.1 μm, respectively. Although these spindles appeared to have decreased tubulin density at the spindle midzone (Fig. 9C), these differences were not significant.

As APC has been shown to bind and be phosphorylated by kinetochore-associated spindle checkpoint kinases (Kaplan et al., 2001), we carefully examined the localisation of Bub1 and BubR1 in the N-APC lines. Interestingly, in metaphase cells, the sister kinetochores of aligned chromosomes appeared closer together (Fig. 10A,B), indicating a reduction in tension across the centromere. Indeed, the average inter-kinetochore distance was significantly reduced from ∼1.4 μm in control cells to ∼1.0 μm in N750 cells (Fig. 10C and Table 1). Likewise, inter-kinetochore distance was significantly reduced in the N1309 and N1807 lines. In addition, there appeared to be more Bub1 and BubR1 bound to metaphase kinetochores in the N-APC cells (Fig. 10A,B and Table 1). Indeed, quantification of pixel intensity demonstrated that Bub1 levels were significantly increased from 1.96×10³ in control cells to 3.08×10³ in N750 cells. Values in the N1309 and N1807 cells were 4.56×10³ and 4.15×10³, respectively (Fig. 10D). Similarly, kinetochore-bound BubR1 was also significantly increased in the N-APC-expressing cell lines (Fig. 10B,E and Table 1). Because dissociation of Bub1 and BubR1 from kinetochores of aligned chromosomes appears to be dependent on microtubule occupancy and tension, respectively (Taylor et al., 2001), the simplest explanation for these observations is that expression of the N-APC mutants weakens kinetochore-microtubule interactions, thus reducing inter-kinetochore tension and resulting in an increased level of checkpoint proteins bound to the kinetochores. This is entirely consistent with the observations recently reported following a careful analysis of CIN cells harbouring endogenous APC mutations (Green and Kaplan, 2003).

To analyse the effect of APC mutation on chromosome stability we generated a panel of stable cell lines expressing N-APC mutants. A major limitation with the standard procedures used to generate stable human cell lines is that the site of integration is random. This makes it difficult to compare independent lines because any differences observed may be due to effects caused by different sites of integration. Therefore we chose a strategy that utilises sequence specific recombination to integrate expression cassettes into a pre-defined site in the genome thus facilitating the generation of isogenic, polyclonal cell lines (O'Gorman et al., 1991). We used the HCT-116 line in this study because these cells are derived from colonic epithelium, have two wild-type APC genes, a near diploid karyotype, a functional mitotic checkpoint and do not display chromosome instability (Abdel-Rahman et al., 2001; Cahill et al., 1998; Lengauer et al., 1997; Rowan et al., 2000; Tighe et al., 2001). Note however that while not CIN, HCT-116 cells have a mutation in hMLH1 resulting in a mismatch repair defect (Liu et al., 1995). Thus they do have an elevated mutation rate. Is it possible therefore that the phenotypes observed here are due to random secondary mutations? We suspect not for three reasons. Firstly, the strategy employed allowed us to pool stably transfected cells and thus generate polyclonal lines. Secondly, three independent polyclonal N750 lines were created and all three showed a similar phenotype (data not shown). And finally, similar results were also obtained with a matched pair of clonal cell lines generated following standard procedures (Fig. 7).

In this experimental system at least, expression of an N-APC mutant can induce chromosome instability in a dominant manner. Do these observations provide any insight into the role APC mutation plays in the evolution of human colon cancers? Significantly, almost all colon cancers exhibit genetic instability. Approximately 85% are aneuploid because of an underlying chromosome instability (Lengauer et al., 1998). While the remaining 15% tend to segregate their chromosomes accurately, they do exhibit another form of genetic instability, namely minisatellite instability (MIN). While its clear that mismatch repair defects are responsible for MIN, the mutations which cause CIN remain elusive (Jallepalli and Lengauer, 2001). However, we reasoned that if we could genetically manipulate a MIN cell line to behave more like a CIN line, we might gain an insight into the mechanisms underlying CIN. Indeed, targeted mutation of either MAD2 or SECURIN, or the expression of dominant Bub1 mutants disrupts chromosome stability in HCT-116 cells (Cahill et al., 1998; Jallepalli et al., 2001; Michel et al., 2001). Although mutations and/or overexpression of these genes has been described, these events are rare in colon cancer cells (Tighe et al., 2001). In contrast, APC mutations are frequent in colon cancer and, like the acquisition of CIN, they occur very early during tumour evolution (Li et al., 1997; Powell et al., 1992; Shih et al., 2001). Significantly, we show here that expression of an APC mutant in a MIN cell line induces a number of phenotypes that are more typical of a CIN line, consistent with the notion that APC mutation does initiate chromosome instability (Fodde et al., 2001; Kaplan et al., 2001).

Expression of N750 reduced the accumulation of cells in mitosis following spindle destruction (Fig. 2). While this may be due in part to reduced proliferation (Fig. 3A), it is notable that cells expressing N750 also exited mitosis faster when released from a nocodazole block (Fig. 4A). Significantly, these cells also exited mitosis in the continued presence of spindle damage. Together, these observations suggest that the N750 mutant can compromise the spindle checkpoint. Although it is unclear how APC dysfunction might disrupt mitotic control, there are several possibilities. APC localises to kinetochores and, in vitro, APC interacts with and is phosphorylated by Bub1 and BubR1 (Kaplan et al., 2001). While the significance of this is not yet clear, it is conceivable that APC sustains or amplifies the Bub1/BubR1 checkpoint signal thus contributing to the fidelity of chromosome segregation. Alternatively, APC might be required to stabilise kinetochore-microtubule interactions (Green and Kaplan, 2003; Kaplan et al., 2001). Consistent with this notion, and the report by Green and Kaplan (Green and Kaplan, 2003), we have shown that mitotic spindles are shorter in cells expressing N-APC fragments (Fig. 9), that tension across metaphase kinetochores is reduced and that levels of kinetochore-bound Bub1 and BubR1 are increased (Fig. 10). Thus, by weakening kinetochore-microtubule interactions, N-APC mutations may directly increase the frequency of chromosome missegregation.

APC and N-APC mutants can also localise to the centrosome ((Louie et al., 2004; Tighe et al., 2001) (see Fig. 8), a key component of the chromosome segregation machinery. Interestingly, centrosome defects occur concurrently with chromosome instability in pre-invasive breast carcinomas (Pihan et al., 2003). In addition, mono- and multicentric spindles and supernumerary centrosomes have been observed in APC-deficient ES cells (Fodde et al., 2001), while spindle defects have been observed in Xenopus extracts depleted of APC (Dikovskaya et al., 2004). APC mutation may therefore disrupt centrosome and/or microtubule function, resulting in spindle assembly and/or positioning defects, which in turn may cause mitosis and/or cytokinesis to completely fail. Indeed, APC-deficient ES cells are near tetraploid and the APC-defective DLD-1 line frequently becomes tetraploid (Fodde et al., 2001). Significantly, the endogenous N-APC mutant in DLD-1 cells localises to a region between the centrosomes during interphase (Fig. 8A) raising the possibility that this mutant may occasionally inhibit centrosome separation thus leading to tetraploidy. Further data supporting the centrosome connection comes from a recent report showing that GSK3β also localises to centrosomes in mitosis and significantly, small molecule GSK-3β inhibitors disrupt normal spindle positioning in human cells (Wakefield et al., 2003).

Clearly therefore, APC mutation affects a number of mitotic parameters including kinetochore-microtubule interactions, spindle assembly and morphology and centrosome function. At present one can only speculate how APC mutation might induce these defects and which of these defects is the most significant in terms of promoting CIN. In addition, in the absence of a defined mechanism, it is also not clear whether these defects arise because APC plays multiple roles in regulating cytoskeletal events, or whether they are all symptoms of an, as yet, undefined single cellular lesion.

Expression of N750 alone did not appear to induce chromosome instability (Fig. 6). Indeed, chromosome rearrangements and aneuploidy only became apparent following recovery from prolonged spindle damage. At present we can only speculate how these conditions initiated the chain of events that resulted in CIN. One possibility is that because cells expressing N750 exited mitosis prematurely (Fig. 4), this increased their chance of bypassing the post-mitotic checkpoint (Michel et al., 2001; Taylor and McKeon, 1997), thus yielding more survivors (Fig. 5). Although continued cell cycle progression in the absence of chromosome and centrosome segregation cannot be without consequences, it is not clear how this might then create dicentric chromosomes (Figs 6 and 7). However, once chromosomes with two functional centromeres are generated, chromosome instability is inevitable (Koshland et al., 1987; McClintock, 1940). Despite the ability of N750 to induce CIN under these conditions, the relevance of this observation to tumour evolution is questionable simply because, to our knowledge, it is highly unlikely that colonic epithelial cells in vivo would encounter conditions that completely ablate the spindle for prolonged periods. However, it has been reported that MIN tumours are usually confined to the proximal segments of the colon whereas CIN tumours are most abundant in the distal colon and rectum (Lindblom, 2001), raising the possibility that the type of genetic instability exhibited reflects either the different selective pressures exerted by carcinogens, which are differentially enriched through the colon (Breivik and Gaudernack, 1999), or the differences in cellular differentiation that occur between morphologically distinct regions of the large intestine (Haigis et al., 2004). Taken together with the many years it takes for a colon cancer to develop after the acquisition of an APC mutation, it is possible therefore that in conjunction with an appropriate APC mutation, specific carcinogens, or other developmental cues, may well trigger a chain of events that initiates CIN in a manner similar to the use of nocodazole in our model system.

Although the N750 mutant compromises the spindle checkpoint, cells expressing the N1309 mutant appear to have a robust checkpoint when exposed to spindle toxins (Figs. 2). If the chromosomal instability induced by N750 is due to its effect on the checkpoint, then this observation does not appear consistent with the `two hits' correlation (Lamlum et al., 1999; Rowan et al., 2000). One might expect that if mutations near codon 1300 predispose to LOH, then the N1309 mutant should have a more dramatic effect on the checkpoint relative to the N750 mutant, yet this does not appear to be the case. While this could be due to the reduced level of expression of N1309 relative N750, it is interesting that when released from a mitotic block into medium lacking spindle toxins, cells expressing N1309 did exit mitosis faster than any other line (Fig. 4) indicating that mutations near codon 1300 can indeed accelerate progression through a normal mitosis. Whether this more subtle effect on the checkpoint can induce chromosome instability in the longer term is unclear but this observation is notable for two reasons. Firstly, although the N750 mutation induces chromosome instability, this is only manifested following total and prolonged spindle destruction, a scenario that is unlikely to occur in vivo (see above). Secondly, it has been argued that excessive instability may be deleterious, whereas a subtle defect may be more favourable during tumour evolution (Cahill et al., 1999). Alternatively, although the instability induced by shorter mutations may be advantageous, perhaps this is offset by other, less advantageous effects. Indeed, the N750 mutant induces a profound proliferation defect (Fig. 3). Thus, mutations around codon 1300 could be more advantageous, not only because they induce an optimal amount of chromosomal instability (i.e. sufficient to accelerate the loss of the second APC allele without compromising clonal expansion) but also because they do not overly compromise other APC functions. Indeed, this balancing act is entirely consistent with the recently proposed `just right signalling' model (Albuquerque et al., 2002). According to this model, truncating mutations that delete all the β-catenin binding sites (as in N750 – see Fig. 1A) are less favourable because complete deregulation of β-catenin control leads to such extensive changes in gene regulation that the risk of apoptosis is increased. In contrast, retaining some of the β-catenin binding sites (as in N1309) allows β-catenin to be partially downregulated thus providing a sufficient proliferation signal without risking cell death.

Our observations suggest that truncating APC mutations can exert oncogenic effects by initiating chromosome instability in a dominant manner. However, our observations also show that different truncating mutations have differing effects on mitotic progression. We propose therefore that by striking a balance between oncogenic and tumour suppressor effects, truncations around the mutation cluster region may provide extremely favourable conditions for the evolution of a colon cancer by combining optimum levels of genetic instability with optimum levels of aberrant proliferation signals.

The authors thank members of the Taylor lab for advice and Andrew Fry (University of Leicester) for the C-Nap1 antibody. The authors are extremely grateful to Inke Nathke (University of Dundee) for providing advice on performing APC immunoblots and APC antibodies. A.T. is supported by the Association for International Cancer Research, and V.L.J. is supported by the Biotechnology and Biological Sciences Research Council (BBSRC). S.S.T. is a Cancer Research UK Senior Research Fellow.